United States Patent: (45) Date of Patent: Aug - 22, 2017

MAIMOTOMONTANT UNUNIT HINDI HII US009741048B2
(12) United States Patent (10) Patent No.: US 9 ,741,048 B2

Bendel et al. (45) Date of Patent: Aug. 22 , 2017
(54 ) PREDICTING ECONOMIC CONDITIONS (58 ) Field of Classification Search
(71 ) Applicant: Bank of America Corporation ,
USPC . . . . . . . . . . . ... 706 /21
@ See application file for complete search history .
Charlotte , NC (US)
(56 ) References Cited
(72 )
@ Inventors : Timothy J . Bendel, Charlotte, NC U .S . PATENT DOCUMENTS
(US ); Mark V . Krein , Charlotte, NC
(US ); David N . Joffe, Charlotte , NC 5 ,444 ,819 A * 8/ 1995 Negishi ................ G06Q 40 /02
(US); Sandi Setiawan , Charlotte , NC 706 /21
(US); William Anthony Nobili, 6 ,546,378 B1 * 4 /2003 Cook ................ A61B 5 /04012
Charlotte , NC (US ); David Joa , San 382/ 159
Francisco , CA (US ) (Continued )
( 73 ) Assignee : Bank of America Corporation ,
@ OTHER PUBLICATIONS
Charlotte , NC (US )
McElroy et al ( Tucker S . McElroy and Thomas M . Trimbur.
( * ) Notice : Subject to any disclaimer, the term of this “ Continuous Time Extraction of a Nonstationary Signal with Illus
patent is extended or adjusted under 35 trations in Continuous Low -pass and Band -pass Filtering,” Finance
U . S .C . 154 (b ) by 0 days . and Economics Discussion Series, Divisions of Research & Statis
tics and Monetary Affairs, Federal Reserve Board , Washington ,
(21) Appl. No.: 15/146, 137 D . C ., Published in SSRN Journal 2007 ) .
(22) Filed: May 4, 2016 (Continued )
Primary Examiner — Kimberly Berona
(65 ) Prior Publication Data Assistant Examiner — William J Jacob
US 2016 /0260114 A1 Sep . 8 , 2016 (74 ) Attorney , Agent, or Firm — Banner & Witcoff, Ltd .;
Michael A . Springs
Related U . S . Application Data (57 ) ABSTRACT
(63 ) Continuation - in -part of application No. 14 /078 ,210 , Computer - implemented methods for identifying or assess
filed on Nov . 12 , 2013 . ing any type of risk and /or opportunity that may arise can
include either, alone or in combination , band pass filtering ,
(51) Int. Ci. principal component analysis, random matrix theory analy
G06Q 40 /06 ( 2012 .01) sis, synchronization analysis, and early -warning detection .
G06Q 30/ 02 ( 2012 .01) Each technique can also be viewed as a process that takes a
G060 40/04 ( 2012 .01) set of inputs and converts it to a set of outputs. These outputs
(52 ) U .S . CI. can be used as inputs for a subsequent process or the outputs
may be directly actionable for formulating certain economic
CPC ... G06Q 30/0205 (2013.01); G06Q 30 /0201 predictions to make certain decisions.
(2013.01); G06Q 40/04 ( 2013 .01) ; GO6Q
40/06 (2013.01) 23 Claims, 40 Drawing Sheets
2000 Take Data of Multiple Time Series of
Start Interest
2002 Build the Standard Pearson Correlation

Coefficients Using the Data
Build the Matrix Whose Elements are

2004 the Correlation Coefficient Values (Call the
Matrix CCM or Correlation
Coefficient Matrix )
2006
Build NearestNeighbor Spacing
Distribution by Generating it from the
Differences Between Nearest Neighbor
Eigen - Values Derived from CCM
2008 Set the Lowest Value of Correlation

Coefficients Iteratively to Zero , and Repeat
Using Chi -Square Test, a Sharp Phase

2010 Transition from Gaussian Orthogonal
Ensemble (RMT ) Distribution to
Poisson Distribution Would be
Observed at a Certain " Threshold " Level
2012 Process Ends Once Phase Transition
Takes Place
US 9 , 741, 048 B2
Page 2
( 56 ) References Cited M . Breakspear, S . Heitmann , A . Daffertshofer. “Generative models

of cortical oscillations : neurobiological implications of the
U .S . PATENT DOCUMENTS Kuramoto model” . Frontiers in Human Neuroscience 4 , 1 ( 2010 ).
Publicly available codes and references therein .
7 ,660 ,737 B1 2 /2010 Lim et al. R . M . May , S . A . Levin , G . Sugihara . “ Ecology for bankers” . Nature
8 , 364, 569 B1 1/ 2013 Lee , Jr. 451, 893 ( 2008 ).
8 ,606 ,620 B2 12/ 2013 Grichnik et al. A . G . Haldane, R . M .May. “ Systemic risk in banking ecosystems” .
2003/ 0074331 A1 * 4 / 2003 Kreft G06Q 10 /06 Nature 469, 351 ( 2011 ) .
705 /400 J. D . Farmer, D . Foley. “ The economy needs agent-based model
2004/0015376 A1 * 1/ 2004 Zhu .... GOOQ 40 / 08 ing” . Nature 460 , 685 (2009 ).
705 / 7 .28 M . Scheffer, J. Bascompte , W . A . Brock , V . Brovkin , S . R . Carpen
2005/0182673 AL 8 / 2005 Marzian et al. ter, V . Dakos , H . Held , E . H . van Nes , M . Rietkerk , G . Sughara .
2007/0130060 A1 6 / 2007 Ariarajah et al. “ Early -warning signals for critical transitions" . Nature 461, 53
2008/ 0033854 Al * 2/ 2008 Nguyen ................. G06Q 30 /02 (2009 ). And references therein .
705 /35 M . Scheffer. “ Foreseeing tipping points " . Nature 467 , 411 (2010 ).
2008/0154821 A1 * 6 / 2008 Poulin ................. G06Q 10 / 04 J. M . Drake , B . D . Griffen . “ Early warning signals of extinction in
706 /21 deteriorating environments" . Nature 467, 456 ( 2010 ) .
2009/0182620 A1 7 / 2009 Sunvold L . Dai, K . S . Korolev, J. Gore . “ Slower recovery in space before
2009 /0265281Al 10 / 2009 Cohen et al. collapse of connected populations" . Nature 496 , 355 (2013 ). And
2009/0292648 A1 11/ 2009 Damschroder et al.
2010 / 0076813
AL 3 / 2010 Ghosh et al. references therein .
2010 /0094672A1 * 4 /2010 Macy .... G06Q 10 /00 K . Borner, S . Sanyal, A . Vespignani. “ Network science ” . In Blaise
705 /7 . 12 Cronin (Ed .), Annual Review of Information Science & Techonol
2013/0103615 Al * 4 / 2013 Mun ................ G06Q 40 /06 ogy, vol. 41, Medford , NJ: Information Today , Inc./American Soci
705 / 36 R ety for Information Science and Technology , Chapter 12 , pp .
2014 / 0330645 A1 11/ 2014 Craft 537 -607 (2007) .
N . Friedman . “ Inferring cellular networks using probabilistic
OTHER PUBLICATIONS graphical models” . Science 303, 799 ( 2004).
M . Wildi. “ Real- time US-Recession Indicator (USRI): A classical
Kastrinaki et al. (An Analysis of UK Merger Cyclers and their cycle perspective with bounceback ” . Publicly available pre -print.
Synchronization with Business Cycles: A Frequency Domain Institute of Data Analysis and Process Design , University of Tech
nical Sciences , Rosentrasse 3 , 8401 Winterthur, Switzerland , email:
Approach , Mar. 2006 , University of Warwick ). marc.wildi @ zhaw .ch ; web -link to the paper: http ://www .idp .zhaw .
Oct. 18 , 2013 — U .S . Appl. No. 14 /058, 119 . ch/ fileadmin /user _ upload/ engineering/ _ Institute _ und _ Zentren /
Carol Alexander, Quantative Methods in Finance, 2008 . IDP /forschungsschwerpunkte/ FRME/sef/US_ Recession _ Indica
Margaret Meyer, Increasing Interdependence of Multivariate Dis tor/ papers/ us_ recession .pdf.
tributions, Mar. 1 , 2011. M . E . J. Newman . “ Networks: An Introduction ” . Oxford University
V . Plerou , P . Gopikrishnan , B . Rosenow , L . A . N . Amaral, T. Guhr, Press ( 2010 ) . And references therein .
H . E . Stanley. “ Random matrix approach to cross correlations in M . O . Jackson . “ Social and Economic Networks ” . Princeton Uni
financial data” . Physics Review E 65, 066126 ( 2002). And refer versity Press (2008 ). And references therein .
ences therein . S . Wasserman , K . Faust. “ Social Network Analysis: Methods and
F. Luo , J. Zhong, Y. Yang, J. Zhou . “ Application of random matrix Applications” . Cambridge University Press ( 1994 ). And references
theory to microarray data for discovering functional gene modules” . therein .
Physical Review E 73 , 031924 ( 2006 ). And references therein . Taylor, R ., Interpretation of the Correlation Coefficient: A Basic
S . Bornholdt. “ Less is more in modeling large genetic networks” . Review , Jan ./ Feb . 1990 .
Science 310 , 449 ( 2005). Saad , Y., NumericalMethods for Large Eigenvalue Problems, 2011.
Z . N . Oltvai, A .-L . Barabasi. “ Life 's complexity pyramid ” . Science Plerou , V ., A Random Matrix Approach to Cross -Correlations in
298 , 763 ( 2002 ). Financial Data .
M . E. J. Newman. “ The physics of networks” . Physics Today, Nov . Benes, J., Vavra D ., Eigenvalue Filtering in VAR Models with
2008 . Application to the Czech Business Cycle , Nov. 2005.
S . H . Strogatz. “ Exploring complex networks” . Nature 410 , 268 Bourgine et al., Cognitive Economics: An Interdisciplinary
(2001 ) . Approach , p . 391, 2004.
A . E .Molter, R . Albert. “ Networks in motion ”. Physics Today, Apr.
2012 . * cited by examiner
U . S . Patent Aug. 22 , 2017 Sheet 1 of 40 US 9, 741, 048 B2
- 109B
107 109A????
109A
0 / 109c
**
*
109D
BALA? %d8
105 ( WAN 101 - 109E
LAN (Opt.)| - 109F
Processor Network Interface

RAM || ROM
RAM ROM || Input/Output
113 / 15 . 119 - 121
Memory
Operating System . Pr123
Control Logic -125
Other Applications L - 127
131 - 129
DB1 | DB2
FIG . 1
U . S . Patent Aug. 22, 2017 Sheet 2 of 40 US 9 ,741,048 B2
Technique Technique
Inputs Outputs
Process Process
Technique Additional
Outputs Inputs
FIG . 2
Top
Bottoms Down
Outputs
Outputs
Other
Outputs
Decision
.. ..
FIG . 3
con cor --- - -
Macro -Micro Complexity Pyramid
General
Large -Scale
Organization
Level 4
Specificity Functional Universality
Modules
Level 3
More Specific
Medium - Sized Network
Level 2
Single Agent to Small - Sized
Agent Circuit or Network
Level 1
Information Database - > Processing - > Execution
FIG . 4
502 504 506 508
Tansforma Apply
Sen tequency band pass Soal
Hilter
FIG . 5
*
*
9
1
-
wa '
44
*
:
*
*
*
*
*
.
.
.
+ - 1. tin vit viintritt vin in v . . . . . ... .. . .+ 1. . . -. + + + + .' . . . " . . .1 . .
FIG . 6A FIG . 6C
*
**
wa rangi yang telah melen
wwwwwwwwwwwwwwwwwwwww
** * * * * * * * ** *
. . ..
* ** **
... ..
* ** * * * * * *
. .. .
** * * * * * * * * * ** * * *
FIG . 6B
0 . 10
0
-0 . 10
1995
1995 2000
2000 2005
2005 2010
2010
FIG . 7
802 804 806 808
Apply Take Find

Time w BandopValue
- Absoluteas ? Correlation Correlation /
SeriesSEPass Volatility of All Concentration !
Data Filter Series Measure
Measure
FIG . 8A
Mommy
CMoerdliaton 00M
m
m
W W wa
-0.2 [ ?
2000
2000 2005
2005 2010
2010
FIG . 8B
ó - - - - -- - - - - - - - - - - - - - - - - - - - - - - - 1
- - - - -
Subcomponent 8
ó-
o Subcomponent 1 w
-- - -- - - I - - - - - - - - -- - - - - - - -
w Subcomponent 3
Subcomponent 7
2-NonSCoymsptoenmict - - www .
ww
ww
= == -E --
www .. . . . . . . . ... . .
Subcomponent 5
wwwwwwwwwwwwww w wwwwww www
Subcomponent 2
is - - - - -
w - - - - - -
w Subcomponent 4
- - - - - - - - -
" Subcomponent 6
- - - - - - - - - - - - - - - - - - - - - - - - -
? - - - -
w
0 .0
- - -
w
0 .2
- - -
0.4 0 .6
w
0 .8 1.0
- - - -
Component 1 - Systemic
FIG . 9
2012
Farm -
- - Credcard
Agri Lease
- -
T 2011
2010
2009
2008
2007
VA 2006
ConstherComre
T
2005
Resi .
-
Bus :
2004
- 2003
?
?
10
.
FIG
.
2002
.
-.
2001
2000
bySUnecoltarblse
)
Scaled
All
(
*
. 6661
1998
CUnocmpletabs 1997
1996
1995
1994
?? ? ? ? ? ?? ??
T1993
1992
1991
04. 2010
EV2EV1
EV4
0
DEV3 922
- 01. 2010
02. 2009
– 00 .2008
04. 2007
EV8DEV7 EVO ØEV5 .
01. 2007
NØ
02.2006
00. 2005
– 04.2004
01. 2004
02. 2003 11
.
FIG
.
O
00. 2002
E
Q
23iugea
bySonUPnecolCtcoaArblse MCoartel ioxn Windownrv t
l e
u s TA
PC1
04.2001
– 01.2001
- 02.2000
00. 1999
04. 1998
01. 1998
02. 1997
%
%
100 % 80
90 % 70
% 60
% % 40
50 % 30
% % 10
20 % 0
Sin' ju?zed ny&· 77 ‘ LIOT v??YS OT JO + 0 ?? ??L°6 *8?0 78
017.201104502..2011
Farm LeaseCredcar
R:ET TT03. 2010 2010
01.2010 02.2010
03.2009 04.2009
017.2009 02
03. 2008
.2009
04 .2008
Agri
-
*
- 017.2008 02.2008
- -
- 03.2007 04.2007
-
- 01.2007 02.2007
03. 2006. 04. 2006
01. 2006029 . 2006
L1FCPoramicpndtoegnarsl ConstherComre
Resi Bus
-
*
01..2004
01.2004
) 03
01 2003
04. 2005
) 03.2005 02.2005
) 03 2005 04.2004
02.2004
404 04.2003
03..2003 02..2003
02 2003
12
.
FIG
.2002 04.2002
- 01
01..2002 02..2001
2002 04 2002
- 01.2001 02.2001
03.2000 04.2000
01.2000 02.2000
03.1999 04.1999
07
01..1999
1909 02
*04..1999
i
1998
) 03.1998 02. 1998
01. 1998 04.1997
03. 1997
01. 1997 02 .1997
034.1996 04.1996
1
-
20
.
0
15
.
0 10
.
0 05
.
0
000
2000200120022003200420052006200720082009201020112012
bySofUDPnicesolpCtrosa1iblen CUoDnicmsplerftanibos T 13
.
FIG
-
015
.
0 -
005
.
0
025 0
.
0 020
. 010
.
0
.
other
-
comes
comment
come
.
.
HPIbyStates VirginiaP-.NewYorkNevadaMHawaiiCeArizona|aonssnaycelhcuvtsaienctiuast WCaslhifnogrtnoina

Island
Rhode
Ohio
Hampshire
New
Michigan
.
Illinois
Florida
- 199519961997199819992000200120022003200420052006200720082009201020112012 14
.
FIG
T :
Tennessee
Oregon
Jersey
New
–
Minnesota
Maryland
.
Georgia
Colorado
-
|
A TIÓ
2011/31/ 10
.. . . . . . . . . . . . . . . .. . . . . . . .... . . . . . . . . . . . .
2010/31 / 10
2009/31/ 10
2008/31/ 10
- 2007 /31/ 10
mm
OEV3BEV127593
W70MECioagrentvldoiuhewxsn EVOZEV14DEV222EV2NEV18OEV10
NEDEV1EVOEEV5AV12713
EV12OEV80EV4EV20DEV16 www
2006/31/ 10
2005/31/ 10
15
.
FIG
2004/31/ 10
2
:
11
- 2003/31/ 10
PC1
2002/31/ 10
PCAonHPLbyStates
2001/31/ 10
TTTTTT 2000/31/ 10
% 90
100 % %
80 70 % 50
% 60 % 40
% %
30 %
20 10
% 0
%
2011/31/10
2010/31/ 10
2009/31/ 10
2008/31/ 10
2007/31/ 10
ofHPIbySCPC1Eiotgneavtnceosr L1FCPoramipncdteognarsl 2006/31/ 10

16
.
FIG
2005/31/ 10
2004 /31/ 10
. 2003/31/10
-
-
-
/7
- 2002/31/ 10
2001/31/ 10
- 2000/31/ 10
07
.
0 06
.
0 05
.
0 04
.
0 03
.
0 0
.
02 01
.
0 00
.
0
200720082009201020112012
HPIbySDofitspaetrseiosn bySIPofHDitnrsopadiemrtecsixon 17
.
FIG
200120022003200420052006
-
0030
.
0
0025 0
.
0 0020
. 0015
.
0 0010 0
.
0 0005
.
( Start
1800
PCA Applied to Time
Series Data
1802
Normalize Eigenvector
Content
1804
as an engineering
Sort Eigenvectors -
Smallest to Largest
1806 )
Calculate Distances
Between Eigenvectors
1808
Calculate Distances
Between Eigenvectors
1810 )
Predict Contagion
Based on
Dispersion Level FIG . 18
1902 1916 1918

Publicly Available Apply Principal
Macro Economic Component List of Driver
Analysis to Find Segments
Time Series Data Drivers
1904 ) 1914 ) 1920 )

Apply Band Pass Concentration Proprietary Micro
Filtering to Status Level Data
Individual Series
1906 1912 1922
Determine Portfolio
Filtered Series Correlation / Optimizer
Concentration
1908 1910 ) 1924
Determine Cycle
State (Equilibrium , Portfolio Outputs
Positive Trend, State Status
Negative Trend )
FIG . 19
Start 2004 TakeData of Multiele TimeSeries of

Start
2000 Take Data of Multiple Time Series of
Interest
2002 Build the Standard Pearson Correlation

on
Build Nearest Neighbor Spacing

2006 Distribution by Generating it from the
Differences Between NearestNeighbor
Eigen -Values Derived from CCM
2008 Set the Lowest Value of Correlation
Coefficients Iteratively to Zero , and Repeat
Using Chi- Square Test, a Sharp Phase

2010 Transition from Gaussian Orthogonal
Ensemble (RMT) Distribution to
Poisson Distribution Would be
Observed at a Certain " Threshold " Level
2012 Process Ends Once Phase Transition

Takes Place
FIG . 20A
Stan 2000, Take

Start
2020
2022
Take Data of Multiple Time Series of
Interest
Build the Standard Pearson Correlation
2026 Build Relevant RMT Where the Density

Distribution of Eigen - Values for Such
Random Matrices is known
2028 Compare the Statistics of the Observed

Eigen - Values, i. e . the Eigen -Values of the
Data Correlation Matrix to Those of the
RMTCounterpart
Number of Observed Eigen -Values Larger
2030 than the Analytical Maximum from RMT
Provides an Estimate of the Number of
Significant Eigen - Value Components
of the Data
2032 Observed Eigen - Values of the Data
Correlation Matrix Larger than the
TheoreticalMaximum from RMT Provides
a Reasonable Approximation to the
Number ofPrincipal Components to Retain
FIG . 20B
O
Site
Senosies www
???
?
??
???
300
???
30
PInshta eous w
www ???
??
???
????
.
-
90
1 - --
D
120
Oise
60 ?? ?? ??? ??? ??? ??? ? ? ?? ?? ???
-- at 270
ww
w when
-
-
-* O
Months
0
=
t
**
www
oo
Cocorota w
w
*
**
**
240
150 sapiens
Phase SDistrbu on Ehnatropny -pOardmetr I

21A
.
FIG
-
9
.
0
98356
.
0
=
S 086128
.
0
=
r
77
.
0 -
6
.
0
ö
-
5
.
0 -
4
.
0
3
.
0
0
2
.
-
1
.
0 obmh FPInrshetaqusnocey
3
-
2
15
1
05
.
0
FrequncDiystrbuion INX
WA
+
0
0
27
.
0 0187
. 16
.
0 14
.
0 12
.
0 08
.
0 06
.
0 04
.
0 02
.
0
Count
330
fo oooooo
XX
PInshta eous 1
an60
90
a
- - -
--
-
-
w www
*w
*
-
--
ty
-
-
www
100DE
300
*
270
Months
5
.
12
=
t
wwww w
-
- *
-- -
-- -
-
-- we w
*
120 240
wSoos TEAU
w
150 Shitesterol
Phase SDistrbu on Ehnatropny -pOardmetr

89163 r
.
0
=
S 50903
.
0
=
? ah PInshta eous
T
0
1
2
3
-
21B
.
FIG
ö 07
IIIIIII
-
9
.
0 -
6
.
0
8
.
0 7
.
0 5
.
0 4
.
0 3
.
0
FrequncDyistrbuon 2
15
1
05
.
0
0
Frequncy
0
2
.
0
18
.
0 16
.
0 14
.
0 12
.
0 08
.
0 06
.
0 04
.
0 02
.
0
Count
??
?
??
?
sro
w w w
??
???
60 ***
?
???
?
-
PInshta eous
**
--
** ??
? --
--
1 *
*
?
- --
www --
90 - - *
*
- --
--
--
R
de300240Leio270
*
*
*
- *
w
Months
20
=
t
-- *
*
* w
-- w
120
150
w
&
w
210
An
Phase EDistrbu on Snhtaronpy -pOardmetr

55355 r
.
0
=
S 88196
.
0
=
-
. . .... . .. . . ... .. .... .... . .. Wwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwww
** * **
W
wwW
w
0
1
2
3
-
PInshta eous 21C
.
FIG
O 8
.
0
-
7
.
0
6
.
0 5
.
0
-
4
.
0
3
.
0 2
.
0 17
.
0
ö 2
.
0
FrequncDyistrbuon 1 !! ! ! ! !
. * . .
ER
. * .* .
* 21
*
* .. . . .
ER
. * . .* *
. , LES
-
15
1
05
.
0
0
Frequncy
2
.
0
18
.
0 16
.
0
-
14
.
0
12
.
0
14
.
0 08
.
0
0
'
90
+
04
.
0 02
.
0
to
Count
ww pitamentoeft 330
PInshta seou 1
60
90
*
-
** *
51
.
0
]-
??? ??? ??? ? ?? ??? ??? ?? ??
*
*
*
??? ?? ??? ?? ? ??? ??? - -

ww w w
.
.
-
.
-
- - “ ww -
300
270
-
--
Months
30
=
t
120
.-
--
-
-
www w .www
WW '
240
-
150
/
150
180
210
Distrbuon
Phase SEhnatropny pO-ardmetr
21485
=
S
9
.
0 99388
=
r
+
8
.
0
TTTTTTT
7
.
0
-
6
.
0 -
5
.
0 -
4
.
0
3
.
0
-
2
.
0 1
.
0
ö
T PInshta eous
T
0
-
0
1
2
3
-
21D
.
FIG
2
.
0
FrequncyDistrbuon |
0
+
15
1
05
.
0
0
Frequncy
27
.
0 18
.
0 16
.
0
-
14
.
0 -
12
.
0
0
.
1
-
08
.
0 06
.
0
-
04
.
0
02
.
0
Count
20 330
60
ca
-
-
**
*
*
*
**
-
w the
--
300
PInshta eous wwwwhat www

ww -
1
1 -
-
-
w -
90
ao -
-
270
-
--
??
???
??
???
?
?
Months
40
=
t
??
- ??
120
120 ??
???
??
??
???
??
240
240
/
150 ???
???
150 ???
210
180
Phase EDistrbu on Snhtaronpy -pOardmetr

18
=
0
22085
. 99669
.
0
=
r
0
1
2
3
-
PInshta eous 21E
.
FIG
9
.
0
-
7
.
0
0
.
6
-
5
.
0 -
4
.
0 -
3
.
0
2
.
0 1
.
0 ot
FrequncDyistrbuon . . .
2 ::*:
1
2
15
1
05
.
0
0
Frequncy
2
.
0
18
.
0 16
.
0 14
.
0 12 0
0
. 17
. 08
.
0
-
06
.
0
04
.
0 02
.
0
Count
330
300
w -
PInshta eous
* --
901d
*
* -
- - -
--
--
w
- - - - - - .
-
-
-- 270
*
-
Months
1
.
60
=
t
-
!
-
w wers
120 wewhen **
**
*
240
/
150
150
180 210
PInshta eous
Phase EDistrbu on Snhtaropyn
FrequncDyistrbuon
951
13335
.
0
=
.
0
parameter
-
Order
8
.
0
DIE
r
99674
.
0
=
. :: , 13
7
.
0
|
6
.
0
RW
.. . . . ..... ... . . . . .. . . . ..
-
5
.
0
.
4
.
0
.. ..
AV111 :
: :
.
3
.
0
. . 1'
-
2
.
0
1
.
0
OT Frequncy
0
1
2
3
-
2
.
0
15
1
05
.
0
21F
.
FIG
0
0
+
-
2
.
0
18
.
0 16
.
0 14
.
0 12
.
0
800 06
.
0
-
04
.
0
02
.
0
Count
o
330
wwwww
PInshta eous
90
w
60
1
w w w **
**
**
www1
-
w we
wwww -
--
300
x
270
Months
80
=
t
www
--
-
--
om my
w
120 240
150
180 210
Phase Distrbuon 0.8Order-parameter

S
E h
n a
t r n
o p y
11695 99674
.
0
[
=
S .
0
=r
0
1
2
3
-
PInshta eous 216
.
FIG
-
2
.
0
7
.
0
-
6
.
0
5
.
0 4
.
0 3
.
0 1
.
0
2
.
0
FrequncDyistrbuon Women
15
1
05
.
0
RE
BAR
0
Frequncy
21
.
0 18
.
0 16
.
0 14
.
0 het?
Count
–
04
.
0
F20
Logo
-
-
-
330
-
- -
300
-
60 -
7 *
-
r
sout
PInshta seou
t
*
* to
1 - *made
-
**
- *
-
*Www
--
- www
90 - - - - -
-
- 1
-
270
-
- *
*
***
- -
*
1
--
Months
100
=
t
-- www
-
120 240
150 210
1180
Distrbuon
Phase ESnhtaropyn pO-ardmetr2251 r
.
0
=
S 99674
.
0
=
0
1
2
3
-
PInshta eous 21H
.
FIG
TTTTTTTT +
0
9
.
0 8
.
0
-
7
.
0 -
6
.
0 -
5
.
0 -
4
.
0
3
.
0
-
2
.
0
ö ö o1
.
0
2
.
0
FrequncDyistrbuon +
0
15
1
05
.
0
0
Frequncy
uency
2
.
0
18
.
0 16
.
0 14
.
0 0
.
12 0
.
17 08
.
0 06
.
0 04
.
0 02
.
0
Count
100
r-PofOEavroldmuetiorn
0
.
1
-
9
.
0 -
8
.
0 -
7
.
0
-
6
.
0 5
.
0
-
4
.
0 3
.
0 2
.
0 1
.
0
T
O
90
80
70
60
50
40
30
20
10
0
) FIG
Sec
(
Time 22A
.
< . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
100
TRAFIWDM8vI1PU9RWY1300DIDIODOS
WAIIWIWIWILI ILE- WELLEM 2 22 :1 3 : : .. .. . : :
90
takutunz
80
70
i ata121129274 . 60
DLyoncakmings
. .. . .. .. .. . .. ... .... .... ... .. . . .. ... .. . -
RESUR EASEDUSERTA
-
ww
2011
- - . . .
+ + + + . . . . . . . . wwwwwwww
. . ..
- - - - -- -
---
--
T
50
40
30
20
20
Time 22B
.
FIG
++ * * * **
W! i
10
other or es
-
350
*
00€
*
250
*
100 .o
100
90
IPHnrodimecx-CARAorueltaio-SDnofARmeroivtahievd-CAorueltaion 80
70
60
Charlote 50
40
)Time(Months 23A
.
FIG
30
20
10
-
nä ä ä0 o0 ác '
eni
Values Scaled
100
ofARDSemroivtahievd 90
IPHnrodimecx-CARAorueltaion -CAorueltaion 80
70
60
50 Months
(
Time
) 23B
.
FIG
40
30
LasVegas 20
T
10
~ - 0 _ Ñ
Values Scaled
> 100
ofARDSemroivtahievd - - - - -
IPHnrodimecx -CAorueltaion
T
Cor elation
Auto
AR
-
90
80
70
60
50
40
30
20
10 Months
(
Time
) 23C
.
FIG
Clev and
LLLLLLLL LL
nä ? - oö · ? ? ?
Values Scaled
100
ofARDSemroivtahievd 90
IPHnrodimecx-ARAutoCorelation -CAorueltaion 80
70
60
50 TimeM()onths 23D
.
FIG
40
30
Phoenix 20
10
o Non
Values Scaled
( Start
2400
Take Data of Time Series of
Interest
2402
Calculate AR ( 1 ) Model Coefficients of the
Time Series (AR Stands for Auto
Regressive )
2404 Compute the Scaled LAG - 1

Auto -Correlation of the Coefficients
Calculate the Scaled Smoothed Derivative
2406 from the Output of the Previous Step (Here
the Resulting Derivative is Used as the
Early Warning Detector)
FIG . 24
INFORMATION
SOURCE ON 2530
2530
2500
2500
2501 2502
DATA PROCESSOR EARLY WARNING
SYSTEM
COMPUTING COMPUTING COMPUTING

DEVICE DEVICE DEVICE
2520 2521
2522
FIGURE 25
INFORMATION
SOURCE 2630
2630
2600
2602
DATA SERVER COMMUNCATION

SERVER
2601
www
2610 LOCAL COMMUNICATION

NETWORK
EXTERNAL
COMMUNICATION 2611
NETWORK
PRIMARY www
COMPUTING
DEVICE
WEB SERVER onwwwwwww

2622 REMOTE
2620 COMPUTING
DEVICE
REMOTE
COMPUTING
DEVICE
2621
2603
FIGURE 26
U . S . Patent Aug. 22 , 2017 Sheet 37 of 40 US 9 ,741,048 B2
2700
2701
DETECT INSTABILTY *NO CONTINUE MONITORING /
EVENT ANALYZING DATA
YES 2702
DETERMINE ASSIGNED
REGISTERED DEVICES (E .G .
ASSOCIATED WITH 2703
GEOGRAPHICAL REGION )
2704
ANY REGISTERED DEVICES L A LALALALALALALALALALALALALALALALA
ACTIVE
YES
.
LUND
GENERATE EARLYWARNING GENERATE EARLY WARNING
?
NOTIFICATION TO PRIMARY 2705 NOTIFICATION TO ALL
DEVICE REGISTERED DEVICES
2707
00
.
0
augod
CONSUJAZIL 2706
2708
RECEIVE
ACKNOWLEDGMENT
LAUNCH EARLY WARNING
Om FROM ANY REGISTERED
APPLICATION AT PRIMARY DEVICES
DEVICE
PERFORM ALTERNATIVE
2710 EARLY WARNING UBon odr
NOTIFICATION YES
ao -2709
LAUNCH EARLYWARNING
APPLICATION AND MODIFY /
RESIZE /MINIMIZE CURRENTLY
DISPLAYED FRAMED
FIGURE 27 GRAPHICAL ELEMENTS AT
RESPONDING REMOTE
DEVICE
2800
2801
DETECT INSTABILTY NO CONTINUEMONITORING /
EVENT ANALYZING DATA
YES 2802
DETERMINE ASSOCIATED
REGISTERED DEVICES (E .G .
ASSOCIATED WITH 2803
GEOGRAPHICAL REGION )
GENERATE EARLY WARNING 2804

NOTIFICATION TO
REGISTERED PRIMARY
COMPUTNG DEVICE
2805
RECEIVE
YE ACKNOWLEDGEMENT www.NO
FROM PRIMARY DEVICE
GENERATE EARLY WARNING
NOTIFICATION TO ALL
2807 REGISTERED REMOTE
LAUNCH EARLY WARNING DEVICES
APPLICATION AT PRIMARY 2806
DEVICE
2808
RECEIVE
ACKNOWLEDGMENT
FROM ANY REGISTERED
REMOTE DEVICES
2810 PERFORM ALTERNATIVE

EARLY WARNING
NOTIFICATION ---2809
LAUNCH EARLY WARNING

APPLICATION AT PRIMARY
COMPUTING DEVICE AND REMOTELY
CONNECT FROM RESPONDING
FIGURE 28 REMOTE DEVICE TO PRIMARY DEVICE
GENERATE EARLY WARNING

2901 v NOTIFICATION TO ALL 2900
REGISTERED DEVICES
2902
RECEIVE
ACKNOWLEDGMENT XXXN
FROM FIRST DEVICE
YES
PROVIDE EARLY WARNING

INFORMATION TO THE FIRST
DEVICE AND SEND INVITE PERFORM ALTERNATIVE
2903 MESSAGE TO ANOTHER EARLY WARNING
DEVICE (E .G ., SECOND NOTIFICATION
DEVICE ) TO COLLABORATE
WITH USER OF THE FIRST
DEVICE
2904
FIGURE 29
3000
3001 RECEIVE ACKNOWLEDEMENT

FROM RESPONSIVE DEVICE
3002
DESKTOP FYYYYYYYYYYYYYYYYYYYYYYYYYYYYY
IDENTIFY DEVICE TYPE YYYYYYYYYYYYYYYYYYYYYYYYYYYYYYY
- TABLET
3003 SMOTOn 3004

3004 3005
SMARTPHONE
OBTAIN FIRST SET OF DISPLAY OBTAIN SECOND SET OF OBTAIN FIRST SET OF DISPLAY
CHARACTERISTICS DISPLAY CHARACTERISTICS CHARACTERISTICS
wwwww Wwwwwwwwwwwwwwwvvwvvwvvwvyuran
DISPLAY SUBSET OF FIGURE 30

3006 DISPLAYED COMPONENTS
FOR DEVICE BASED ON
DISPLAY CHARACTERISTICS
US 9,741,048 B2
PREDICTING ECONOMIC CONDITIONS the following description in consideration of the accompa
nying drawings, in which like reference numbers indicate
This application is a continuation - in -part of co - pending like features, and wherein :
U .S . patent application Ser. No. 14 /078 ,210 entitled “ Pre - FIG . 1 illustrates one example of a network architecture
dicting Economic Conditions” and filed on Nov . 12 , 2013, 5 and data processing device thatmay be used to implement
the entire disclosure of which is hereby incorporated by one or more illustrative aspects discussed herein .
reference . FIG . 2 illustrates a flow diagram for an exemplary process
disclosed herein .
BACKGROUND 10
FIG . 3 illustrates a flow diagram for another exemplary
process disclosed herein .
Modern global finance depends on a high level of con FIG . 4 illustrates an exemplary diagram relating to one or
nectivity among financial institutions. In stable market con - more aspects of the disclosure herein .
ditions such connections allow capital to flow freely with FIG . 5 illustrates a flow diagram for another exemplary
little regard to geography. Transactions among investors / 15 process disclosed herein .
savers and liquidity providers and liquidity users can be FIGS. 6A -6C illustrate an exemplary data filtering tech
globally efficient. The same networked structure may , how - nique disclosed herein .
ever, become a channel of economic instability and distress FIG . 7 illustrates a graph relating to an exemplary process
amplification during global, financial stress. In the case of of FIG . 5.
banks, where savers (deposits ) fund investments (loans ), a 20 FIG . 8A illustrates a flow diagram for another exemplary
devaluation of loans can create abnormal liquidity demands filtering process disclosed herein .
from deposits . In the same way , in the case of shadow banks, FIG . 8B illustrates an exemplary graph relating to the
where savers (money markets) fund investments (equity, exemplary process of FIG . 8A .
debt and derivatives ), a devaluation of collateral can also FIG . 9 illustrates another exemplary graph relating to
create an abnormal liquidity demand . These abnormal 25 principle component analysis .
liquidity demands can escalate due to the interconnectedness FIG . 10 illustrates another exemplary graph relating to
of the economic system . principle component analysis .
Studying systemic risk as compared with the emphasis on FIG . 11 illustrates another exemplary graph relating to
developing conventional risk management techniques in principle component analysis .
individual entities may be helpful in understanding eco - 30 FIG . 12 illustrates another exemplary graph relating to
nomic conditions . The losses resulting from systemic risk principle component analysis .
may be taken into account. Focusing on the individual firm FIG . 13 illustrates another exemplary graph relating to
level in risk management may not be enough in managing principle component analysis.
the risk of a complex, interconnected system of companies. 26 FIG . 14 illustrates another exemplary graph relating to
A larger, systemic perspective may be desired . Therefore , principle component analysis .
micro -macro connections can be studied , though individual FIG . 15 illustrates another exemplary graph relating to
components unique to a given microeconomic system . The principle component analysis .
topological properties of the networks in macroeconomic FIG . 16 illustrates another exemplary graph relating to
system share similarities with universal organizing prin - 40 principle component analysis .
ciples . FIG . 17 illustrates another exemplary graph relating to
principle component analysis .
BRIEF SUMMARY FIG . 18 illustrates an exemplary flow chart pertaining to
principle component analysis .
The following presents a simplified summary of various 45 FIG . 19 illustrates a flow chart pertaining to an exemplary
aspects described herein . This summary is not an extensive data filtering technique and principle component analysis .
overview , and is not intended to identify key or critical FIG . 20A illustrates a flow chart pertaining to an exem
elements or to delineate the scope of the claims. The plary random matrix theory analysis .
following summary merely presents some concepts in a FIG . 20B illustrates another flow chart pertaining to an
simplified form as an introductory prelude to the more 50 exemplary random matrix theory analysis .
detailed description provided below . FIGS. 21A - 21H show the evolution of exemplary eco
Computer- implemented methods for identifying or nomic data of an exemplary system over time.
assessing any type of risk and/ or opportunity that may arise FIG . 22A shows the evolution of an exemplary order
can include either, alone or in combination , band pass parameter over time.
filtering, principal component analysis , random matrix 55 FIG . 22B illustrates an exemplary phase distribution of
theory analysis , synchronization analysis, and early -warning economic data .
detection . Each technique can also be viewed as a process FIGS. 23A - 23D illustrate exemplary Home Price Index
that takes a set of inputs and converts it to a set of outputs . data and derivatives thereof.
These outputs can be used as inputs for a subsequent process FIG . 24 shows an exemplary process for early warning
or the outputs may be directly actionable for formulating 60 detection of economic conditions .
certain economic predictions to make investment decisions FIG . 25 illustrates an apparatus with a data processor and
and the like. an early warning system in accordance with an aspect of the
embodiments .
BRIEF DESCRIPTION OF THE DRAWINGS FIG . 26 illustrates a computer system that supports pro
65 cessing of information about a real-world system and gen
A more complete understanding of the present disclosure erating an early warning signal in accordance with an aspect
and the advantages thereof may be acquired by referring to of the embodiments .
US 9, 741, 048 B2
FIG . 27 illustrates a flow diagram for generating an early As shown in FIG . 1 , devices 109A - F may include per
warning signal from processed information and responding sonal computers such as desktops, laptops , notebooks ,
to the early warning signal in accordance with an aspect of mobile telephones or smartphones with applications and
the embodiments . other functionality, a handheld device with Wi-Fi or other
FIG . 28 illustrates a flow diagram for generating an early 5 wireless connectivity (e.g ., wireless enabled tablets, tablet
warning signal from processed information and responding computers , PDAs, and the like ), displays with built - in or
to the early warning signal in accordance with an aspect of external memories and processors, or any other known
computer, computing device , or handheld computer can also
the embodiments . be connected to one or more of the networks described
FIG . 29 illustrates a flow diagram for sending an early
warning signal to a plurality of computing devices in accor 10 herein . It is also contemplated that other types of devices
such as ATMs, kiosks, and other cash handling devices can
dance with an aspect of the embodiments . be connected to one or more of the networks described
FIG . 30 illustrates a flow diagram for displaying infor herein . These devices can be enabled to communicate with
mation obtained from an early warning signal at a comput wireless access points which in one example can be a series
ing device in accordance with an aspect of the embodiments . 15 of cellular towers hosted by a service provider. Additionally ,
DETAILED DESCRIPTION the wireless access points may be Wi-Fi (e .g ., compatible
with IEEE 802. 11a / b / g / and the like wireless communication
standards) connections and the computing devices may
In the following description of the various embodiments , obtain access to the Internet at these connections. Other
reference is made to the accompanying drawings , which 20 known techniques may be used to allow devices to connect
form a part hereof, and in which is shown by way of with a network .
examples various embodiments in which the disclosure may The term “ network ” as used herein and depicted in the
be practiced . It is to be understood that other embodiments drawings refers not only to systems in which remote storage
may be utilized and structural and functional modifications devices are coupled together via one or more communica
may be made without departing from the scope of the 25 tion paths, but also to stand -alone devices that may be
present disclosure . The disclosure is capable of other coupled , from time to time, to such systems that have storage
embodiments and of being practiced or being carried out in capability . Consequently , the term " network ” includes not
various ways. Also , it is to be understood that the phrase only a " physical network ” but also a " content network ,"
ology and terminology used herein are for the purpose of which is comprised of the data — attributable to a single
description and should not be regarded as limiting . Rather , 30 entity — which resides across all physical networks .
the phrases and terms used herein are to be given their The components may include data server 103, web server
broadest interpretation and meaning. For example , the use of 105 , and client computers 107, and devices 109A - F . Data
" including” and “ comprising" and variations thereof is server 103 provides overall access , control and administra
meant to encompass the items listed thereafter and equiva - tion of databases and control software for performing one or
lents thereof as well as additional items and equivalents 35 more illustrative aspects as described herein . Data server
thereof, and the use of the terms “mounted ," " connected ," 103 may be connected to web server 105 through which
" coupled ," " positioned ," " engaged ” and similar terms, is users interact with and obtain data as requested . Alterna
meant to include both direct and indirect mounting , con - tively , data server 103 may act as a web server itself and be
necting, coupling , positioning and engaging. directly connected to the Internet . Data server 103 may be
As noted above, various aspects of the disclosure relate to 40 connected to web server 105 through the network 101 ( e . g .,
predicting economic conditions . Before discussing these the Internet), via direct or indirect connection , or via some
aspects in greater detail, however, several examples of a other network . Users may interact with the data server 103
network architecture and a data processing device that may using remote computers 107, devices 109A - F , e. g ., using a
be used in implementing various aspects of the disclosure web browser to connect to the data server 103 via one or
will first be discussed . 45 more externally exposed web sites hosted by web server
Detailed Description of Example Network Architecture 105. Client computers 107 , 109 may be used in concert with
and Data Processing Device thatMay be Used to Implement data server 103 to access data stored therein , or may be used
Methods for Determining Economic Conditions: for other purposes. For example , from client device 107 or
FIG . 1 illustrates one example of a network architecture devices 109A - F a user may access web server 105 using an
and data processing device that may be used to implement 50 Internet browser, as is known in the art, or by executing a
one ormore illustrative aspects. Various network nodes 103, software application or app that communicates with web
105 , 107, and 109A -F may be interconnected via a wide area server 105 and/ or data server 103 over a computer network
network (WAN ) 101, such as the Internet. Other networks (such as the Internet ).
may also or alternatively be used , including private intra Servers and applications may be combined on the same
nets, corporate networks, LANs, wireless networks, per - 55 physical machines , and retain separate virtual or logical
sonal networks (PAN ), and the like. Network 101 is for addresses, or may reside on separate physical machines .
illustration purposes and may be replaced with fewer or FIG . 1 illustrates just one example of a network architecture
additional computer networks. A local area network (LAN ) that may be used , and those of skill in the art will appreciate
may have one ormore of any known LAN topology and may that the specific network architecture and data processing
use one or more of a variety of different protocols , such as 60 devices used may vary , and are secondary to the function
Ethernet . Devices 103, 105 , 107, 109A -F and other devices ality that they provide, as further described herein . For
(not shown ) may be connected to one or more of the example , services provided by web server 105 and data
networks via twisted pair wires, coaxial cable , fiber optics , server 103 may be combined on a single server.
radio waves or other communication media . For example , Each component 103, 105 , 107 , 109 may be any type of
the above connections can be made via the internet, blue 65 known computer, server, or data processing device as dis
tooth , WiFi, infrared , or any other known method ofwireless cussed herein . Data server 103 , e .g .,may include a processor
transmission . 111 controlling overall operation of the data server 103 . Data
US 9, 741, 048 B2
server 103 may further include RAM 113 , ROM 115 , these studies, an early -warning system of economic -insta
network interface 117, input/output interfaces 119 ( e.g ., bility detection can be developed . This early -warning sys
keyboard , mouse , display, printer , or the like ), and memory tem can also be used in identifying and assessing any type
121. I/O 119 may include a variety of interface units and of risk and opportunity that may arise . It can be used to
drives for reading , writing, displaying, and/ or printing data 5 mitigate risk or maximize gain .
or files . Memory 121 may further store operating system Additionally , aggregate macroeconomic events at the
software 123 for controlling overall operation of the data largest scale are built up from individual microeconomic
processing device 103 , control logic 125 for instructing data agents or activities at the most granular level. This may be
server 103 to perform aspects as described herein , and other reflected in utilizing network analysis to identify economic
application software 127 providing secondary , support, and 10 instability in financial systems.
or other functionality which may or may not be used in The interlinked ebb - and -flows of financial markets, which
conjunction with one ormore aspects described herein . The show the dynamic structure of financial systems, affect
control logic may also be referred to herein as the data server systemic risk inherent in it . In an ideal situation where
software 125 . Functionality of the data server software may relevant data or their possible proxies are available , banking
refer to operations or decisions made automatically based on 15 systems that are composed of a number of several connected
rules coded into the control logic , made manually by a user banks can be constructed . Key parameters that define the
providing input into the system , and /or a combination of structure of the system can be varied accordingly . These may
automatic processing based on user input ( e . g ., queries, data include the level of capitalization , the degree to which banks
updates, or the like ). are connected , the size of interbank exposures and the
Memory 121 may also store data used in performance of 20 degree of concentration of the system . The result of the
one or more aspects , including a first database 129 and a analysis of the influence of these parameters can be used to
second database 131. In some embodiments , the first data - potentially modify the financial system and mitigate risks as
base may include the second database (e . g ., as a separate needed .
table , report, or the like). That is , the information can be Network analysis can be applied to analyze the increas
stored in a single database , or separated into different 25 ingly complex and globally interlinked financialmarkets . A
logical, virtual , or physical databases , depending on system network is simply a collection of points ( or nodes ) joined by
design . Devices 105 , 107 , 109 may have similar or different lines (or edges ). Networks provide a simple but useful
architecture as described with respect to device 103 . Those representation of real-world systems of interacting compo
of skill in the art will appreciate that the functionality of data nents . The internet, for instance , can be represented as a
processing device 103 ( or device 105 , 107 , 109A - F ) as 30 network of computers linked by data connections. Other
described herein may be spread across multiple data pro - examples include social networks of friendships between
cessing devices, for example , to distribute processing load individuals and networks of business contacts .Networks are
across multiple computers , to segregate transactions based increasingly common in the study of biology, epidemiology ,
on geographic location , user access level, quality of service genetics , transportation , computer software, and so on . Net
(QoS), or the like . 35 works can also be applied in exploring some possible
One or more aspects may be embodied in computer - applications of network analysis in financial systems.
usable or readable data and / or computer- executable instruc It is possible to model the propagation of failures in a
tions , such as in one or more program modules, executed by financial system as an epidemic spreading process in a
one or more computers or other devices as described herein . network of interlinked stocks and flows of money. Starting,
Generally , program modules include routines , programs, 40 for example , from a small number of failed banks, the aim
objects, components, data structures, or the like that perform is to characterize the probability that failures propagate at
particular tasks or implement particular abstract data types the systemic level as a function of some relevant parameters ,
when executed by a processor in a computer or other device . like the connectivity of the network and concentrations,
The modules may be written in a source code programming correlations, sensitivities , leverage and liquidity of asset and
language that is subsequently compiled for execution , or 45 liability classes at banks. One of the key problems in this
may be written in a scripting language such as (but not context is to understand the role of the network structure in
limited to ) HTML or XML . The computer executable relation to economic instability .
instructions may be stored on a computer readable medium System variables that show purely random fluctuations or
such as a hard disk , optical disk , removable storage media , perfectly periodic rhythmsdefine idealized extremes . In fact,
solid state memory , RAM , or the like . As willbe appreciated 50 some parameters whose fluctuations may seem random
by one of skill in the art, the functionality of the program when viewed in isolation can be highly predictable in the
modules may be combined or distributed as desired in temporal context of variation in other parameters; hence,
various embodiments . In addition , the functionality may be variations in one system variable can convey substantial
embodied in whole or in part in firmware or hardware information about variation in another. An organism (or a
equivalents such as integrated circuits , field programmable 55 company that is capable of learning these correlations can
gate arrays (FPGA ), and the like. Particular data structures exploit them in order to anticipate vital changes in the
may be used to more effectively implement one or more system . Stressful stimuli may be important not because of
aspects , and such data structures are contemplated within the their immediate and direct consequences , but in the infor
scope of computer executable instructions and computer- mation they convey about the overall state of the system and
usable data described herein . 60 its likely trajectory . Therefore, survival may depend much
Detailed Description of Example Methods and Systems on the art of correctly identifying the system 's trajectory
for Determining Economic Conditions: coupled with the ability to adapt to changes in the system .
Identifying structural attributes of the interconnected Robustness of the economic system and the ability to
world economic network can provide clues about which withstand economic shock can be promoted by modular
characteristics of the system correlate with a certain degree 65 ity — the degree to which nodes of a system can be decoupled
under systemic distress leading to economic instability or into relatively discrete components. For example , a basic
economic stability. Based on the understanding derived from principle in management of forest fires and epidemics is that
US 9, 741, 048 B2
if there is strong interconnection among the system ele moment becomes more and more like its past state . The
ments , a perturbation will encounter nothing to stop it from resulting increase in “memory ” of the system can be mea
spreading. But if the system is appropriately compartmen - sured in various ways from the frequency spectrum of the
talized through , say, introduction of fire breaks and quaran system . One approach is to look at the lagging autocorrela
tining, disturbance or risk is more easily countered . 5 tion , which can directly be interpreted as slowness of
The introduction of modularity , however, will often recovery in such natural perturbation regimes . Analysis in
involve a trade-off between local and systemic risk . More the models exposed to stochastic forcing confirms that if the
over, the wrong compartmentalization in financial markets system is driven gradually closer to a catastrophic state ,
could preclude stabilizing feedbacks, such as mechanisms there is a marked increase in autocorrelation that builds up
for maintaining liquidity of cash flows through the financial 10 long before the critical transition takes place . This is true
system , where fragmentation leading to illiquidity could even for realistic models . Increased variance in the pattern
actually increase systemic risk . Redundancy of components of fluctuations can be seen as another possible consequence
and pathways , in which one can substitute for another, is of critical slowing as a critical transition is approached . As
also a key element in the robustness of complex systems, the Eigenvalue approaches zero , the impacts of shocks do
and effective redundancy is not independent of modularity . 15 not decay , and their accumulating effect increases the vari
There are only coarse or indirect options for control of the ance of the state variable . In principle , critical slowing could
financial system . The tools available to policymakers are reduce the ability of the system to track the fluctuations, and
designed to modify individual incentives and individual thereby produce an opposite effect on variance . However,
behaviors in ways that will support the collective good . Such analysis shows that an increase in the variance usually arises
top - down efforts to influence individual behaviors can often 20 and may be detected before a critical transition takes place .
be effective , but in certain instances itmay be a challenge to This detection technique can be implemented both tempo
control the spread of synchronized behaviors or to manage rally and spatially .
financial crises in an optimal way . Methods for identifying or assessing any type of risk
Although the study of payment flows is of interest to and /or opportunity thatmay arise can include either, alone or
central bankers, in certain instances it may miss aspects of 25 in combination , band pass filtering , principal component
systemic risk , namely public perception and asset valuation analysis, random matrix theory analysis, synchronization
associated with the interaction of counterparties (the mutual analysis, and early -warning detection . Each technique can
financial obligations and exposures that link companies ). also be viewed as a process that takes a set of inputs and
Such company networks are helpful in studying the effects converts it to a set of outputs . These outputs can be used as
of inflated asset - pricing , credit crises and the poorly under - 30 inputs for a subsequent process or the outputs may be
stood but potentially worrying effects of the current wide- directly actionable for formulating certain economic predic
spread use of derivatives by investment banks to manage tions to make investment decisions and the like.
risk in real-time. Whatever the case , it seems that networks A general diagram of this approach is illustrated in FIG .
that define financial reality and globalmarkets may be of use 2 . FIG . 2 shows a diagram of an analysis where the system
to understanding the market robustness and ability to thrive 35 analyzes a series of inputs , which can be any financial data ,
as well as its potential vulnerability to collapse . for example, consumer leverages , commercial leverages ,
Complex dynamical systems, financial system included , and unemployment rates and the like. These inputs are
may have transitional points where a sudden shift to a totally processed using any of the techniques described herein to
different dynamical regimemay take place . Though predict- produce an output or economic indicator. These economic
ing such tipping points is very difficult , the generic early - 40 indicators can be used to develop strategies for making
warning signs that may indicate if a critical threshold is investments ,mitigating risks, creating new products , and the
approaching may exist in reality . like. The various outputs from the techniques described
Many complex dynamical systems have critical thresh - herein can be used to arrive at a decision , e . g . to sell, invest,
olds called tipping points where the system abruptly shifts m itigate risk , and the like.
from one state to another. In global finance , there is big 45 As schematically illustrated in FIG . 3 , either approach
concern about a sudden systemic decline in market prices works by converting various sets of inputs into outputs that
that can threaten the global financial system which , in turn , may be used to guide financial decisions . Based on the
may lead to global economic downturns. It is traditionally understanding derived from these studies, an early -warning
difficult to predict such critical transitions since the state of system of assessing and dealing with various risks (includ
the system may show little change before the tipping point 50 ing, but not limited to , economic instability ) is also devel
is reached . However, it seems that certain generic signs may o ped . In one example , as shown in FIG . 3 , various outputs
take place in a wide class of systems as they approach a can be reviewed to arrive at the financial decision , for
critical point . The dynamics of systems near a critical point example , outputs from a top - down review of financial data ,
have generic properties, regardless of differences in the bottom up review of financial data , and other outputs .
details of each system . Critical thresholds for such transi- 55 Certain techniques can be used in a top -down approach
tions correspond to bifurcations. In the so -called cata - that reviews inputs from the macro level down to the micro
strophic bifurcation , once a threshold is exceeded , a positive level or a bottom up approach that works in an opposite
feedback mechanism pushes the system through a phase of fashion . Considerable noises make modeling time series
directional change towards a contrasting state . The transi- data in its most granular level ( such as modeling account
tions from a stable equilibrium to a cyclic or chaotic state 60 level in portfolio analysis or relationship networks in
can also happen in some other types of bifurcations. finance ) more difficult. To deal with these difficulties, a
Three possible early -warnings in the dynamics of a sys - top -down approach can be conducted . This approach is
tem approaching a tipping point may be used in emerging intended to firstly understand the behavior at the most
risk detection : slower recovery from perturbations, aggregate level (top ), where the relevant national level
increased autocorrelation and increased variance . 65 macroeconomic time series are used . Then the analysis is
Because slowing down causes the intrinsic rates of change gradually performed towards the more granular level
in the system to decrease, the state of the system at any given (down ), where time series data being used includes, but not
US 9, 741, 048 B2
10
limited to , relevant geographic or state level, and segmental predict certain economic trends , for example, the level of
level. As information is viewed down from the most aggre leverage of consumer and commercial sectors . An example
gate level to the more granular level, any useful derivable output of the absolute values is illustrated in FIG . 8B . As
information is taken into consideration , such as how shocks, shown in 8B , in this example, a higher correlation value may
policies, and strategies may contribute to the behaviors 5 be associated with a weaker state in the economy, whereas
being observed . a lower correlation value may indicate a stronger performing
A complexity pyramid composed of various levels of economy. However, higher correlation value may reflect a
granularity is shown in FIG . 4 . The bottom of the pyramid better state of the economy, if the level of saving is consid
shows the representation of specific individual to small- ered . Therefore, higher correlation may be associated with
sized economic agent network or circuit (level 1 ) . Insights 10 stronger or weaker performing economy based on the data
into the logic ofmost granular organization can be achieved being observed .
when the economic system is viewed as a complex network Principal Component Analysis :
in which the components are connected by some kind of Once it is determined whether or not the economy is
relationship links . At the lowest level, these components moving in a positive or negative direction , the next step is
form medium -sized networks ( level 2 ), which in turn are the 15 to determine if there are any particular elements driving the
building blocks of functional modules (level 3 ). These change . On a macro level this could be certain industries or
modules are nested , generating scale- free or other kinds of asset classes. On a micro level it could be a certain compa
network architectures (level 4 ). Although individual com - ny 's economic performance or group of companies ' eco
ponents are unique to a given microeconomic system , the nomic performances . Principal component analysis or PCA
topologicalproperties of the networks share similarities with 20 can be used to identify the particular " drivers” of the trends .
universal organizing principles. This can indicate which subcomponents are accounting for
Band Pass Filtering Technique: the change in state .
A band pass filtering technique can be applied to eco - As shown in FIG . 9 , largest ratios of two predetermined
nomic inputs in order to extract a signal from noise by components may illustrate a particular economic driver. As
removing certain data or frequencies that are too rapid or too 25 shown in FIG . 9 , two subcomponents can be compared by
slow to be “ true,” similar to electronic signal processing . calculating the ratio between their first and second principal
Certain filters can be applied to economic time series data to components . In one example , the first principal component
identify the particular phase in an economic cycle . The can represent systemic risk , for example , a recession or other
analysis may help guide certain decisions such as whether to event external to the data being studied . The second prin
loosen or tighten credit availability . 30 cipal component can be non - systemic (idiosyncratic ) risk ,
FIG . 5 illustrates a process flow of an application of a which can be specific to a particular subcomponent. In FIG .
band pass filter to a series of inputs 502 . These inputs can be 9 , Subcomponent 5 may be the market driver because the
consumer and commercial leverages and unemployment ratio of the first principal component to the second principal
rates, and can be either macroeconomic or microeconomic component is the largest.
data . First the series of inputs is collected and transformed 35 In another example , certain economic instability markers
to a proper frequency domain at 504. The band pass filter can can be identified using spectral analysis and principal com
then be applied at step 506 to generate a meaningful cycle ponent analysis . In this case , because the typical size of a
signal in step 508 . FIGS. 6A -6C show a raw time series U .S . business cycle is approximately 70 months, this data
transformed into a frequency domain with the signal can be analyzed as part of a study. Spectral analysis of
extracted . In particular, a band pass filter is applied to the 40 correlation matrices can be conducted to determine eco
data depicted in FIG . 6A and FIG . 6B to produce the filtered nomic instability markers .
data shown in FIG . 6C . The data shown in FIG . 6 can then In certain examples , the analysis can be applied to uncol
be evaluated to determine a synchronization point or region lectables , where it has been declared that certain amounts of
of the data that can be used to make economic predictions money are unlikely to be collected , by sectors as shown in
based on the synchronization events . 45 FIGS. 10 - 13 . This analysis can also be applied to the Home
As shown in FIG . 7 , the band pass filtering technique can Price Index (“ HPI” ) by states as shown in FIGS . 14 - 17. By
be applied to multiple economic indicators simultaneously. first scaling the time series, a correlation matrix is generated
If the correct band pass filter is implemented , when each of within a moving timewindow with the length of roughly one
the signals synchronize , it can indicate a transition from an business cycle (~ 70 months ). In each of the moving win
equilibrium state to either a positive state (economic boom ) 50 dows, principal component analysis (PCA ) can be con
or a negative state (recession ). FIG . 7 illustrates a synchro -ducted . The first principal component ( PC1) is usually called
nization event in the period around 2010 , which shows a “ systemic risk ," which basically represents the collective
positive state of the economy. response of all component time series under consideration to
During a synchronization event, the risk or reward can exogenous stimuli. The behavior of the PC1 eigenvector
increase and the benefits of diversification are diminished . 55 contents ( or factor loadings ) over time can be investigated .
The synchronization level can be measured based on the A quality called dispersion level is provided to give a
collective correlation of a group of time series. This corre measure of disparity or diversity of the time series being
lation could also be applied to additional measures other analyzed .
than the simple filtered series . For each analysis, a set of figures are shown : the time
For example , the absolute value of the volatility of the 60 series PCA ( FIGS. 10 and 14 ), PC1 Eigenvector contents
changes can be determined . This process is depicted gener (FIGS. 12 and 16 ), and PC1 dispersions (FIGS. 13 and 17).
ally in FIG . 8A . As shown in FIG . 8A , time series data 802 The vertical black ribbons show recessions, and the vertical
can be processed by a band pass filter, at step 804 , can be green ribbons show the 1995 / 1996 period where the bad
applied to the economic data and the absolute value of the consumer cycle took place without being accompanied by a
data can be taken at step 806 to locate correlations across all 65 recession . As can be seen in FIGS . 10 and 14 , the time series
of the series of data at step 808 . The correlation and data can be scaled . The eigenvalues obtained using PCA are
concentration measures can then be reviewed at step 810 to shown in FIGS. 11 and 15 and can be normalized such that
US 9, 741, 048 B2
11 12
their total summation is equal to 100 % . FIGS. 12 and 16 comparison between the observed eigenvalues and the ana
show the contents of the first principal component ( or, lytical “ null” distribution of RMT can be used to obtain an
principal component number 1 ) . estimate of the number of components . Specifically , the
Exemplary charts of the dispersion level over time are number of observed eigenvalues larger than the analytical
shown in FIGS. 13 and 17 . The dispersion level may be 5 maximum provides an estimate of the number of significant
taken as representing the levelof economic instability risk of eigenvalue components . Hence, the observed eigenvalues of
the system under consideration . From the dynamical behav the data correlation matrix larger than the theoreticalmaxi
ior of the dispersion level, an early -warning tool of eco mum provides a reasonable approximation to the number of
nomic instability risk may be developed . In particular, the principal components to retain , i.e ., the number of eigen
higher the value of the dispersion level the less contagious 10 value components obtained using PCA that really matter in
the system is . As shown in FIGS. 13 and 17 recessions the calculation . RMT is analytical in nature , and the com
promoted the decrease of the dispersion level . This can be putational cost of RMTmay be small in certain instances.
interpreted as the phenomenon of the economic instability of
systemic risk . Taking a matrix whose elements are the correlation coef
ng
FIG . 18 shows an exemplary flow chart for implementing ficient values ofmultiple time series of interest, the standard
15 nici
an algorithm thatmay detect economic instability risk based Pearson correlation coefficients can be defined as :
on eigenvector dispersion “ level.” In step 1800 , principal
component analysis is applied to several time series under
consideration by using an appropriate time window . In step
1802, the eigenvector content of principal component num - 20
ber 1 is normalized for each time window . The eigenvectors
(8:8)= 2 ( May 184
k = 1 .NL
are then sorted , e .g., from smallest to largest, say from 1 to where Mgi and Mg; are the average of g; and g; respectively ,
8 for each time window in step 1804 . In step 1806 , the and og and 0 .; are their corresponding standard deviations ,
distances between the eigenvectors are then calculated e .g .
1 to 2 , 2 to 3 , 3 to 4 , and so on , specific for each time window 25 andThe
N is the total number of observations .
statistical properties of the eigenvalues of random
(note that the order of eigenvectors may change for different matrices are known in the limit of very large dimensions .
time window ). In step 1808, for each time window , the mean Particularly, in the limit N > 0, L > 0, such that Q = L /N is
( or, the variance ) of the remaining distances is calculated , fixed , the distribution Prm a ) of eigenvalues À of the
where this entity represents the dispersion “ level” at specific
timewindow ( the higher the value, the higher the dispersion 30 random correlation matrix is given by
level). In step 1810 , economic instability is then predicted
based on the dispersion level.
Band Pass Filtering and Principal Component Analysis : V (14 - 1a)( - )
PRM ((4A)) == 2 27 +
PRM
Band pass filtering and principal component analysis can
be used in conjunction with each other to make economic 35
predictions. FIG . 19 shows an exemplary flow chart of an for å within the bounds à sà ;sì , where à _ and at, are the
example implementation ofband pass filtering and principal minimum and maximum eigenvalues of the random corre
component analysis being used together to determine eco - lation matrix , respectively , given by
nomic instability risk . In this example , publicly available
macroeconomic time series data 1902 can be applied to a 40
band pass filter 1904 to obtain a filtered data series 1906 .
The filtered data series can then be analyzed to determine a + = 1 + = + 2
O
10
particular economic cycle state , e. g . positive or negative
trend at step 1908. Once a state of the market is detected
(steps 1910 - 1914 ), principle component analysis can be 45 RMT focuses on the study of statistical properties of
applied to determine the particular driver of the state of the eigenvalue spacing between consecutive eigenvalues. From
market at step 1916 . Using the drivers obtained in the RMT, distribution of eigenvalue spacing of real and sym
analysis ( as well as the correlation / concentration ) at step metrical random matrices follows two universal laws
1918 , a portfolio optimizer 1922 can , for example , be depending on the correlativity of eigenvalues. Strong cor
implemented for making investments , mitigating risks, cre - 50 relation of eigenvalues leads to statistics described by the
ating new products , and the like at step 1924. Additionally, Gaussian Orthogonal Ensemble (GOE). On the other hand ,
proprietary micro level data can be analyzed at step 1920 . eigenvalue spacing distribution follows Poisson statistics if
Random Matrix Theory : there is no correlation between eigenvalues . To express it
Another approach to spectral analysis of economic data is differently , eigenvalue spacing distribution of a random
using Random Matrix Theory (RMT) to analyze economic 55 matrix with non -zero values only for its diagonal (or block
data and predict trends in the market. Random Matrix diagonal parts) follow Poisson statistics , because eigenval
Theory (RMT) was initially proposed by Wigner and Dyson ues of this system are not correlated due to the absence of
in 1960s for studying the spectrum of complex nuclei. RMT interaction between diagonal (or block -diagonal) parts .
can be used to identify and model phase transitions and To validate RMT estimate and to ensure that the theoreti
dynamics in physical systems and can also be used to create 60 cal null distribution does not deviate significantly from that
financial and economical models . of empirical null (deviations might be expected because of
For example , RMT can be used to estimate the number of the finite size of the matrix and because the data may not be
dimensions (components ) of a data correlation matrix by gaussian ), the data matrix may be scrambled ( for each row ,
comparing the statistics of the observed eigenvalues , i.e . the a distinct permutation of the columns is performed ) and
eigenvalues of the data correlation matrix to those of a 65 verified that RMT predicted zero significant components .
random matrix counterpart. The density distribution of The nearest neighbor spacing distribution (NNSD ) of
eigenvalues for such random matrices is known, so that the eigenvalues, P ( s), of RMT can also be used in some calcu
US 9 ,741,048 B2
13 14
lations, as discussed below . This is defined as the probability An example flow chart using RMT to estimating dimen
density of the so -called unfolded eigenvalue spacing s ionality of the data used for PCA is depicted in relation to
S = e;+ 1 -e; where e =Nay (E ;), and E ; (i= 1, N ) is the eigenval FIG . 20B . First, the data of multiple time series of interest
ues of the matrix (N being the order of the matrix ), and Nav is determined at step 2020 . Next, the standard Pearson
is the smoothed integrated density of eigenvalues obtained 5 correlation coefficients are formulated using the data at step
by fitting the original integrated density to a cubic spline or 2022 . The matrix whose elements are the correlation coef
by local density average . From RMT, P (s ) of the GOE ficient values ( call the matrix CCM or Correlation Coeffi
statistics closely follows Wigner-Dyson distribution cient Matrix ) is then determined at step 2024 . The relevant
PGOE (S)~ 1/27ts exp ( - as -/4 ). In the case of Poisson statistics, RMT is formed where the density distribution of eigen
P (s) is given by Poisson distribution P Poisson (s) = exp ( - s). " values for such random matrices is known at step 2026 . The
The difference between Wigner -Dyson and Poisson distri- statistics of the observed eigen - values, i.e . the eigen - values
butions manifests in the regime of small s, where PGOE of the data correlation matrix is then compared to those of
s( 0 ) = 0 and PPoisson
Poi (s -> 0 ) = 1 . the RMT counterpart at step 2028 . At step 2030, the number
Note that RMT technique can be applied to both higher 15 of observed eigen -values larger than the analytical maxi
and lower level views. In higher level, it may be used to mum from RMT provides an estimate of the number of
extract the important companies/ groups which are most significant eigen - value components of the data. The
responsible for the economic dynamics being observed . At observed eigen -values of the data correlation matrix larger
a lower level, it may be used to extract the underlying than the theoretical maximum from RMT provides a rea
skeleton of some complex networks. 20 sonable approximation to the number of principal compo
To test the “ modularity ” of the clustering, lower values of nents to retain , i.e. the number of eigen -value components
correlation coefficients can be removed as given by the obtained using Principal Component Analysis (PCA ) that
equation are relevant in the calculation . The retained principal com
ponents (whose numbers are usually much lower than the
25 original dimensionality of the data ) can be used in the
analysis .
esi,8,)= 2 ( 7M884 MIN)
k = 1, 1 Early Warning Detection of Economic Conditions:
Complex dynamical systems, financial system included ,
can have transitional points where a sudden shift to a totally
from the data of interest, starting from the lowest . Using 30 different dynamical regimemay take place . Though predict
chi-square test, a sharp transition from a Wigner-Dyson ing such tipping points can be difficult , generic early
distribution to a Poisson distribution would be observed at a warning signs may indicate if a certain threshold is
certain defined “ cutoff” level q . Once this takes place , the approaching .
desired clustering can be obtained . It should be noted that Many complex dynamical systems have critical thresh
this approach is different from existing clustering methods , 35 olds called tipping points where the system abruptly shifts
where here cutoffs or thresholds used for clustering are from one state to another. For example, in global finance ,
determined self - consistently by the transition given by there is big concern about a sudden systemic decline in
RMT. For segmentation , the time series that are used could market prices that can threaten the global financial system
be FICO scores, geographical performance , loss, revenue, which , in turn , may lead to a global financial crisis. Itmay
and so on . 40 be difficult to predict such critical transitions since the state
FIG . 20A shows an example flow chart using the RMT of the system may show little changes before the tipping
method to determine economic instability in financial sys - point is reached . Additionally , models of complex systems
tems in segmentation or clustering , for example in risk may not be accurate for predicting where critical thresholds
segmentation or population clustering . First data ofmultiple are located . However, it seems that certain generic signsmay
time series of interest is obtained in step 2000 . Next, 45 take place in a wide class of systems as they approach a
standard Pearson correlation coefficients using the data are critical point. The dynamics of systems near a critical point
built at step 2002 . The matrix whose elements are the have generic properties, regardless of differences in the
correlation coefficient values ( call the matrix CCM or Cor - details of each system . Critical thresholds for such transi
relation CoefficientMatrix ) is then formulated at step 2004 . tions correspond to bifurcations. In the so - called cata
The nearest neighbor spacing distribution by generating it 50 strophic bifurcation , once a threshold is exceeded , a positive
from the differences between nearest neighbor eigen - values feedback mechanism pushes the system through a phase of
derived from CCM is then obtained at step 2006 . The lowest directional change towards a contrasting state . The transi
value of correlation coefficients is iteratively set to zero , and tions from a stable equilibrium to a cyclic or chaotic state
the calculation is repeated at step 2008 . Using a chi-square can also happen in some other types of bifurcations.
test, a sharp phase transition from Gaussian Orthogonal 55 FIGS. 21A - 21H illustrate eight panels of the same exem
Ensemble (RMT) distribution to Poisson distribution would plary economic data , which show the evolution of an exem
be observed at a certain " threshold ” level at step 2010 . Once plary system over time. Each panel in FIGS. 21A -21H show
this takes place , optimal segmentation or clustering is three graphs: the left graph shows a frequency distribution ,
achieved . The mathematical philosophy is basically trying to the middle graph shows a phase distribution , and the right
make CCM as diagonal as possible , but not necessarily 60 graph shows the phases in polar form on a unit circle . As
completely diagonal, and the process should be stopped shown in FIGS. 21A - 21C , from time= 0 to 20 months, the
once a phase transition takes place at step 2012 . The output exemplary economic data is out of phase with each other.
could be in form of optimal segmentation or clustering, and However, starting at time= 30 months, the exemplary eco
software could be built to detect the threshold where the nomic data starts to move into phase with each other.
transition takes place . This could be associated with eco - 65 FIG . 22A shows the evolution of the order parameter over
nomic instability risks or simply optimal risk segmentation time, and FIG . 22B illustrates how the phase distribution is
or population clustering only . locked to have the values which are confined within a certain
US 9 ,741,048 B2
15
(moving ) range only . The order parameter, in this case,
shows the level of synchronicity of the system as a whole .
In one example , indicators of whether a system is getting
close to a critical threshold may be related to a phenomenon
called critical slowing down, for example , fold catastrophe . 5
f (x ) + = ?x
XE
ser)tot =100) eao e bret
At a fold bifurcation point, the dominant eigenvalue char- It is easy to see from the above equations
acterizing the rates of change around the equilibrium is zero .
Therefore , as the system approaches such critical points , it
becomes increasingly slow to recover from small perturba
tions. Such slowing down typically starts far from the 10
bifurcation point, and that recovery rates decrease smoothly and
Az = |c==3( –a).
to zero as the critical points is approached .
The recovery rate after a small perturbation will be
reduced , and will approach zero when a system moves
12 = ax |b = =y(b –a).
towards a catastrophic bifurcation point, which can be 15
further explained by the following simple dynamical system , that at the bifurcation (b = a ) the recovery rates à , and 12 are
where y is a positive scaling factor and a and b are param both zero and perturbations will not recover. Farther away
eters : from the bifurcation , the recovery rate in this model is
linearly dependent on the size of thebasin of attraction (ba).
20 For more realistic models , this is not necessarily true but the
dx
dr = y(x – a )(x – b ). relation is still monotonic and is often nearly linear.
The most direct implication of critical slowing down is
that the recovery rate after small perturbation can be used as
This model has two equilibria , x , -a and X2 =b , of which an indicator of how close a system is to a bifurcation point.
one is stable and the other is unstable . If the value of a equals 25 For most natural systems, it would be impossible to monitor
to that of b , the equilibria collide and exchange stability in them by systematically observing recovery rates. However,
a transcritical bifurcation ). Assuming that Xi is the stable it can be shown that as a bifurcation is approached in such
equilibrium , it can be determined what happens if the state a system , certain characteristic changes in the pattern of
of the equilibrium is slightly perturb fluctuations are expected to take place . An important pre
30 diction is that the slowing down should lead to an increase
in autocorrelation in the resulting pattern of fluctuations .
( x = x1 + E ): d- (71 + E) – f(Ti + ?). Critical slowing down will tend to lead to an increase in
the autocorrelation and variance of the fluctuations in a
Here f(x ) is the right hand side of the above equation 35 stochastically forced system approaching a bifurcation at a
threshold value of a control parameter. The example given
here illustrates why this is the case . In certain instances it
may be assumed , there is a repeated disturbance of the state
= y (x – a )(x – b ). variable after each period At (that is, additive noise ).
10 Between disturbances, the return to equilibrium is approxi
mately exponential with a certain recovery speed , A . In a
Linearizing the equation using a first -order Taylor expansion simple autoregressive model this can be described with the
yields which simplifies to following equations: Xn + 1 -X = exp (24t)(x , - X ) + 0€7 , and
Yn + 1 =exp (^ At)yn + + 0€m. Here yn is the deviation of the state
45 variable x from the equilibrium , e , is a random number from
d (X1 + E )- = a standard normal distribution and o is the standard devia
dt
tion . If and At are independent of yn , this model can also
f(x1 + 5)= f(81)+ON
f ( x1 + E ) {|X,8,f(fi)+ de = f(8 )+ 16= date = 11€.
f ( x ) + - - | x | f, f ( x ) + . +
de
- | kyº = — = / E
dt
be written as a first-order autoregressive (AR ( 1 )) process :
ymu , ray,, + 0€, . The autocorrelation a = (ast) is zero for
50 white noise and close to one for red ( autocorrelated ) noise .
With eigenvalues 41 and 4 , in this case , The expectation of an AR ( 1) process : Yn + 1 = c + ayn + oen is
1 = a= -yb-a),
Ja = - yb - a ), 55 E (Yn + 1 ) = E ( c) + a E (yn ) + E (Ten ) = M = c + api + 0 = p = 1
and , for the other equilibrium For c = 0 , the mean equals zero and the variance is found to
be
60
12 = b= =y(b-a) Var (yn+1) = E( ym ) –4 ? = 2

are obtained
If b > a then the first equilibrium has a negative eigenvalue , 65 Close to the critical point, the return speed to equilibrium
ai, and , therefore , it is stable ( as the perturbation goes decreases , implying that approaches zero and the autocor
exponentially to zero ; relation a tends to one. Therefore , the variance tends to
US 9, 741, 048 B2
17 18
infinity. These early -warning signals are the result of critical discussed herein . Alternatively an apparatus comprising : a
slowing down near the threshold value of the control param - processor ; and a memory for storing computer readable
eter. instructions that, when executed by the processor, can cause
Slowing down may cause the intrinsic rates of change in the apparatus to perform one or more aspects of the methods
the system to decrease , and the state of the system at any 5 discussed herein . In other embodiments , one or more non
given moment may become more and more like its past transitory computer- readable media may have instructions
state . The resulting increase in “ memory ” of the system can stored thereon that, when executed , cause at least one
be measured in various ways from the frequency spectrum computing device to perform one or more aspects of the
of the system . One example approach is to calculate a lag - 1 methods discussed herein .
autocorrelation , which can be directly interpreted as slow - 10 Features of Methods and Systems for Determining Eco
ness of recovery in such natural perturbation regimes . nomic Conditions According to Examples of the Disclosure :
Additionally , analysis in the models exposed to stochastic In one example , a method can include one or more of the
forcing confirms that if the system is driven gradually closer following steps : receiving an input of economic data includ
to a catastrophic bifurcation , there is a marked increase in ing a plurality of macroeconomic and microeconomic indi
autocorrelation that builds up long before the critical tran - 15 cators , applying a band pass filter to the economic data ,
sition takes place . This is true even for realistic models . extracting a cycle signal from the band pass filter , determin
Whereas increased variance in the pattern of fluctuations can ing a synchronization event of the plurality of macroeco
be seen as another possible consequence of critical slowing nomic and microeconomic indicators , and outputting an
down as a critical transition is approached . As the eigenvalue economic prediction based on the synchronization event.
approaches zero , the impacts of shocks do not decay, and 20 The band pass filter can be applied to the plurality of
their accumulating effect increases the variance of the state macroeconomic and microeconomic indicators simultane
variable . In principle, critical slowing down could reduce the ously . After applying a band pass filter to the economic data
ability of the system to track the fluctuations , and thereby the absolute value of the plurality of the filtered economic
produce an opposite effect on variance. However, analysis data series is calculated . The absolute value of the filtered
shows that an increase in the variance usually arises and may 25 data can be reviewed to determine correlations and concen
be detected before a critical transition takes place. tration measures across the plurality of economic indicators
Phenomenon of critical slowing down may lead to three to predict economic trends. A principal component analysis
possible early -warning signals in the dynamics of a system can be applied to the economic data to determine a market
approaching a bifurcation that may be used in emerging risk driver.
detection : slower recovery from perturbations, increased 30 In another example , a method can include one ormore of
autocorrelation , and increased variance. the following steps: determining a first principal component
An exemplary process is depicted in FIG . 24 , which number from the economic data, determining eigenvector
shows an exemplary process for determining an early warn - contents of the first principal component number, normal
ing signal for economic instability. In the first step , a data izing the eigenvector contents of the first principal compo
series of interest is obtained in step 2400 to determine the 35 nent number, sorting the normalized eigenvector contents
particular early warning signal . Next, the Auto Regressive from smallest to largest, calculating the distances between
(“ AR ( 1 )" ) Model coefficients of a time series can be calcu - the normalized eigenvector contents , calculating the mean of
lated in step 2402 . The scaled lag- 1 auto - correlation of the the distances to obtain a dispersion level, and outputting an
coefficients can be determined in step 2404 . Finally , the economic instability prediction based on the dispersion level
scaled smoothed derivative of the previous computation can 40 of the distances. The predetermined time window can be set
be determined at step 2406 . The resulting derivative can be to approximately 70 months , though it can have any value
used as an early warning detector. For example, an indicator deemed reasonable . The first principal component number
as a positive or negative warning could be provided to the can represent systemic risk . The eigenvalue contents can be
user. In particular, the system could be configured to output normalized such that the eigenvalue contents total summa
a positive indicator when the scaled smoothed derivative is 45 tion is 100 % . A lower dispersion level may indicate a greater
above a predetermined threshold value and can be config - risk of economic instability . The inputof economic data can
ured to output a negative indicator when the scaled include both microeconomic data and macroeconomic data .
smoothed derivative is below a predetermined threshold Prior to applying principal component analysis the economic
value . data can be passed through a band pass filter to determine a
In one example , the above-mentioned process described 50 cycle state .
in relation with FIG . 24 can be applied to the Home Price In another example , a method can include one or more of
Index (“ HPI” ) as shown in FIGS . 23A - 23D to detect certain the following steps: receiving economic data of a predeter
economic conditions . FIGS. 23A - 23D show the scaled HPI mined time window and applying principal component
(blue curve ), the scaled lag - 1 auto - correlation of AR ( 1 ) analysis to the economic data , determining a first principal
coefficients (green curve ), and the scaled smoothed deriva - 55 component number, determining a second principal compo
tive of the lag - 1 auto - correlation of AR1 coefficients ( red nent number , calculating the ratio between the first principal
curve). The results are shown for Charlotte in FIG . 23A , Las component number and the second principal component
Vegas in FIG . 23B , Cleveland in FIG . 23C , and Phoenix in number, determining the largest ratio between the first
FIG . 23D . The last monthly calendar time being used is principal component number and the second principal com
August 2009. In this example , the scaled lag - 1 auto - corre - 60 ponent number, and outputting a market driver prediction
lation of AR ( 1 ) coefficients and the scaled smoothed deriva - based on the largest ratio between the first principal com
tive of the lag - 1 auto -correlation of AR ( 1 ) coefficients can ponentnumber and the second principal component number.
serve as the leading indicators for HPI. A monitoring process The predetermined time window can be approximately 70
can be configured to monitor the leading indicators for the months . The first principal component can be systemic risk
HPI. 65 and the second principal component can be non - systemic
In certain examples a computer -implemented method can risk . The input of economic data may include both micro
be employed to perform one or more aspects of the methods economic data and macroeconomic data .
US 9, 741, 048 B2
19 20
In another example , a method can include one ormore of phase of directional change towards a contrasting state . The
the following steps: receiving economic data of a multiple transitions from a stable equilibrium to a cyclic or chaotic
time series , determining correlation coefficient values using state can also happen in some other types of bifurcations.
the data of multiple time series, creating a matrix using the Three possible early -warnings in the dynamics of a system
correlation coefficient values, determining the nearest neigh - 5 approaching a tipping point may be used in emerging
bor spacing distribution , determining lowest values of cor - instability detection : slower recovery from perturbations ,
relation coefficients and set the lowest value of correlation increased autocorrelation and increased variance .
coefficients iteratively to zero , determining a threshold level When an economic system approaches a critical threshold
by using a chi -square test, and outputting an economic (e . g ., a tipping point), the system may abruptly shift from
instability prediction based on the threshold level. The 10 one state to another ( e . g ., where the value of underlying
nearest neighbor spacing distribution can be determined by economic instrument becomes appreciably lower or higher
taking the differences between nearest neighbor eigen -val than a previous value ). Consequently, it may be important
ues derived from a correlation coefficientmatrix . The thresh - that an early warning signal be expeditiously conveyed to a
old level can be determined by detecting a sharp phase user who is responsible for the underlying economic instru
transition from a Gaussian Orthogonal Ensemble distribu - 15 ment. Those of skill in the art will appreciate that there may
tion to a Poisson distribution . A standard Pearson correlation be instances where instability may occur within minutes or
coefficient values can be determined using the data of seconds after receiving an early warning signal. Those of
multiple time series. skill in the artwill also appreciate that economic instruments
In another example , a method can include one or more of may be a tradable resource of any kind . Economic instru
the following steps: obtaining economic data of multiple 20 ments may be cash , evidence of an ownership interest in an
time series of interest, determining correlation coefficient entity , a contractual right to receive or deliver cash or
values using the data , forming a matrix of the correlation another economic instrument, or a derivative . Cash instru
values and determining the eigen - values of the correlation ments may be securities which are readily transferable or
matrix , forming a random matrix counterpart with known may be instruments such as loans, where both borrower and
eigen - value density distribution , comparing statistics of the 25 lender have to agree on a transfer. Derivative instruments
eigen - values of the data correlation matrix to the values of may derive their value from the value and characteristics of
the random matrix counterpart of known eigen - value distri- one or more underlying entities such as a resource, index , or
bution , determining the difference between the number of interest rate . Derivative instruments may be exchange
the eigen - values of the data correlation matrix and the traded derivatives and over -the -counter derivatives.
random matrix counterpart, and outputting a the number of 30 While the term “ network ” includes not only a " physical
principal components to retain based on the difference network " and a “ content network ," as previously discussed ,
between the number of the eigen - values of the data corre the term “ network ” may also refer to " platform networks.”
lation matrix and the random matrix counterpart. The stan - For example , a Unified Network Platform (UNP ) comprises
dard Pearson correlation coefficients can be obtained using an open platform architecture that enables the convergence
the data in the method . 35 of network applications including data networking , security ,
In another example , a method can include one ormore of network infrastructure , network applications, and network
the following steps : receiving data of a predetermined time management.
series, determining auto regressive model coefficients for the FIG . 25 illustrates apparatus 2500 that comprises data
data of the time series, determining a scaled lag - 1 auto - processor 2501 and early warning system 2502 in accor
correlation of the coefficients , and outputting an indicator 40 dance with an aspect of the embodiments . For example ,
based on the scaled lag - 1 auto - correlation of the coefficients . processor 2501 obtains information about a real -world sys
The indicator can be positive when the scaled lag - 1 auto tem (which is typically external to apparatus 2500 ) such as
correlation of the coefficients is above a predetermined an underlying complex economic network . The information
threshold value. The indicator can be negative when the is gathered from information source 2530 , which may com
scaled lag - 1 auto - correlation of the coefficients is below a 45 prise a network of computers and /or databases that monitor
predetermined threshold value. The data of the predeter - the real-world system . As previously discussed, processor
mined time series can be Home Price Index data . A scaled 2501 processes the obtained information to determine if the
smoothed derivative from the scaled lag - 1 auto -correlation underlying real -world system exhibits characteristics are
of the coefficients can be determined. A second indicator can indicative that the real -world system is approaching a criti
be output based on a scaled smoothed derivative from the 50 cal threshold . In such a situation , early warning system 2502
scaled lag - 1 auto -correlation of the coefficients . generates an early warning signal to one or more computing
As previously discussed , many complex dynamical sys - devices 2520 - 2522 to inform devices 2520-2522 about the
temshave critical thresholds called tipping points where the potential instability of the system .
system abruptly shifts from one state to another. In global Those of skill in the art will appreciate that apparatus
finance, there is big concern about a sudden systemic decline 55 2500 may process and transform information from different
in values that can threaten the global economic system types of real-world systems in the realms of biology , epi
which , in turn ,may lead to global economic downturns. It is demiology , genetics, transportation , computer software, and
traditionally difficult to predict such critical transitions since economics .
the state of the system may show little change before the Referring to FIGS . 23A - D , early warning system 2502
tipping point is reached . However, it seems that certain 60 may determine from the scaled smoothed derivative that the
generic signs may take place in a wide class of systems as home price index (HPI ) for Las Vegas ( a geographical region
they approach a critical point. The dynamics of systems near as shown in FIG . 23B ) is approaching a critical threshold
a critical point have generic properties , regardless of differ - while the HPI for Cleveland (a geographical region as
ences in the details of each system . Critical thresholds for shown in FIG . 23C ) is not. In such a situation , early warning
such transitions correspond to bifurcations. In the so - called 65 system 2502 may generate an early warning signalhaving a
catastrophic bifurcation , once a threshold is exceeded , a positive warning signal component corresponding to Las
positive feedback mechanism pushes the system through a Vegas and a negative warning signal component for Cleve
US 9 ,741,048 B2
22
land . For example , the early warning signal may comprise a ous types of data as shown in FIGS. 23A - D ) may be sent
message having indicators for Las Vegas and Cleveland with from web server 2603 to device 2622 via communication
positive and negative values, respectively . server 2602 and external communication network 2611 .
An early warning signal may comprise one or more Detailed information may include information specific to a
messages that are sent to one or more computing devices . 5 particular geographical region associated with a registered
For example , as will be discussed , a firstmessage may notify device 2620 - 2622 as well as information about other geo
computing devices 2520 -2522 that a critical threshold is graphical regions that may be correlated with the particular
being approached while a second message has more detailed geographical region . For example , a potential instability for
information about the possible instability . Based on the Las Vegas may be detected so that detailed information may
information contained in the early warning signal, action via 10 be sent a device registered for a user responsible for the Las
a computing device may be initiated to address the possible Vegas geographical region . However, the detailed informa
instability . tion may include HPI information about Charlotte , Cleve
FIG . 26 illustrates computer system 2600 that supports land , and/ or Phoenix and /or other related information ( e . g .,
apparatus 2500 for processing information from information home liquidity ) so that the registered user for Las Vegas
source 2630 and generating an early warning signal pertain - 15 region can correlate the potential instability with other
ing to the stability /instability of a real-world system in geographical regions to ascertain whether there is a systemic
accordance with an aspect of the embodiments . Computer instability for the HPI.
system 2600 comprises servers 2601 , 2602 , and 2603. With With some embodiments , the detailed information in the
some embodiments , data server 2601 obtains information early warning signal may indicate whether the economy
about a real -world system from information source 2630 . As 20 associated with the instability is moving in a positive or
previously discussed , the real-world system may comprise negative direction and further indicate if there are any
an underlying economic network that exhibits structural particular elements driving the change . On a macro level this
attributes that may provide clues about economic instability could be certain industries or asset classes. On a micro level
or economic stability . Based on processing the information it could be a certain company ' s economic performance or
from information source 2630 , data server 2601 may detect 25 group of companies' economic performances . Also , eco
early -warning signs of the real-world system approaching a nomic instability may be affected by political instability ,
critical threshold where the system becomes unstable . If so , e .g ., government grid - lock , elections, and the like. As pre
data server 2601 generates an early warning signal. Notifi- viously discussed , principal component analysis or PCA
cation of the event may be sent to one or more users may be used to identify the particular “ drivers ” of the trends .
( associated with computing devices 2620 - 2622) via com - 30 This can indicate which subcomponents are accounting for
munication server 2603 . the change in state . For example, as previously discussed in
Communication server 2603 may send an event notifica - reference to FIG . 9 , largest ratios of two predetermined
tion to some or all of computing devices 2620 - 2622 through components may illustrate a particular economic driver. Two
one or more different communication media. For example, a subcomponents can be compared by calculating the ratio
notification may be sent via a short message service (SMS ) 35 between their first and second principal components . In one
and /or via an intranet/ internet. Referring to FIG . 26 , com - example , the first principal component can represent sys
munication server 2603 interacts with primary computing temic risk , for example , a recession or other event external
device 2620 via local communication network 2610 and to the data being studied . The second principal component
with remote computing devices 2621- 2622 via external can be non -systemic (idiosyncratic ) risk , which can be
communication network 2611 . With some embodiments, 40 specific to a particular subcomponent. Subcomponent 5 may
primary computing device 2620 may support a user when be the market driver because the ratio of the first principal
the user is working in an office environment. However, the component to the second principal component is the largest.
user may not be available in the office environment when the With some embodiments , an early warning application
real-world system is exhibiting characteristics of possible may be automatically launched at computing device 2620 ,
instability and immediate action must be initiated by the 45 2621 , or 2622 whenever an acknowledgment is generated at
user. In such a situation , the event notification may be sent the device in response to receiving an early warning noti
to remote computing device 2621 and/ or remote computing fication . Once launched , the early warning application may
device 2622 , which are registered to the user. As will be extract additional information about a potential instability
discussed , if an acknowledgement to the notification is from one or more messages contained in an early warning
received by communication server 2602 from one of remote 50 signal, appropriately filter the additional information , format
computing devices 2621 - 2622, more detailed information the processed information , and display the information at the
about the possible instability may be sent to the responding computing device . Moreover, the early warning application
computing device . For example , if communication server may be device - independent so that the application can
2602 receives an acknowledgement from device 2622 , an execute on a variety of different device types .
early warning application may be manually or automatically 55 Servers 2601 , 2602 , and 2603 may be combined on the
launched at device 2622 in which a communication connec - same physical machines, and retain separate virtual or
tion is established between web server 2603 and device 2622 logical addresses , or may reside on separate physical
via external communication network 2611. One or more machines . FIG . 26 illustrates just one example of a network
messages may be sent to device 2622 via the early warning architecture that may be used , and those of skill in the art
application through the communication connection to pro - 60 will appreciate that the specific network architecture and
vide additional information about the detected event. For data processing devices used may vary, and are secondary to
example , textual information identifying the corresponding the functionality that they provide, as further described
geographical area (e .g., Charlotte, Las Vegas, Cleveland , herein . For example , services provided by data server 2601,
and/ or Phoenix as shown in FIGS . 23A - D , respectively ) and web server 2602, and communication server 2603 may be
including quantitative information (e . g ., AR auto -correlation 65 supported by a single physical server.
and/ or smoothed derivative of the AR ) as well as graphical Each computing component ( servers 2601-2603 and com
information ( e.g ., plots showing relationshipsbetween vari puting devices 2620 - 2622 ) may be any type of known
US 9 ,741,048 B2
23 24
computer, server, or data processing device as discussed devices at block 2707 . For example , if device 2620 is not
herein , e . g ., as shown in FIG . 1 . active , system 2600 may send an early warning notification
FIG . 27 illustrates flow diagram 2700 for generating an to remote computing devices 2621- 2622 , which may be
early warning signal from processed information and portable devices having lesser capabilities than device 2620 .
responding to the early warning signal in accordance with an 5 If system 2600 receives an acknowledgement from a
aspect of the embodiments. With some embodiments, flow registered device , as detected at block 2708 , the early
diagram 2700 may be performed by one or more processors warning application may be automatically launched at the
executing computer readable instructions from one or more responding device at block 2709 . In addition , based on the
non - transitory computer -readable media spanning data display capabilities of the device , some or all of the currently
server 2601 , communication server 2602 , and/ or web server 10 displayed framed graphical elements ( sometimes referred as
2603 . windows) may be modified , resized , and / or minimized to
Referring to FIGS . 26 and 27 , system 2600 obtains more effectively display the detailed content of the early
information about a real-world system from information warning signal. For example , a currently displayed window
source 2630 . At block 2701 , system 2600 determines may be obscuring a window depicting a graphical represen
whether there is a potential instability within the real -world 15 tation about the HPI (FIGS. 23A - D ). If so , the currently
system . For example , system 2600 may detect the dynamics displayed window may be modified ormoved to reduce the
of a system approaching a tipping point as indicated by obscurity .
slower recovery from perturbations , increased autocorrela If system 2600 does not receive an acknowledgement to
tion , and/ or increased variance. If not, system 2600 contin the early warning notification from any of the registered
ues to monitor /analyze information from source 2630 . 20 devices , system 2600 may initiate an alternative early warn
If a potential instability is detected at block 2701, system ing notification . A notification may be sent to alternative
2600 determines the assigned registered devices for the devices ( e . g ., devices registered for a different geographical
potential instability at block 2703 , where registered devices region ) or via e -mail sent to registered users at block 2710 .
may be associated with one or more users. For example , the FIG . 28 illustrates flow diagram 2800 for generating an
real-world system may exhibit instability at some or all of 25 early warning signal from processed information and
the associated geographical regions (entities ), where regis - responding to the early warning signal in accordance with an
tered devices may be assigned to a particular geographical aspect of the embodiments . Referring to FIG . 27 , blocks
region . Based on the determined geographical region , one or 2801- 2808 and 2810 are similar to blocks 2701- 2708 and
more registered computing devices 2620 - 2622 ( e . g ., desk - 2710 , respectively . However , if system 2600 receives an
tops, laptops, notebooks, mobile telephones or smartphones 30 acknowledgment from a registered remote device at block
with applications and other functionality , handheld device 2808 , the early warning application is launched at the
with Wi-Fi or other wireless enabled tablets, tablet comput primary computing device (e.g., the user 's computer located
ers, PDAs, and the like ), may be registered for the geo on the office environment) to retrieve detailed information
graphical region on behalf of the user or users responsible from web server 2603. A communication connection may
for the identified geographical region . 35 then be automatically established between the remote device
Devices 2620 - 2622 may be registered according to dif and the primary device via communication server 2602 .
ferent criteria , e .g., associated with specific geographical Consequently, a user may obtain detailed information about
regions as discussed above . Moreover, embodiments may the potential instability from the primary device via the
consider an identified economic system spanning geogra communication connection at the remote device .
phies, industries , and / or other economic neighborhoods 40 Referring to FIGS. 27 and 28 , while early warning
when determining criteria for the registration of devices notifications may target specific devices, embodiments may
2620 - 2622 . target access - id applications, e . g ., social media user
System 2600 determines whether any of registered com - accounts , domain - specific registered users , and the like . The
puting devices 2620 - 2622 are active at block 2704 . If so , additional layer of notification may not have very specific
system 2600 generates an early warning notification to the 45 details of the alert other than notify the recipient of the
primary device at block 2705 . For example , a user may be message to check the devices where the recipient chooses to
active at primary computing device 26er20 (which may be receive the secured alert .
registered as the primary device since it operates in a more FIG . 29 illustrates flow diagram 2900 for sending an early
secure office environment). With some embodiments , the warning signal to a plurality of computing devices in accor
notification may be a short text message identifying the 50 dance with an aspect of the embodiments . As with flow
possible instability and the associated geographical region . diagrams 2700 and 2800 , system 2600 generates an early
At block 2706 , an early warning application may be warning notification to all registered computing devices
automatically launched at the primary device after the 2620- 2622 at block 2901 . When the first device acknowl
notification is received at the computing device . The early edges the notification at block 2902 , system 2600 provides
warning application may obtain detailed information about 55 detailed early warning information to the first responding
the detected event. For example , a communication connec - device . At block 2903 , an invite message may be generated
tion may be established between web server 2603 and device and sent to another device so that the user of the first device
2620 via communication server 2602 to provide quantitative can collaborate with another user of the other device , where
information ( e . g ., AR auto - correlation and/ or smoothed the other device may or may not be registered . When
derivative of the AR ) as well as graphical information . The 60 communication is established between the first device and
additional information may be formatted in one or more the other device , the detailed information may be shared
webpages by web server 2603 and displayed at computing with the other device . Furthermore , a subsequent invite
device 2620 via the early warning application and/ or web massage may be generated by the first device , other device ,
browser. or system 2600 to include a further device during the
However, if system 2600 determines that none of the 65 collaboration.
registered devices is active at block 2704 , system 2600 FIG . 30 illustrates flow diagram 3000 for displaying
generates an early warning notification to all registered information derived from an early warning signal at a
US 9 ,741,048 B2
25 26
computing device in accordance with an aspect of the separate communication connection may be established
embodiments . At block 3001, system 2600 receives an among the decision -makers (e.g., a subset of the partici
acknowledgment to an early warning notification from one pants ) so that messaging/ conversation among the decision
of the registered devices. System 2600 identifies the device makers is not available to other participants. For example ,
type of the responding device at block 3002 and determines 5 some of the participants in a first communication connect
the display characteristics based on the identified device may not have the authority to make a decision about an
type at blocks 3003 - 3005 , which corresponds to a desktop , instability . If so , a second communication connection may
smartphone , and tablet, respectively . However, additional or be established among the decision -makers, where the con
different types of devices may be supported . Based on the versation /messaging of the second communication connec
obtained display characteristics , web server 2603 may con - 10 tion is not available to the first communication connection .
sider the viewable image size , resolution , and aspect ratio to Although the subject matter has been described in lan
determine the amount of detailed information provided to guage specific to structural features and / or methodological
the device. Web server 2603 may partition the detailed acts, it is to be understood that the subject matter defined in
information into a plurality of information components and the appended claims is not necessarily limited to the specific
provide selected information components of the detailed 15 features or acts described above . Rather , the specific features
information to the device based on the display characteris - and acts described above are disclosed as example forms of
tics at block 3006 . For example , different information com - implementing the claims.
ponents may comprise different graphical elements having What is claimed is :
varying degrees of information and/ or resolution . 1 . An apparatus comprising :
Information components of the detailed may span a 20 an early warning system ;
variety of forms including textual information and visual an electronic processing device ; and
izations that may convey the systematic uncertainty associ a memory device storing computer readable instructions
ated with the potential instability . Visualizations may be that, when executed by the electronic processing
presented in different forms, including graphs, network device , cause the apparatus to :
diagrams, correlograms, and the like. The information com - 25 receive economic data for a plurality of time series ,
ponents may be pushed from computer system 2600 or may wherein each time series of the plurality of time
be pulled by computing devices 2620 - 2622 . When pushed , series corresponds to a different geographical entity ;
computer system 2600 may obtain the display characteristics determine auto regressive model coefficients for the
of a responding computing device 2620 - 2622 , and send economic data of the plurality of time series ;
information components that the responding computing 30 determine , for a first time series of the plurality , a first
device can effectively display. When pulled , an early warn scaled lag - 1 auto -correlation of the coefficients;
ing application may support a dashboard from which a user determine, for the first time series , a first scaled
can select available information components that are to be smoothed derivative from the first scaled lag - 1 auto
displayed on the computing device . correlation of the coefficients ;
When early warning system 2502 has notified registered 35 output, for the first member, a first indicator based on
devices about the potential instability, early warning system the first scaled smoothed derivative;
2502 may further insure that the computing devices of generate , by the early warning system based on at least
appropriate individuals are connected via a communication the first indicator for the first time series of the
system ( for example , by generating invite messages as plurality of time series , an early warning signal,
discussed above ) so that a decision about the instability can 40 wherein the first time series corresponds only to a
be reached by a participant or collectively by a group of first geographical entity and wherein the early warn
participants . With some embodiments , a data structure ( e.g., ing signal comprises an early warning notification
role -based matrix ) may map a level of responsibility for and detailed information about a potential instability
each of the participants and with identities of associated of the economic data ;
computing devices for those who are needed in the decision - 45 determine a set of registered computing devices asso
making process . The data structure may be based on an ciated with the first geographical entity , wherein the
organization chart, where typically the higher one 's level in set includes a first device and a second device ;
organization chart, the greater the authority one has for send the early warning notification about the first
making a decision . The data structure may relate the indi geographical entity to the first device and second
vidual, the level of responsibility , and the device identity of 50 device ;
the individual. The data structure may reside within system initiate launching of an early warning application at the
2600 or may reside in a separate computing entity ( e .g ., a first device ; and
separate server ). send the detailed information about the potential insta
With some embodiments, early warning system 2502 may bility to the first device via the early warning appli
interact with an emergency notification system that may 55 cation .
have been established to reach personnel in the event of 2 . The apparatus of claim 1 wherein the first indicator is
natural and non -natural events . The emergency notification positive when the first scaled smoothed derivative is above
system may follow a contact tree to contact employees via a predetermined threshold value .
telephone calls , e -mails, and so forth . The early warning 3 . The apparatus of claim 1 wherein the first indicator is
signal from early warning system 2502 may induce the early 60 negative when the first scaled smoothed derivative is below
notification system to function as an added contact mecha - a predetermined threshold value .
nism . Moreover, the contact mechanism may include deliv - 4 . The apparatus of claim 1 wherein the economic data for
ering early warning signals through social media and soft- the plurality of time series comprises Home Price Index
ware -defined networks . data .
In some situations, only some of the participants of the 65 5 . The apparatus of claim 1, wherein thememory device
communication connection may have the authority to make stores computer readable instructions that, when executed
a decision about the instability . With some embodiments a by the electronic processing device , further cause the appa
US 9, 741, 048 B2
27 28
ratus to output a second indicator based on the first scaled ing signal comprises an early warning notification and
lag - 1 auto -correlation of the coefficients and the early warn detailed information about a potential instability of the
ing signal is based on the first indicator and the second economic data ;
indicator. processing the early warning signal to determine whether
6 . The apparatus of claim 1 , wherein the memory device 5 there is an instability of the economic data for the first
stores computer readable instructions that, when executed geographical entity ;
by the electronic processing device , further cause the appa determining a set of registered computing devices asso
ratus to : ciated with the first geographical entity, wherein the set
select a selected time series from the plurality of time 10 includes a first device and a second device;
series; and sending the early warning notification about the first
generate the early warning signal from the first indicator geographical entity to the first device and second
of the selected time series . device ;
7 . The apparatus of claim 1 , wherein the memory device initiating launching of an early warning application at the
first device ; and
stores computer readable instructions that, when
When executed
executed 15 sending the detailed information about the potential insta
by the electronic processing device , further cause the appa bility to the first device via the early warning applica
ratus to : tion .
output a second indicator based on the first scaled lag - 1 12. The method of claim 11 wherein the first indicator is
auto -correlation of the coefficients ; positive when the scaled lag - 1 auto - correlation of the coef
select a selected indicator from the first and second 20 ficients is above a predetermined threshold value .
indicators; and 13 . The method of claim 11 wherein the first indicator is
generate the early warning signal from the selected indi- negative when the scaled lag - 1 auto - correlation of the coef
cator of the selected time series. ficients is below a predetermined threshold value.
8 . The apparatus of claim 1, wherein the plurality of time 14. Themethod of claim 11 wherein the economic data for
series comprises the first time series and a second time 25 the time series comprises Home Price Index data .
series , the second time series corresponds to a second 15 . The method of claim 11 further comprising:
geographical region , and the early warning signal is appli outputting a second indicator based on the scaled
cable only for the first time series . smoothed derivative from the scaled lag- 1 auto - corre
9 . The apparatus of claim 1 , wherein the memory device lation of the coefficients.
stores computer readable instructions that when executed 30 16 . One or more non -transitory computer -readable media
by the electronic processing device , further cause the appa having instructions stored thereon that, when executed ,
cause at least one computing device to :
ratus to : receive economic data for a plurality of a time series,
send an invite message to another device, wherein the wherein each time series of the plurality of time series
invite message invites the other device to communicate 35 corresponds to a different geographical entity ;
with the first device. determine , for a first time series of the plurality , auto
10 . The apparatus of claim 1 , wherein thememory device regressive model coefficients for the economic data of
stores computer readable instructions that, when executed the plurality of time series , wherein the first time series
by the electronic processing device , further cause the appa corresponds to a first geographical entity ;
ratus to : 40 determine, for the first member, a scaled lag - 1 auto
when the acknowledgment to the warning notification is correlation of the coefficients ;
received from the first device , identify a device type of determine a scaled smoothed derivative from the scaled
the first device; lag - 1 auto - correlation of the coefficients ;
determine a first display characteristic of the first device output, for the first time series , a first indicator based on
from the device type; and 45 the scaled smoothed derivative , wherein the first indi
send a subset of information components to the first cator is applicable only to the first geographical entity ;
device based on the first display characteristic , wherein generate , based on the first indicator for the first time
the detailed information comprises a plurality of infor series, an early warning signal to an early warning
mation components . system via a computer network , wherein the first times
11 . A method comprising: 50 series corresponds to a first geographical entity ;
receiving economic data for a plurality of time series, determine , from the early warning signal, whether there is
wherein each time series of the plurality of time series an instability of the economic data for the first geo
corresponds to a different geographical entity ; graphical entity , wherein the early warning signal com
determining, for a first time series of the plurality, auto prises an early warning notification and detailed infor
regressive model coefficients for the economic data of 55 mation about a potential instability of the economic
the plurality of time series , wherein the first time series data ;
corresponds to a first geographical entity ; determine a set of registered computing devices associ
determining , for the first member, a scaled lag - 1 auto ated with the first geographical entity , wherein the set
correlation of the coefficients; includes a first device and a second device ;
determining a scaled smoothed derivative from the scaled 60 send the early warning notification about the first geo
lag - 1 auto -correlation of the coefficients ; graphical entity to the first device and second device ;
outputting , for the first member, a first indicator based on initiate a launching of an early warning application at the
the scaled smoothed derivative , wherein the first indi first device ; and
cator is applicable only to the first geographical entity ; send the detailed information about the potential instabil
generating, based on the first indicator for the first time 65 ity to the first device via the early warning application .
series of the plurality of time series , an early warning 17 . The one or more non -transitory computer -readable
signal via a computer network , wherein the early warn - media of claim 16 wherein the first indicator is positive
US 9 ,741,048 B2
29 30
when the scaled lag - 1 auto -correlation of the coefficients is cessing device, further cause the apparatus to establish
above a predetermined threshold value. a first communication connection with the plurality of
18 . The one or more non -transitory computer -readable devices based on the data structure .
media of claim 16 wherein the first indicator is negative 22 . The apparatus of claim 21 wherein the memory device
when the scaled lag - 1 auto - correlation of the coefficients is 5 stores computer readable instructions that, when executed
below a predetermined threshold value. by the electronic processing device , further cause the appa
19 . The one or more non -transitory computer -readable ratus to :
media of claim 16 wherein the economic data for the time establish a second communication connection with a
series comprises Home Price Index data . subset of the plurality of devices based on the data
20 . The one or more non -transitory computer - readable 10 structure ; and
media of claim 16 wherein the at least one computing device prevent content of the second communication connection
performs: from being conveyed to the first communication con
outputting a second indicator based on the scaled nection .
smoothed derivative from the scaled lag - 1 auto -corre 23 . The apparatus of claim 1 wherein the memory device
lation of the coefficients. 15 stores computer readable instructions that, when executed
21 . The apparatus of claim 1 further comprises : by the electronic processing device, further cause the appa
a data structure , wherein the data structure maps decision ratus to :
making roles of individuals with associated device send an alert indication to a designated recipient through
identities, wherein the device identities identify a plu an access- id application , wherein the recipient is
rality of devices ; and 20 informed to check at least one of the set of registered
wherein the memory device stores computer readable computing devices.
instructions that, when executed by the electronic pro * * * *

United States Patent: (45) Date of Patent: Aug - 22, 2017

Uploaded by

Copyright:

Available Formats

United States Patent: (45) Date of Patent: Aug - 22, 2017

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

United States Patent: (45) Date of Patent: Aug - 22, 2017

Uploaded by

Copyright:

Available Formats

MAIMOTOMONTANT UNUNIT HINDI HII US009741048B2

(12) United States Patent (10) Patent No.: US 9 ,741,048 B2

2002 Build the Standard Pearson Correlation

Build the Matrix Whose Elements are

2008 Set the Lowest Value of Correlation

Using Chi -Square Test, a Sharp Phase

( 56 ) References Cited M . Breakspear, S . Heitmann , A . Daffertshofer. “Generative models

105 ( WAN 101 - 109E

LAN (Opt.)| - 109F

Processor Network Interface

Macro -Micro Complexity Pyramid

502 504 506 508

+ - 1. tin vit viintritt vin in v . . . . . ... .. . .+ 1. . . -. + + + + .' . . . " . . .1 . .

802 804 806 808

Apply Take Find

HPIbyStates VirginiaP-.NewYorkNevadaMHawaiiCeArizona|aonssnaycelhcuvtsaienctiuast WCaslhifnogrtnoina

ofHPIbySCPC1Eiotgneavtnceosr L1FCPoramipncdteognarsl 2006/31/ 10

1902 1916 1918

1904 ) 1914 ) 1920 )

Start 2004 TakeData of Multiele TimeSeries of

2002 Build the Standard Pearson Correlation

Build Nearest Neighbor Spacing

Using Chi- Square Test, a Sharp Phase

2012 Process Ends Once Phase Transition

Stan 2000, Take

2026 Build Relevant RMT Where the Density

2028 Compare the Statistics of the Observed

Phase SDistrbu on Ehnatropny -pOardmetr I

Phase SDistrbu on Ehnatropny -pOardmetr

Phase EDistrbu on Snhtaronpy -pOardmetr

??? ?? ??? ?? ? ??? ??? - -

PInshta eous wwwwhat www

Phase EDistrbu on Snhtaronpy -pOardmetr

Phase Distrbuon 0.8Order-parameter

2404 Compute the Scaled LAG - 1

COMPUTING COMPUTING COMPUTING

DATA SERVER COMMUNCATION

2610 LOCAL COMMUNICATION

WEB SERVER onwwwwwww

GENERATE EARLY WARNING 2804

2810 PERFORM ALTERNATIVE

LAUNCH EARLY WARNING

GENERATE EARLY WARNING

FROM FIRST DEVICE

PROVIDE EARLY WARNING

3001 RECEIVE ACKNOWLEDEMENT

3003 SMOTOn 3004

DISPLAY SUBSET OF FIGURE 30

12 = b= =y(b-a) Var (yn+1) = E( ym ) –4 ? = 2

You might also like