The current issue and full text archive of this journal is available on Emerald Insight at:
https://www.emerald.com/insight/1753-8270.htm
Forecasting house prices in Iran
using GMDH
Forecasting
house prices
Behrooz Nazemi and Mohsen Rafiean
Faculty of Art and Architecture, Yazd University, Yazd, Iran
Abstract
Purpose – An accurate predictive model for forecasting urban housing price in Isfahan can be useful for
Received 29 May 2020
Revised 4 July 2020
Accepted 13 July 2020
sellers and owners to take more appropriate actions about housing supplying. Also, it can help urban housing
planners and policymakers in managing of the housing market and preventing an urban housing crisis in
Isfahan. The purpose of this paper is forecasting housing price in Isfahan city of Iran until 2022 using group
method of data handling (GMDH).
Design/methodology/approach – This paper presents an accurate predictive model by applying the
GMDH algorithm by using GMDH-Shell software for forecasting housing price in municipal boroughs of
Isfahan city till the second half of 2022 based on creating time series and existing data. Alongside housing
price, some other affecting factors have been also considered to control the forecasting process and make it
more accurate. Furthermore, this research shows the housing price changes of boroughs on map using
ArcMap.
Findings – Based on forecasting results, the housing price will increase at all boroughs of Isfahan till second
half of the year 2022. Amongst them, Borough 15 will have the highest percentage of the price increasing
(28.27%) to year 2022 and Borough 6 will have the lowest percentage of the price increasing (8.34%) to the
year 2022. About ranking of the boroughs in terms of housing price, Borough number 6 and 3 will keep their
current position at the top and Borough number 15 will stay at the bottom.
Research limitations/implications – In this research, just few factors have been selected alongside
housing price to control the forecasting process owing to limitation of reliable data availability about affecting
factors.
Originality/value – The most remarkable point of this paper is reaching to a mathematical formula that
can accurately forecast housing price in Isfahan city which has been rarely investigated in former studies,
especially in simplified form. The technique used in this paper to forecast housing price in Isfahan city of Iran
can be useful for other cities too.
Keywords Forecasting, GIS, Algorithm, Housing price, Analyzing, GMDH
Paper type Research paper
1. Introduction
Nowadays, housing is not considered just as a roof over one’s head, and it has a crucial role
in the urban economy and sustainable development (Golubchikov and Badyina, 2012). As an
irreplaceable, consumable, marketable, producible and durable commodity (Ruonavaara,
2018), price of housing is determined based on supply and demand in the market (Rahadi
et al., 2015). Forecasting the future trend of housing price is very necessary for all decisionmaker and takers to balance the housing market and react to the changes more appropriate
(Liu and Liu, 2019). In today’s financial markets, including the housing market, analyzing
existing data can be very helpful for predicting price movements (Tsantekidis et al., 2017).
Choosing the suitable forecasting method depends significantly on data availability, kind of
data, case study, the context of the research, financial and human resources and expected
accuracy (Hyndman and Athanasopoulos, 2018). In previous studies different kinds of
methods have been used to forecast housing price, some of the most important of them are
auto regressive integrated moving average, random forest, fuzzy logic, support vector
International Journal of Housing
Markets and Analysis
© Emerald Publishing Limited
1753-8270
DOI 10.1108/IJHMA-05-2020-0067
IJHMA
machine, Bayesian approach, geographical detector technique, satellite imagery, spatial
hedonic models, co-regionalized models, panel data-based approach, factor analysis, hedonic
model and artificial neural networks (ANNs) (Li et al., 2017; Khalafallah, 2008; Morano et al.,
2019; Stamou et al., 2017; Rossini, 2000; Morano et al., 2015; Wei and Cao, 2017; Bork and
Møller, 2015; Kouwenberg and Zwinkels, 2014; Gu et al., 2011; Azadeh et al., 2012; Park and
Bae, 2015; Ahn et al., 2012; Hadavandi et al., 2011; Waltl, 2016).
In recent years, machine learning and ANN methods have been improved
considerably in order to forecast housing price because of their advantages over
traditional methods (Xu and Li, 2020; Dong, 2020). For example, Rui Liu and Lu Liu in the
year 2019 predicted housing price based on long short-term memory incorporating
modified genetic algorithm in their article that has shown more acceptable results over
traditional techniques (Liu and Liu, 2019). Most of the previous studies about forecasting
housing price using machine learning and artificial neural network methods have not
presented a simplified and obvious predictive mathematical formula. So, one of the main
goals of this article is using group method of data handling or group method of data
handling (GMDH)-type artificial neural network for making an accurate predictive model
in order to forecast housing price in different municipal boroughs of Isfahan (which is
divided by the municipality of Isfahan) [1] and representing the relevant mathematical
formula for each borough. Another main goal is showing the housing price changes of
boroughs on map utilizing one of the most important geographic information system
(GIS) software that known as ArcMap. Isfahan is one of the most important main cities in
Iran located in 32°38’ 30” N latitude and 51°38’ 40” E longitude, about 340 km south of
Tehran and the capital of Isfahan Province. The population of Isfahan city has been
increased rapidly and considerably, mostly because of immigration, such as that based
on last national census statistics it has grown more than 435% (more than 4 times) in
comparison with 50 years ago (Assari et al., 2017) [1] [2]. Facility inequalities in regional
scale, owning large and maternal industries, tourism and other important economics
determinants have caused by knowing Isfahan as an attractive destination in terms of
immigration (Assari et al., 2017). The certain consequence of this population growing has
been increasing demand for housing in Isfahan city. Owing to appropriate response to the
increasing demand, an accurate predictive model for forecasting urban housing price in
Isfahan can be useful for developers, sellers, and owners to make more precise decisions
and take more appropriate actions about housing supplying. Also, it can help urban
housing planners and policy makers in managing of the housing market, price changes
and prevent an urban housing crisis in Isfahan city of Iran.
2. Material and methods
2.1 Data
In this research, data of housing price in different boroughs of Isfahan city have been
collected from 1995 to 2019 for every six months. In order to increase accuracy and
controlling process of the forecasting, 5 other factors have been selected alongside housing
price (Rahadi et al., 2015). The value of these factors varies from one borough to another
borough. The selected factors are average rent of 1 m2 residential unit (RP), average price of
1 m2 land (LP), population (P), average total costs of a household (CH) and average income of
a household (IH). Relevant data to the research variables have been collected from Iran
national statistics bureau (www.amar.org.ir), central bank of Iran (www.cbi.ir), World Bank
(www.worldbank.org) and Isfahan municipality (www.isfahan.ir).
2.2 Group method of data handling algorithm
The group method of data handling (GMDH) algorithm is very useful for modeling work like
mathematics and computer ones (Onwubolu, 2014). This algorithm has been introduced by a
scientist named Ivakhnenko as a multivariate analysis approach to solving complicated
system problems (Stepashko et al., 2017). The basis of the GMDH approach is to create an
analytical function in a quadratic node-based transfer network (Moroz, 2016). The GMDH
network makes a connection between input and output data, that is indicated to the Volterra
function series or the Kolmogorov–Gabor polynomial function (Yang et al., 2018):
y ¼ a0 þ
m
X
i¼1
ai xi þ
m X
m
X
aij xi xj þ
i¼1 j¼1
XXX
aijk xi xj xk þ . . .
(1)
Such mathematical expression can be produced by a set of partial quadratic polynomials,
including only two variables (neurons) as follows (Kordnaeij et al., 2015):
y ¼ Gðxi ; xj Þ ¼ a0 þ a1 xi þ a2 xj þ a3 xi xj þ a4 x2i þ a5 x2j
(2)
By this means, in a network of related neurons the partial quadratic description is
recursively used to create the general mathematical relation between giving inputs and
output in Equation (Kordnaeij et al., 2015). The coefficients ai in equation (B) is classified by
regression methods to minimize the difference between the actual output y and the assigned
one y for each pair of input parameters xi, x j (Shaghaghi et al., 2017). In fact, a hierarchy of
polynomials is built using the quadratic form produced in equation (A) whose constants are
achieved by least-squares logic, then the constants of every quadratic function Gi are
obtained to optimally fit the output throughout the whole set of output–input data pairs, in
the form of Ahmadi et al. (2015):
M
X
E¼
ðyi
i
M
Gi Þ
2
! min
(3)
Basically, all the probabilities of two independent variables out of the total n input variables
are taken in order to create the regression in the GMDH algorithm, that best fits the
dependent observations (yi, i = 1, 2, . . ., M) in terms of least squares logic (Ahmadi et al.,
2019). For every row of M data triples, the following matrix equation can be readily achieved
as Aa = Y, where a is the vector of unknown coefficients of the quadratic polynomial in the
equation (B) (Shahsavar et al., 2019).
a ¼ fa0 ; a1 ; a2 ; a3 ; a4 ; a5 g
(4)
Here Y is the vector of the output’s value of observation, as shown in equation (5)
(Shahsavar et al., 2019):
Y ¼ fy1 ; y2 ; y3 ; . . . ; yM gT
(5)
The least squares technique from multiple regression analysis leads to the solution of the
normal equations in the form of Ghanadzadeh et al. (2014):
Forecasting
house prices
IJHMA
0
1
B
B1
B
A ¼ B.
B ..
@
1
x1p
x1q
x2p x2q
..
..
.
.
xMp xMq
x1p x1q
x21p
x2p x2q
x22p
..
..
.
.
xMp xMq
x2Mp
1
a ¼ ð AT AÞ AT Y
x21q
x22q
..
.
x2Mq
1
C
C
C
C
C
A
(6)
(7)
which assignes the vector of the best coefficients of the quadratic equation (B) for the entire
set of data triples, this procedure is repeated for each neuron of the next hidden layer based
on the connectivity topology of the network (Varahrami, 2012). In the training of each layer,
the neuron candidates are represented according to all the feasible combinations of input
variables, then these neuron candidates are screened automatically regarding their ability to
estimate the target variable and only those neurons having good predication powers (meets
the criterion) are fed forward for the training of the next layer, and the rest is ignored (Jia
et al., 2018).
Special behavior properties of GMDH make it a perfect option to use in fields such as
knowledge exploration, data mining, predicting, different types of modelling, optimization
and pattern finding (Onwubolu, 2014; Moroz, 2016).
In order to forecast housing price in Isfahan of Iran in this research, GMDH Shell
software that is an advanced analytics suite that can represent complicated time series
predicting, regression, classification, clustering and curve fitting tools based on GMDH
algorithm was used [3]. Some of the most important characteristics of GMDH Shell software
are assigning model structure automatically, preventing overfitting and performing
acceptable with very small data sets and being quick even with more than 1,000 input
variables. Some of the greatest global corporations like NASA, Xcel Energy, Cork Supply,
Fike, Marketstar and Honeywell are customer of GMDH Shell software3.
2.3 Parameters setting
After inserting data from Excel software to GMDH Shell software in data explorer section,
the corresponding parameters have been set in the solver section of the software where the
values of considered parameters have a significant influence on the performance of GMDH
Shell to make a predictive model. The parameters are set based on trial and error method to
obtaining the best result as follows:
Reorder observations parameter is utilized to obtain uniform statistical properties of
training and testing samples and to make them equally informative, this technique works
perfectly for almost all problem types including time series models. This parameter has been
set on pseudorandom function that is a deterministic function of a key and an input that is
indistinguishable from a truly random function of the input.
Validation strategy parameter is utilized to choose a method for model validation and
sorting out. In this case the training/testing strategy, that splits the dataset into two parts
and utilizes the training part to find model coefficients and implements the testing part to
compare and choose a set of the best models, has been assigned as validation strategy. In
order to obtain best results, train/test ratio differs from one borough to another one.
For the Validation criterion parameter, that defines the model selection criterion, the Root
Mean Square Error (RMSE) balance that is focused on the difference between training and
testing sample RMSE, has been selected.
The neuron function parameter sets the type of the internal function of neurons. The
neurons are active, and each neuron can drop some of the function terms to increase overall
predictive power of the model. The selected function for this parameter is:
Forecasting
house prices
a þ xi þ xi :xj þ . . .
The Max. Number of layers parameter values that sets the upper limit for the number of
network layers created by the algorithm is 10.
The selected value of the Initial layer width parameter that defines how many neurons
are added to the set of inputs at each new layer is 50.
2.4 Geographic Information System
GIS is a system that created to analyze, checking, manage and show spatial or geographic
data [4]. Generally, GIS uses two kinds of data: spatial data and attribute data
(Tomaszewski, 2014). Spatial data describe the absolute and relative locations of spatial
(geographic) entities, while attributing data indicate to characteristics of spatial entries
(Ehrgott et al., 2010). In this research Arc Map software as one of the most important parts of
Esri’s ArcGIS suite of geospatial processing programs that is useful to display, modify,
produce and analyze geospatial data [5] is regarded to show forecast results on the map.
ArcMap makes user able to discover data within a set of data, symbolize properties and
produce maps [6].
3. Results
In this article, only model details of Borough 1 is shown as an example. For other boroughs,
just final forecasting results are presented on relevant maps using the same technique. The
obtained results for Borough number 1 are shown in Table 1, Figure 1 and Table 2.
About Table 2, errors have been calculated based on differences of existing data and
predictions (till 2019) (Figure 2).
The first 27 of existing data were used to recognize the pattern of created time series by
the model. To forecast housing price till 2022, the next 6 observations after observation
number 54 were considered. Predictions have been achieved based on below equations
(according to GMDH algorithm).
0
HP ½t ¼ þ0:940701*a þ 0:0615838* b
2:33448
ID
Actual value
Predictions
ID
Actual value
Predictions
ID
Actual value
Predictions
28
29
30
31
32
33
34
35
36
37
38
360.09
409.75
440.08
475.02
528.13
523.08
552.16
548.9
587.19
599.63
650.89
385.19
398.01
403.05
440.81
492.21
538.27
566.01
599.23
584.69
591.81
651.19
39
40
41
42
43
44
45
46
47
48
49
712.70
875.03
1000.02
1080.1
1076.3
1115.25
1104.12
1180.1
1245.01
1280.02
1440.3
769.94
866.81
958.37
997.66
1079.56
1126.92
1195.48
1179.14
1263.67
1291.17
1476.04
50
51
52
53
54
55
56
57
58
59
60
1550.81
1786.79
1900.91
2214.23
2350.11
–
–
–
–
–
–
1531.67
1765.63
1877.47
2204.67
2351.28
2338.13
2504.01
2459.51
2659.31
2852.97
2798.66
Table 1.
Actual values and
predictions
IJHMA
Figure 1.
Forecasting housing
price for Borough 1
Index
Table 2.
Obtained errors
Figure 2.
Residuals
Mean absolute percentage error (MAPE)
Root Mean Square Error (RMSE)
Mean absolute error (MAE)
Residual sum
Value
3.26%
4.98
24.84
0.77
a¼
b ¼ þ0:0792739*RP ½t
0:0254963*v þ 1:02278*Xþ2:77809
6*P ½t
X¼
d ¼
6
5919:2*RP ½t
6
Forecasting
house prices
0:0315856*P ½t
6 þ 3076:71
0:491843*ˆ þ 1:4904*d þ1:47333
0:000407005*LP ½t
6* w þ 1:24173* w
68:9823
w ¼ þ0:0000697197*s *l þ 0:231829*s þ 0:608187*l þ 70:452
l ¼ þ0:00000169618*CH ½t
6* m þ 0:0117867*CH ½t
m ¼ þ0:00445097*CH ½t
ˆ¼
6
0:000375373*LP ½t
6
2:93543* m þ 543:806
0:0157098*P ½t
6 þ 1092:93
6*u þ 1:22365*u
64:5554
u ¼ þ0:0000843966* c *s þ 0:548268* c þ 0:255702*s þ 87:6753
s ¼ þ740:362*RP ½t
c ¼ þ0:00000000287695*CH ½t
0:00223708*IH ½t
v ¼ þ0:000504971*LP ½t
6
1:77642*LP ½t
6*IH ½t
6 þ 199:15
6 þ 0:00475612*CH ½t
6
6 þ 178:369
6*P ½t
6
37:7863*LP ½t
6
0:0582863*P ½t
6
þ 5199:12
In above
equations:
0
HP ½t = Housing price prediction
[t]
= Number of observation
*
= Multiply sign
For example, to forecasting housing price about observation number 60 (2022 year):
t
= 60; and
t 6 = 54
Value of considered factors for observation number 54 are as follows: (Table 3).
When these values have been put in sub equations:
Factor
P
LP
RP
CH
IH
Value
80986
1100
4.413
764925.67
830675.47
Table 3.
Value of considered
factors
a ¼ 2798:877
IJHMA
b ¼ 2729:48
Finally, the value of prediction for observation number 60 was achieved using a and b in
main equation. The final predicted value for observation number 60 is 2798.66 $.
Results of housing price prediction for all boroughs of Isfahan city have been presented
in Table 4.
In Figure 3, housing price in the year 2019 has been shown for each borough of Isfahan
city that assigned by numbers (1, 2, 3. . ., 15) at relevant borders. Each borough has its own
color on a map based on average housing price in there. The ranges of housing price on the
map (Figure 3) have been selected based on geometrical interval classification method of Arc
map software.
The rate of housing price changes (per cent) based on forecasted values to the year 2022
for Isfahan boroughs that obtained by the used model, have been presented in Figure 4.
4. Discussion
To avoid over-increasing the pages of the article, the process of housing price forecasting
using considered technique has been presented only for Borough 1 of Isfahan city as an
example. To measure the accuracy of housing price forecasting in Borough 1 of Isfahan city
in comparison to existing data, interpretation of MAPE, RMSE and MAE is essential (Liu
and Liu, 2019; Kordnaeij et al., 2015; Ahmadi et al., 2019; Ganti et al., 2018; Chen et al., 2017;
Alfiyatin et al., 2017). To interpret of MAPE value, generally less than 10% can be
acceptable (Borzemski and Wojtkiewicz, 2011). According to the obtained results, MAPE
value of this work for the Borough number one of Isfahan is 3.26%. As shown in table 2,
RMSE value is 4.98 in this research. Regarding scale of used data, RMSE value can be
approved (Balouchi et al., 2015; Zhang, 2015). Considering used data range (maximum and
minimum values), obtained MAE value, that is 24.84, can be confirmed as an acceptable
value (Roy et al., 2016; Kim et al., 2017). For future housing price forecasting (from 2019 to
Name of
borough
Table 4.
Results of housing
price forecasting for
all boroughs
Borough 1
Borough 2
Borough 3
Borough 4
Borough 5
Borough 6
Borough 7
Borough 8
Borough 9
Borough 10
Borough 11
Borough 12
Borough 13
Borough 14
Borough 15
Housing price in first half Housing price in second half Forecasted housing price to second
of year 1995
of year 2019
half of year 2022
17.08 ($)
9.52 ($)
24.87 ($)
12.32 ($)
18.64 ($)
31.02 ($)
8.56 ($)
20.69 ($)
10.52 ($)
8.12 ($)
6.51 ($)
7.25 ($)
15.53 ($)
6.05 ($)
5.29 ($)
2350.11 ($)
1658.21 ($)
3700.23 ($)
2001.11 ($)
2763.25 ($)
4707.79 ($)
1502.38 ($)
3102.52 ($)
1608.33 ($)
1459.63 ($)
1096.78 ($)
1203.33 ($)
2640.74 ($)
999.32 ($)
850.85 ($)
2798.66 ($)
1923.52 ($)
4150.36 ($)
2464.01 ($)
3051.33 ($)
5100.58 ($)
1821.11 ($)
3450.22 ($)
1901.87 ($)
1702.65 ($)
1370.59 ($)
1450.81 ($)
3011.38 ($)
1250.61 ($)
1091.47 ($)
Forecasting
house prices
Figure 3.
Housing price in
Isfahan city 2019
2022) that does not exist actual relevant data about them to compare, accuracy and
validation of model results can be evaluated according to the following arguments:
Accuracy of the predictive model is acceptable for existing data owing to the
obtained results and its ability for recognizing the general trends correctly.
In this work a multivariate method has been used for forecasting, other variables
alongside housing price have been utilized to decrease the errors and provide more
reasonable results.
In this article the targeted year is 2022, so it is a short-term forecasting. Generally,
the lower selected period to forecasting will increase the accuracy.
Housing price is a stock variable, it cannot increase or decrease far more than the
current value. In other words, housing price needs time to change considerably.
Given that the results, the forecasted prices to 2022 are reasonable in comparison to
current values.
About other boroughs of Isfahan city, housing price has been forecasted based on the same
technique, arguments and interpretations. Of course, every borough has its own forecasting
equations that obtained by the model.
IJHMA
Figure 4.
The rate of housing
price changes in
Isfahan city 2022
Owing to the obtained results, accuracy of housing price forecasting in Isfahan city housing
market using GMDH-type artificial neural network that has been conducted in this work, is
acceptable. Also, the relevant mathematical equations that produced by the model to
forecast housing price in Isfahan city have been presented in the results section and
simplified to be used easier by users. Finally, forecasting results for each borough of Isfahan
in 2022 have been shown on the relevant maps that created by ArcMap GIS software to
better analyze the price changes. So totally, the main goals of this article have been achieved.
Based on forecasting results, the housing price will increase at all boroughs of Isfahan in
the second half of 2022. Amongst them, Borough 15 will have the highest percentage of the
price increasing (28.27%) to year 2022 and Borough 6 will have the lowest percentage of the
price increasing (8.34%) to the year 2022. About ranking of the boroughs in terms of
housing price, Borough number 6 and 3 will keep their current position at the top until the
second half of 2022 because of predicting results. Accordingly, Borough number 15 of
Isfahan will stay at the bottom of the housing price ranking.
5. Conclusion
Humans have always been very interested in forecasting the future, predicting the prices in
the housing market has not been exempted from this rule. Knowing about future housing
price changes let the planners and policy makers be ready for quick reaction and handle the
situation of the housing market. Also, using accurate forecasting results can help urban
housing programs to be more feasible. Furthermore, recognizing the future trends of
housing price changes can help owners, developers and consumers take and make better
decisions. So, this article presents an accurate predictive model based on GMDH approach
utilizing GMDH-Shell software for forecasting housing price in a case study of Isfahan city
based on created time series and existing data. One of the most remarkable points of this
work is reaching to a mathematical formula that accurately can forecast housing price in
Isfahan city. This rarely has been presented in former studies, especially in simplified form.
Achieving the formula and simplifying that, can help users to use it easily for analyzing and
describing the status of the housing market and the changes in Isfahan city of Iran. Another
most important point is using ArcMap GIS to providing quality and legible maps in this
article. Descriptive maps can show the housing price changes more objective and visible
over the time; this is useful for users to analyze the changes more precise with regard to the
relevant locations. It should be said that the used technique in this research is not only for
Isfahan city and can be used for other cities. As a limitation, in this research just few factors
have been selected alongside housing price to control the forecasting process because of
limitation of reliable data availability about affecting factors. For the future studies about
the title of this article, using more related indices and periods to improve the accuracy of
model is recommended. Also, using a more powerful computer in terms of processing
abilities can improve the accuracy of results, in this research all calculations conducted on a
personal computer with an Intel Core i7-3630QM @2.4 GHz CPU, 8 GB RAM and Windows
10 (64 bit) operational system.
Notes
1. www.Isfahan.ir
2. www.amar.org.ir
3. www.gmdhsoftware.com
4. www.nationalgeographic.org
5. www.arcgis.com
6. www.esri.com
References
Ahmadi, M., Ahmadi, M.-A., Mehrpooya, M. and Rosen, M. (2015), “Using GMDH neural networks to
model the power and torque of a stirling engine”, Sustainability, Vol. 7 No. 2, pp. 2243-2255.
Ahmadi, M.H., Sadeghzadeh, M., Raffiee, A.H. and Chau, K-W. (2019), “Applying GMDH neural
network to estimate the thermal resistance and thermal conductivity of pulsating heat pipes”,
Engineering Applications of Computational Fluid Mechanics, Vol. 13 No. 1, pp. 327-336.
Ahn, J.J., Byun, H.W., Oh, K.J. and Kim, T.Y. (2012), “Using ridge regression with genetic algorithm to
enhance real estate appraisal forecasting”, Expert Systems with Applications, Vol. 39 No. 9,
pp. 8369-8379.
Alfiyatin, A.N., Febrita, R.E. and Taufiq, H. (2017), “Modeling house price prediction using regression
analysis and particle swarm optimization”, International Journal of Advanced Computer Science.
Assari, A., Maghreby, S. and Nik, M.M. (2017), “Investigation of smart growth in traditional islamic
culture: case study of isfahan city in Iran”, Journal of Geography and Regional Planning, Vol. 10
No. 4, pp. 47-56.
Forecasting
house prices
IJHMA
Azadeh, A., Ziaei, B. and Moghaddam, M. (2012), “A hybrid fuzzy regression-fuzzy cognitive map
algorithm for forecasting and optimization of housing market fluctuations”, Expert Systems with
Applications, Vol. 39 No. 1, pp. 298-315.
Balouchi, B., Nikoo, M.R. and Adamowski, J. (2015), “Development of expert systems for the prediction
of scour depth under live-bed conditions at river confluences: application of different types of
ANNs and the M5P model tree”, Applied Soft Computing, Vol. 34, pp. 51-59.
Bork, L. and Møller, S.V. (2015), “Forecasting house prices in the 50 states using dynamic model
averaging and dynamic model selection”, International Journal of Forecasting, Vol. 31 No. 1,
pp. 63-78.
Borzemski, L. and Wojtkiewicz, M. (2011), “Evaluation of chaotic internet traffic predictor using
MAPE accuracy measure”, in International Conference on Computer Networks, Springer,
pp. 173-182.
Chen, X., Wei, L. and Xu, J. (2017), House Price Prediction Using LSTM, The Hong Kong University of
Science and Technology.
Dong, S., Wang, Y., Gu, Y., Shao, S., Liu, H., Wu, S. and Li, M., (2020), “Predicting the turning points of
housing prices by combining the financial model with genetic algorithm”, PloS One, Vol. 15
No. 4, p. e0232478.
Ehrgott, M., Figueira, J. and Greco, S. (2010), Trends in Multiple Criteria Decision Analysis, Springer.
Ganti, R.K., Rallapalli, S. and Srivatsa, M. (2018), “predicting real property prices using a convolutional
neural network”, ed: US Patent App. 15/255, Vol. 687.
Ghanadzadeh, H., Daghbandan, A., Akbarizadeh, M. J. I. J. O C. and Engineering, C. (2014), “Applying
pareto design of GMDH-Type neural network for Solid-Liquid equilibrium of binary systems
(isotactic poly 1-Butene (1)-organic solvents (2))”, Iranian Journal of Chemistry and Chemical
Engineering (IJCCE), Vol. 33 No. 1, pp. 67-73.
Golubchikov, O. and Badyina, A. (2012), Sustainable Housing for Sustainable Cities: A Policy
Framework for Developing Countries, UN-HABITAT, Nairobi,
Gu, J., Zhu, M. and Jiang, L. (2011), “Housing price forecasting based on genetic algorithm and support
vector machine”, Expert Systems with Applications, Vol. 38 No. 4, pp. 3383-3386.
Hadavandi, E., Ghanbari, A., Mirjani, S.M. and Abbasian, S. (2011), “An econometric panel data-based
approach for housing price forecasting in Iran”, International Journal of Housing Markets and
Analysis, Vol. 4 No. 1, pp. 70-83.
Hyndman, R.J. and Athanasopoulos, G. (2018), “Forecasting: principles and practice”, OTexts.
Jia, X., Di, Y., Feng, J., Yang, Q., Dai, H. and Lee, J. (2018), “Adaptive virtual metrology for
semiconductor chemical mechanical planarization process using GMDH-type polynomial neural
networks”, Journal of Process Control, Vol. 62, pp. 44-54.
Khalafallah, A. (2008), “Neural network based model for predicting housing market performance”,
Tsinghua Science and Technology, Vol. 13 No. 1, pp. 325-328.
Kim, K.-Y., Park, J. and Sohmshetty, R. (2017), “Prediction measurement with mean acceptable error for
proper inconsistency in noisy weldability prediction data”, Robotics and Computer-Integrated
Manufacturing, Vol. 43, pp. 18-29.
Kordnaeij, A., Kalantary, F., Kordtabar, B. and Mola-Abasi, H. (2015), “Prediction of recompression
index using GMDH-type neural network based on geotechnical soil properties”, Soils and
Foundations, Vol. 55 No. 6, pp. 1335-1345.
Kouwenberg, R. and Zwinkels, R. (2014), “Forecasting the US housing market”, International Journal of
Forecasting, Vol. 30 No. 3, pp. 415-425.
Li, R.Y.M., Fong, S. and Chong, K.W.S. (2017), “Forecasting the REITs and stock indices: group method
of data handling neural network approach”, Pacific Rim Property Research Journal, Vol. 23 No. 2,
pp. 123-160, doi: 10.1080/14445921.2016.1225149.
Liu, R. and Liu, L. (2019), “Predicting housing price in China based on long short-term memory
incorporating modified genetic algorithm”, Soft Computing, Vol. 23 No. 22, pp. 11829-11838.
Morano, P., Tajani, F. and Torre, C.M. (2015), “Artificial intelligence in property valuations an
application of artificial neural networks to housing appraisal”, Advances in Environmental
Science and Energy, Planning, pp. 23-29.
Morano, P., Tajani, F., Guarini, M.R., Liddo, F.D. and Anelli, D. (2019), “A multivariate econometric
analysis for the forecasting of the interdependences between the housing prices and the socioeconomic factors in the city of Barcelona (Spain)”, in International Conference on Computational
Science and Its Applications, Springer, pp. 13-22.
Moroz, O. (2016), “Using hybrid algorithms based on GMDH-Type neural networks for solving
economic problems”, ICTERI, pp. 558-565.
Onwubolu, G. (2014), Gmdh-Methodology and Implementation in C [with CDROM], World Scientific
Publishing.
Park, B. and Bae, J.K. (2015), “Using machine learning algorithms for housing price prediction: the case
of Fairfax county, Virginia housing data”, Expert Systems with Applications, Vol. 42 No. 6,
pp. 2928-2934.
Rahadi, R.A., Wiryono, S.K., Koesrindartoto, D.P. and Syamwil, I.B. (2015), “Factors influencing the
price of housing in Indonesia”, International Journal of Housing Markets and Analysis, Vol. 8
No. 2.
Rossini, P. (2000), “Using expert systems and artificial intelligence for real estate forecasting”, in Sixth
Annual Pacific-Rim Real Estate Society Conference, Sydney, Citeseer, pp. 24-27.
Roy, K., Das, R.N., Ambure, P. and Aher, R.B. (2016), “Be aware of error measures. Further studies on
validation of predictive QSAR models”, Chemometrics and Intelligent Laboratory Systems,
Vol. 152, pp. 18-33.
Ruonavaara, H. (2018), “Theory of housing, from housing, about housing”, Housing, Theory and
Society, Vol. 35 No. 2, pp. 178-192.
Shaghaghi, S., Bonakdari, H., Gholami, A., Ebtehaj, I. and Zeinolabedini, M. (2017), “Comparative
analysis of GMDH neural network based on genetic algorithm and particle swarm optimization
in stable channel design”, Applied Mathematics and Computation, Vol. 313, pp. 271-286.
Shahsavar, A., Khanmohammadi, S., Karimipour, A. and Goodarzi, M. (2019), “A novel comprehensive
experimental study concerned synthesizes and prepare liquid paraffin-Fe3O4 mixture to develop
models for both thermal conductivity and viscosity: a new approach of GMDH type of neural
network”, International Journal of Heat and Mass Transfer, Vol. 131, pp. 432-441.
Stamou, M., Mimis, A. and Rovolis, A. (2017), “House price determinants in Athens: a spatial
econometric approach”, Journal of Property Research, Vol. 34 No. 4, pp. 269-284.
Stepashko, V., Bulgakova, O. and Zosimov, V. (2017), “Construction and research of the generalized
iterative GMDH algorithm with active neurons”, in Conference on Computer Science and
Information Technologies, Springer, pp. 492-510.
Tomaszewski, B. (2014), Geographic Information Systems (GIS) for Disaster Management, CRC Press.
Tsantekidis, A., Passalis, N., Tefas, A., Kanniainen, J., Gabbouj, M., (2017), and A. and Iosifidis,
“Forecasting stock prices from the limit order book using convolutional neural networks”, in
2017 IEEE 19th Conference on Business Informatics (CBI), Vol. 1, pp. 7-12.
Varahrami, V. (2012), “Good prediction of gas price between MLFF and GMDH neural network”,
International Journal of Finance and Accounting, Vol. 1 No. 3, pp. 23-27.
Waltl, S.R. (2016), “A hedonic house price index in continuous time”, International Journal of Housing
Markets and Analysis, Vol. 9 No. 4.
Wei, Y. and Cao, Y. (2017), “Forecasting house prices using dynamic model averaging approach:
evidence from China”, Economic Modelling, Vol. 61, pp. 147-155.
Forecasting
house prices
IJHMA
Xu, L. and Li, Z. (2020), “A new appraisal model of Second-Hand housing prices in China’s First-Tier
cities based on machine learning algorithms”, Computational Economics, pp. 1-21.
Yang, L., Yang, H. and Liu, H. (2018), “GMDH-based semi-supervised feature selection for electricity
load classification forecasting”, Sustainability, Vol. 10 No. 1, p. 217.
Zhang, Y., Wang, H., Liu, S., Lei, Q., Liu, J., He, J., Zhai, L., Ren, T. and Liu, H. (2015), “Identifying critical
nitrogen application rate for maize yield and nitrate leaching in a haplic luvisol soil using the
DNDC model”, Science of the Total Environment, Vol. 514, pp. 388-398.
Corresponding author
Mohsen Rafiean can be contacted at: mrafian@yazd.ac.ir
For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: permissions@emeraldinsight.com