Information Processing and Management 60 (2023) 103319

Contents lists available at ScienceDirect

Information Processing and Management

journal homepage: www.elsevier.com/locate/infoproman

Information analysis for dynamic sale planning by AI decision

support process
Feng Wang a, b, Yu Luan b, *, Abdel Nour Badawi c, Abbad Ayad c,
Abdel Fattah Abdallah d, Mansour Ali d, Zobair Ahmad d, Wu Jiang d
a
Big Data Management Research Center, Jilin University, Jilin,130012, China
b
School of Business and Management, Jilin University, Jilin, 130012, China
c
School of Business, MIS Department, The University of Jordan, Amman, Jordan
d
School of Business and Management, Universiti Putra Malaysia, 43400, Malaysia

A R T I C L E I N F O A B S T R A C T

Keywords: Marketing is all about finding out what your customers want and need. The information is used to
Artificial intelligence AI create the model and improve the quality of products and services to satisfy customers. There are
Decision support process two major parts to a sale’s planning: containing sales tactics and sales strategy. To aid in business
Information analysis
decision-making, a data warehouse (DW) for sales collects and organizes relevant and historical
Particle swarm optimization
Sale planning
data. DW is also a method for combining data from various heterogeneous databases (DB) and
Database management system other sources of information for analysis. In this paper, Information analysis for dynamic Sale
planning by AI Decision support process was done. Particle Swarm Optimization (PSO) and
Quantum Particle Swarm Optimization (QPSO) algorithms have been used to find optimal MVs
for sale planning since several authors prove that these algorithms are better than other existing
algorithms. MVs selection is based on three factors: time running of MVs, area of MVs, and access
frequency of MVs by manipulating the selection result using a weighted combination of each
factor as needed by the designer. For PSO, the most cost of frequency was 1.924 for View (V) 11,
while for QPSO, it was 1.722 for V11. For PSO, the most cost of time was 1.931 for V2, while for
QPSO, it was 1.221 for V22. For PSO, the most cost of the area was 1.800 for V17, while for
QPSO, it was 1.071 for V17. The results revealed that Quantum Particle Swarm Optimization is
much more accurate than Particle Swarm Optimization.

1. Introduction

Nowadays, the business competitive advantage comes from having access to business information (Yang et al., 2022). Marketing is
all about finding out what your customers want and need (Huang et al., 2022). The information was used to create the model and
improve the quality of products and services to satisfy customers (Wu & Liao, 2021). The revenue generated by the sale of goods and/or
services. There are two major parts to a sale’s planning: containing sales tactics and sales strategy. Sale planning is frequently the
source of problems. When the cost of goods sold is raised or lowered, the demand for a product may shift (Gupta et al., 2022). This issue

Abbreviations: DW, data warehouse; DB, Database; QPSO, quantum particle swarm optimization; PSO, particle swarm optimization; DBMS,
database management system; PG, polynomial greedy; HPSO, human particle swarm optimization; DSS, decision support system.
* Corresponding author.
E-mail address: luanyu20@mails.jlu.edu.cn (Y. Luan).

https://doi.org/10.1016/j.ipm.2023.103319
Received 27 May 2022; Received in revised form 23 January 2023; Accepted 6 February 2023
Available online 16 February 2023
0306-4573/© 2023 Elsevier Ltd. All rights reserved.
F. Wang et al. Information Processing and Management 60 (2023) 103319

should be thoroughly investigated to devise a strategy for determining the selling price (Seeling et al., 2021). To aid in business
decision-making, a DW for sales collects and organizes relevant and historical data (Zhu & Hao, 2020). In DW, to assess performance,
the data is queried and analyzed, and the process of trends and business will help in making better decisions and developing operations.
DW is also a method for combining data from various heterogeneous DB and other sources of information for analysis. Analytical and
query-ready copies of operational data reside in the DW. The query response time is the most useful problem in any very large DB
(Jaleel & Abbas, 2020a). There’s another technique to reduce the time of query, such as a pre-calculated view of the data stored in the
DW for online use, such as MVs. When compared to accessing the base tables, this can achieve the fastest response time but this
approach has several drawbacks because it necessitates synchronizing response time and space requirements. After all, materializing
all viewpoints would be impossible (Heydarpoor et al., 2020).
The fastest time of response is achieved, but this necessitates a lot of space and a lot of maintenance because any changes made to
the base tables must be reflected in the logical tables. MVs a useful role in the performance of the DB. In the database management
system (DBMS), MVs are a data structure used to speed up access to data in databases. Physical structures such as MVs and indexes can
significantly improve assessment (Fouladi et al., 2022). An efficient physical DB building software for DSS systems must, therefore,
take into account the interaction between sales planning, MVs, and optimization algorithm. A system’s physical design can be
improved by taking these factors into account together. In the literature, several algorithms are applied to choose optimal MVs like the
greedy algorithm, simulated annealing algorithm, PSO algorithm, QPSO, and human PSO, etc.
Some previous research has been done on the best MV to pick in the DW and some of the algorithms used, some of which are related
to this paper. In the study of Yao et al. (2015) the authors described a method for the effective and improved algorithm of Polynomial
Greedy (PG) for the selection of MV. The algorithm took into account the impact of the entire space as well as the cost of adding
candidate MVs and choosing a lower cost for selecting views limits the number of candidate MVs that can be added or removed. The
findings of the tests and analysis offered that the algorithm worked well and was efficient. In the study of Gosain (2016), the authors
offered a suggested method for running the algorithm of PSO on the method of lattice for the selection of MV in DW, the experiment
was carried out by running algorithms on the TPC–H benchmark from the Transaction Processing Performance Council, using a
different frequency and number of dimensions to implement the technique, the findings described that the PSO algorithm out­
performed the algorithm of genetic in terms of selecting a suitable set of MV with lower processing cost. In the study of Salah and
Hamad. (2018), the authors offer a method for selecting the best MV by applying PSO to realize an efficient set of good running time,
and low query processing costs, findings show that the MV recommendation technique is superior to other techniques. In the study of
Abdullah et al. (2019) methodology for creating ideal MVs based on three parameters has been developed (MV- time, MV-space, and
MV-frequency) using aggregations on bitmap indexes and the Firefly-Algorithm (F-A) can be used to find the best option for complex
queries and minimizing queries response time. In the study of Jaleel and Abbas (2020b) the authors provided a way for selecting the
optimal MV using the QPSO algorithm, resulting in effective collection of low query handling cost and good query response of the time.
The results reveal that the QPSO algorithm-based strategy for selecting the best MV is superior to other techniques via computing and
compare the ratio of response time to the time it took to respond to the same query on the MVs. In the study of Fadhil (2020) the
authors the initial framework for accelerating the response time to queries is presented depending on the algorithm of Human Particle
Swarm Optimization (HPSO) for describing the top viewpoints in the DW.
From the related works, the problem statement is concluded that the main issue of the admin of DW is which view will be extracted
first to be materialized in the DSS if the DW. No practical way of keeping track of multiple queries’ MVs until the disk space re­
quirements, as the MV is realized physical table that’s why there’s a big consumption and/or a big update cost. To materialize, one
option is to extract a set of derived views and reduce the sum of the total query maintenance costs and times of the views selected. That
is the issue of view selection.
For the wide applicability of PSO, the following can be mentioned: Kumari and Sahana (2022), proposed a ’Hybrid ACO-PSO
Meta-Heuristic (HAPM)’, a combination of ACO, PSO, and a dynamic queue mechanism to improve QoS constraints and minimize
QoS the data dropping. The Hop count of HAPM has been reduced by 90% and 87% in comparison with ACO and PSO, individually.
Sahana (2022) proposed an advanced form of PSO (balanced PSO (Ba-PSO)) to solve the scheduling problem of the computational grid.
The proposed, Ba-PSO, was scalable and worked for small as well as large datasets. Srivastava and Sahana (2020) presented an
overview of biologically inspired techniques and their various implementations for traffic optimization to minimize wait time, improve
safety and reduce pollution.
In this paper, PSO and QPSO have been used to find optimal MVs for sale planning since several authors prove that these algorithms
are better than other existing algorithms. MVs selection is based on three factors (time running of MVs, area of MVs, and access
frequency of MVs by manipulating the selection result using a weighted combination of each factor as needed by the designer. This
research offered the methodology to achieve high query performance and this work various from other papers and research firstly by
applying PSO and QPSO algorithm for MVs in sale planning to guarantee the DW’s high-quality standards and secondly, the DW is
specialized for DSS of sale planning, the exponential-distribution of locations in QPSO makes it global convergent, which is a sig­
nificant advantage over PSO. A second improvement of QPSO is the inclusion of the mean best position in the evolution equation.
This study develops the sale planning for marketing by introducing the decision support process which has been improved by
artificial intelligence. This research will explain how to select the most suited materialized view using various main parameters; cost of
the area, frequency, and time for queries.
The main objective of the current study is to develop the sale planning for marketing by artificial intelligence method and while the
most pressing issue for the DW designer/administrator is deciding which view should be used first, SIDSS-ESP will be suggested to
extract and analyze important information.
For the limitations of this model, maintaining MVs for each question is problematic since the MV is conducted on a physical table

2
F. Wang et al. Information Processing and Management 60 (2023) 103319

Fig. 1. Components of a Decision Support System (Kumar, 2020; Malekshah et al., 2016).

Fig. 2. Flowchart for PSO (Piotrowski et al., 2020; Ang et al., 2020).

that is limited by disc capacity, because of this, the consumption and/or updating costs are extremely expensive.
For the assumptions of this model, a possible solution is to realize a group of derived perspectives, as a result of which the total
reaction time of the selected views will be reduced.
The rest of the paper is as follows: the related works and problem statement are offered in section 2, followed by the materials and
methods in section 3. Section 4 illustrates the experiments and discusses the suggested ISDSS-ESP results. Conclusions are offered in
section 5.

2. Materials and methods

2.1. Decision support system

The Decision Support System (DSS) is a set of tools that can be used for analytical processing, as well as query and decision analysis

3
F. Wang et al. Information Processing and Management 60 (2023) 103319

Fig. 3. Flowchart of QPSO (Xin-gang et al., 2020b; Tu et al., 2020).

in an integrated DW. According to this research’s view, data analysis and decision-making are both protected by DW’s protection of
business intelligence and its ease of use. For nearly 40 years, information systems researchers and technicians have constructed and
examined (Kumar, 2020). The benefits expected from DSS discovered are improved employee satisfaction, higher decision quality, cost
reduction, improved customer satisfaction, time savings increased productivity, and improved communication. DSS are systems that
assist managers to optimize the decision-making process, by giving the possibility to retrieve efficient information quickly from
multiple blocks of data. All DSS include three essential components (Fig. 1) (Kumar, 2020; Malekshah et al., 2016).

2.2. Particle swarm optimization

Global optimization using PSO is a method that uses swarm intelligence theory to find the best solution. It behaves similarly to the
fish schools’ birds and flocks in search of food. When the birds search for food, they exchange information about the best possible
method to find food and finally flock to that place. PSO, similarly, works as follows: It begins with a non-consistent initial population of
n particles, each of which represents a potential solution to the problem. In D-dimensional space, each particle has a position and some
velocity (Pervaiz et al., 2021). The algorithm’s goal is to find the smallest possible value for an objective function. Particles’ position
and velocity are changed continuously according to the best position found by the whole population (global best) and by the particle
itself (particle best) so far. Hence, to obtain the objective function’s idealism value, particles learn from one another and then influence
the swarm’s best member. PSO binary form is used, where a binary string defines each candidate solution (Mohseni & Sohrabi, 2020).
The range of [0, 1] also includes the velocity parameter and determines the likelihood of a bit on the string being altered. Velocity is
normalized using the sigmoid function and updated according to g best and p best values (Fig. 2) (Piotrowski et al., 2020).
Simplifying query processing by selecting the best possible set of views within a certain amount of available space is at the heart of
MV Selection. To discover the lowest cost value, the PSO algorithm examines the entire space and progresses in the direction of the
best-found solution to run the user queries as quickly as possible and return the results (Ang et al., 2020). The following equation
represents (PSO) algorithm (Ang et al., 2020):
p = p + v (1)

with

4
F. Wang et al. Information Processing and Management 60 (2023) 103319

Fig. 4. Difference between PSO and QPSO (Tu et al., 2020; Lai et al., 2020).

V= V
⏟⏞⏞⏟ + C1 ∗ rand ∗ (pBest − − p) + C2 ∗ rand ∗ (gBest − − p)
⏟̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅⏞⏞̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅⏟
Diversification Intensification

where v: Path direction, p: position of a particle, c2: global-information weight. pBest: optimal placement of the particle, c1: Weight of
local information, rand: Random variable, gBest: Best position of the swarm, Intensification: Investigates prior solutions, finds the best
solution of a given region, Diversification: novel solutions are sought, and the parts with the best solutions in PSO are identified.

2.3. Quantum particle swarm optimization

The algorithm of QPSO establishes the current and optimal locations of particles, the mean best location of the swarm’s particle is
updated and computed at each iteration (Xin-gang et al., 2020a). The algorithm of QPSO is a novel intelligent optimization algorithm;
it has fewer parameters and is easily implemented. Before the current position is updated, each particle assesses and changes its fitness
value, its best position. The term particle refers to a bee in a flock or colony of birds, and the function of fitness is computed by applying
statistical calculations for each location in the sample space. These computations are variance, mean, position Y, and position X.
position Y and position X indicate the coefficient’s 2-dimensional locations in the sample space (Fig. 3). The variance and mean for
each position is produced by the formulas (1) and (2) (Xin-gang et al., 2020b; Tu et al., 2020).
∑ /
Mean = (Xi M j = 1) M (2)

Variance = (Xi − mean) 2/ M (3)

Where M is the number of locations in sample space and xi (j) is the value in a specific position. Continue until the specified end
conditions are met (The required location number is less than or equal to the number of iterations) are met. Fig. 1 shows a flowchart for
the algorithm of QPSO. The PSO algorithm may reach convergence if each particle concentrates on its local attractor. The local
attractor Pji,t = (P1i,t , P2i,t , ……., PDi,t ) of particles is defined as coordinate (Xin-gang et al., 2020b; Tu et al., 2020):
( )
Pji,t = Qji,t Pji,t + 1 − Qji,t Gtj (4)

Where Qji,t is a sample of random numbers with a normal distribution [Yang et al., 2022), 0]. The previous equation denotes that
particle i’s stochastic (local) attractor, allocations are made to the hyper rectangle of Gt and Pji,t that is being the coordinate of its
diagonally so that it in the direction Gt andPji,t . Using a different N-dimensional Hilbert space than Newtonian space, the individual
particles in PSO can have a quantum attitude and motion, leading to a different PSO from the perspective of quantum mechanics. In
QPSO, the particle’s position is updated by solving any of the Eqs. (5) and ((6) (Xin-gang et al., 2020b; Tu et al., 2020):
( )
⃒ ⃒ 1
⃒ ⃒
d
Xi,t+1 = Pdi,t ± α⃒Xi,td − Ctd ⃒ln (5)
Ui,td

/
d
Li,t
Xi,td = Pdi,t ± ln1 u (6)
2

A random number that is evenly distributed over the range [(Yang et al., 2022), 0] is called U. L (i,t)d can be calculated using the
formula (7) (Xin-gang et al., 2020b; Tu et al., 2020):
⃒ ⃒
⃒ ⃒
Ldi,t = 2α⃒Xi,td − Pdi,t ⃒ (7)

2.4. Difference between PSO and QPSO

QPSO differs from the PSO algorithm in some characteristics. Firstly, QPSO is globally convergent because of the exponential

5
F. Wang et al. Information Processing and Management 60 (2023) 103319

Fig. 5. Proposed Methodology.

Fig. 6. Area of Queries for PSO.

6
F. Wang et al. Information Processing and Management 60 (2023) 103319

Fig. 7. Frequency of Queries for PSO.

distribution of positions (Malekshah et al., 2022). A second improvement of QPSO is the addition of the mean best position (mbest) to
the evolution equation. In SPSO, each particle converges to its global best position independently (Tu et al., 2020). The QPSO is
described by the equation in which the distance between the current position of a particle and the mean best position obtains the
particle’s next iteration’s location distribution (Fig. 4) (Lai et al., 2020).

2.5. Proposed methodology

The Framework was carried out on sales DSS and the block diagram of it is shown in Fig. 5, after building DW architecture, there are
twenty-five complex queries from requirements of users which include data from the sales planning system, MVs are needed to improve
query performance because of the complexity of the queries. After making setup and upload for the complex queries (views and tables)
in the SQL server, then the step of calculating time, frequency, and area for each query has been done (these parameters are considered
as the base for operating the PSO and QPSO). The next step is calculating the cost of each query or view depending on the important
parameters. And then the PSO and QPSO are set up to apply it to the table of the cost for the complex queries. The first step of the cost
table is making it a matrix to operate the optimization algorithms on it. After the algorithms have been done and select the best views
or queries, PSO (Salah & Hamad., 2018) and QPSO (Jaleel & Abbas, 2020b) are compared in selecting the best queries. And it is
founded that the QPSO is best than PSO.

7
F. Wang et al. Information Processing and Management 60 (2023) 103319

Fig. 8. Time for Queries of PSO.

Fig. 9. Area for Queries of QPSO.

3. Results

3.1. Selection parameter of views

DW of swarm intelligence for sale planning contains several tables (Invoices, Supplier, Items, Details, and Warehouse), there are
twenty-five complex (SQL) operational aggregation (OLAP) queries assumed in this part such as (min, max, and sum, count, select,
join, filtering operations like applying group by operation and condition where to choose data from tables of sale planning DSS, after

8
F. Wang et al. Information Processing and Management 60 (2023) 103319

Fig. 10. Time of Queries for QPSO.

Fig. 11. Cost of Parameters for PSO.

obtaining inquiries from the DW. Each query’s area has been calculated, and the greatest value of area from all queries is saved as offers
in Fig. 6, which displays the area findings for twenty-five queries. After that, the time spent on each query is computed, and the longest
total time spent on all inquiries is then summed, as offered in Fig. 7, which displays the outcome of twenty-five queries. After that, the
frequency of each query was computed, and after that, the maximum value of frequency possible from across all searches was saved as
shown in Fig. 8, this demonstrates that the outcome is between 1 & 0, 1 indicates that the query is demanded and 0 indicates that the
query is not demanded as offers in Figs. 8-10. In Fig. 6, the maximum number of areas for Q1 and Q12 which is 300. Also, the minimum
areas are for Q9, Q16, Q17, and Q22 which are 200. In Fig. 8, the minimum times are 21 which is for Q22, and 22 which is for Q11.
Also, the maximum times are 40 which is for Q1, and 38 which is for Q10 (Fig. 9).

3.2. Selection cost of queries

The cost of parameter selection for each query such as cost of time, cost of frequency, and cost of the area has been computed and
the frequency of each query divided by the maximum frequency for all frequencies is used to calculate the frequency of each inquiry.

9
F. Wang et al. Information Processing and Management 60 (2023) 103319

Fig. 12. Cost of Parameters for QPSO.

Table 1
Area and time for queries of PSO.
No. View Area Time

1 V1 300 40
2 V2 250 35
3 V3 250 30
4 V4 225 32
5 V5 275 37
6 V6 250 30
7 V7 270 25
8 V8 225 30
9 V9 200 35
10 V10 210 38
11 V11 220 22
12 V12 300 34
13 V13 275 29
14 V14 290 32
15 V15 295 24
16 V16 200 31
17 V17 200 34
18 V18 225 27
19 V19 210 30
20 V20 205 32
21 V21 240 22
22 V22 200 21
23 V23 230 24
24 V24 240 29
25 V25 250 28

The cost of MV for each query selection has been computed, as well as for time and area (Fig. 11 and Fig. 12). The outcomes are shown
in Tables 1 and 2. The cost of Parameters for PSO is shown in Fig. 11 and for QPSO is shown in Fig. 12. For PSO, the most cost of
frequency is 1.924 for V11, while for QPSO, it is 1.722 for V11. For PSO, the most cost of time is 1.931 for V2, while for QPSO, it is
1.221 for V22. For PSO, the most cost of the area is 1.800 for V17, while for QPSO, it is 1.071 for V17. The mentioned values are shown
in Tables 3 and 4. Using a formula shown in the study of Salah and Hamad and the study of Jaleel and Abbas (Salah & Hamad., 2018;
Jaleel & Abbas, 2020b), the cost of MV for each query selection was computed. Table 5 (cost selection of PSO) and 6 (cost selection of
QPSO) show the cost of MV for each query. For PSO, the maximum values of weights 1, 2, and 3 are 0.921479 (V25), 0.991277 (V3),
and 0.986100 (V15), individually. Also, the maximum cost selection is 1.952915 for V25. For QPSO, the maximum values of weights 1,
2, and 3 are 0.431289 (V22), 0.791274 (V3), and 0.783218 (V9), individually. Also, the maximum cost selection is 1.007293 for V15.

10
F. Wang et al. Information Processing and Management 60 (2023) 103319

Table 2
Area and time for queries of PSO.
No. View Area Time

1 V1 250 35
2 V2 225 30
3 V3 220 28
4 V4 200 30
5 V5 245 33
6 V6 240 28
7 V7 239 22
8 V8 200 28
9 V9 170 32
10 V10 180 31
11 V11 190 20
12 V12 240 30
13 V13 231 29
14 V14 244 30
15 V15 217 30
16 V16 200 28
17 V17 150 30
18 V18 190 25
19 V19 170 28
20 V20 175 24
21 V21 180 20
22 V22 179 19
23 V23 212 21
24 V24 222 24
25 V25 244 25

Table 3
Cost for area, time, and frequency of PSO.
View Area Cost Time Cost Frequency Cost

V1 0.912 1.061 1.328

V2 0.819 1.931 1.232
V3 0.124 1.281 1.451
V4 0.318 1 1.231
V5 1.123 1.028 1.8
V6 1.7 1.311 1.341
V7 0.891 1.232 1.521
V8 1.012 1.222 1
V9 0.113 1.921 1.901
V10 0.412 1.223 1.322
V11 0.511 1.381 1.924
V12 0.211 1.541 1.311
V13 0.378 1.277 1.2
V14 0.212 1.114 1
V15 0.341 1.291 1
V16 1.5 1 1.333
V17 1.8 1.024 1.533
V18 0.317 1.103 1.215
V19 0.112 1.421 1.777
V20 0.321 1.021 1.002
V21 0.215 1.112 1.322
V22 0.321 1.313 1
V23 0.212 1.101 1.035
V24 0.141 1.212 1.009
V25 0.201 1 1.877

3.3. Implementing PSO and QPSO algorithms

The PSO and QPSO algorithms have been built up to improve the difficult processing of view time using swarm intelligence. Work
begins with all inquiries in a two-dimensional matrix as indicated in Figs. 13 and 14 of the PSO and QPSO algorithms respectively,
which describe the matrix of two-dimensional of all queries. Then, the PSO and QPSO algorithms will choose the optimal query pa­
rameters. For each row of data, the cost of each query is divided by the number of rows to get the optimum MV, which saves cost,
depressed processing, and good response for the query. After that cost for all views has been subtracted from the rate of the lower ratio
of error then the findings have been squared to take them positively. The locations with the lowest error ratios were then chosen from
all inquiries, as per the algorithms of PSO and QPSO. Finally, the query with the lowest cost and highest frequency will be chosen based

11
F. Wang et al. Information Processing and Management 60 (2023) 103319

Table 4
Cost for area, time, and frequency of QPSO.
View Area Cost Time Cost Frequency Cost

V1 0.802 1.002 0.118

V2 0.711 1.021 0.122
V3 0.024 1.071 0.341
V4 0.215 0.211 1.121
V5 1.003 0.028 0.891
V6 0.9 1.201 0.229
V7 0.781 0.232 0.421
V8 1.004 1.009 0.871
V9 0.007 0.221 0.702
V10 0.302 0.113 1.112
V11 0.401 1.001 1.722
V12 0.101 1.018 1.201
V13 0.278 0.177 0.271
V14 0.102 1.004 0.222
V15 0.241 1.181 0.921
V16 1.022 0.902 0.213
V17 1.071 0.024 0.423
V18 0.214 0.103 0.217
V19 0.111 0.421 1.241
V20 0.121 0.021 0.102
V21 0.015 1.002 1.112
V22 0.021 1.221 0.421
V23 0.002 1.011 0.225
V24 0.101 1.112 0.109
V25 0.109 0.713 0.777

Table 5
Cost selection of PSO.
View Weight 1 Weight 2 Weight 3 Cost Selection

V1 0.3216111 0.0124721 0.3171022 0.2456009

V2 0.3451023 0.2271023 0.1420481 0.1378329
V3 0.0571224 0.9912771 0.0412911 0.0235193
V4 0.0281821 0.3890213 0.1930111 0.4127893
V5 0.1031791 0.0120541 0.4186923 0.0247429
V6 0.0214778 0.0311788 0.7145021 0.2189124
V7 0.4103991 0.3217902 0.0099325 0.4190042
V8 0.4217928 0.2210875 0.4819002 0.1673893
V9 0.6170321 0.4120842 0.8832179 0.3184002
V10 0.0235812 0.9136729 0.9145827 1.0034023
V11 0.5218932 0.8317802 0.3271926 0.8210063
V12 0.0712924 0.8127032 0.3418003 0.0312798
V13 0.0321964 0.7210972 0.0214893 1.0087421
V14 0.1134098 0.6138992 0.3120883 1.0284032
V15 0.1297003 0.0237119 0.9861003 1.0074932
V16 0.0071742 0.0145889 0.4900321 0.8714728
V17 0.4180651 0.2810072 0.9217801 0.4183628
V18 0.7190265 0.1245002 0.7123033 0.9943821
V19 0.0221793 0.1243004 0.1347024 0.2922102
V20 0.4218798 0.3217932 0.9216883 0.1178432
V21 0.7215683 0.8873452 0.1352894 0.0217882
V22 0.7312893 0.7772197 0.8126392 0.2996743
V23 0.0247118 0.9884003 0.4217893 0.1217943
V24 0.0533218 0.2189332 0.4117993 1.0342994
V25 0.9214789 0.7123008 0.0321892 1.9529148

on the chosen position for the lowest ratio of error.

3.4. Time of view in MVs

The response time of view in DW of swarm intelligent DSS for sale planning is important, to speed up decision-making, performing
queries on MVs gives users a quick response time such as four complicated queries that are operating outside the DW, as depicted in
Fig. 15 and time of query is then calculated and compared to MVs’ response time for queries of the same type, for PSO algorithm, As
shown in Fig. 15, the time it takes to implement a query on the base-table is 16 times longer than the time it takes to implement a query
on the MV (Kitsios & Kamariotou, 2021; Maggi & Marrella, 2021). And for QPSO, the time it takes to implement a query on the base

12
F. Wang et al. Information Processing and Management 60 (2023) 103319

Fig. 13. Cost Matrix of PSO.

Fig. 14. Cost Matrix of QPSO.

Fig. 15. Running Time.

table is 8 times longer than the time it takes to implement a query on the MV of PSO (Table 6).

4. Conclusions

In this paper, SIDSS-ESP has been suggested to extract and analyze important information. For the DW designer/administrator, the
most pressing issue is deciding which view should be used first. Maintaining MVs for each question is problematic since the MV is
conducted on a physical table that is limited by disc capacity, because of this, the consumption and/or updating costs are extremely
expensive. A possible solution is to realize a group of derived perspectives, as a result of which the total reaction time of the selected
views will be reduced. This paper explains how to select the most suited materialized view using various main parameters; cost of the
area, frequency, and time for queries.

13
F. Wang et al. Information Processing and Management 60 (2023) 103319

Table 6
Cost selection of QPSO.
View Weight 1 Weight 2 Weight 3 Cost Selection

V1 0.1116111 0.0024712 0.2171021 0.1246007

V2 0.2350023 0.1271022 0.0220481 0.0278329
V3 0.0211024 0.7912741 0.0212911 0.0224192
V4 0.0171721 0.2890212 0.1130111 0.2127892
V5 0.0031791 0.0110441 0.2086922 0.0127427
V6 0.0112778 0.0211777 0.4145011 0.1189122
V7 0.2003991 0.2217901 0.0099324 0.0190022
V8 0.2217928 0.1110874 0.2819002 0.0473892
V9 0.4070321 0.2120841 0.7832177 0.2184001
V10 0.0115812 0.7136727 0.7145824 0.9934022
V11 0.4218932 0.7317801 0.2271924 0.7210062
V12 0.0410924 0.7127031 0.2418002 0.0212797
V13 0.0221962 0.4210942 0.0214891 0.0087422
V14 0.0124097 0.4138991 0.2120881 0.9284032
V15 0.0197002 0.0227117 0.4861102 1.0072932
V16 0.0040442 0.0145888 0.2900221 0.7714727
V17 0.2170441 0.1810071 0.7207801 0.2183627
V18 0.4190242 0.0245001 0.4123002 0.7743822
V19 0.0221722 0.0243002 0.0347022 0.1722112
V20 0.2218787 0.2217931 0.4216003 0.0178431
V21 0.4215682 0.7873451 0.0442894 0.0117884
V22 0.4312891 0.4772194 0.7122292 0.1774742
V23 0.0117117 0.7884002 0.2217892 0.0217922
V24 0.0223217 0.1189331 0.4007993 1.0042974
V25 0.2214784 0.4123007 0.0021872 0.9529147

• It has been implemented and compared between PSO and QPSO algorithms to obtain a high level of query efficiency.
• The findings of this study show that the query’s response time of sale planning can be improved by optimizing it using PSO and
QPSO.
• For PSO, the most cost of frequency is 1.924 for V11, while for QPSO, it is 1.722 for V11. For PSO, the most cost of time is 1.931 for
V2, while for QPSO, it is 1.221 for V22. For PSO, the most cost of the area is 1.800 for V17, while for QPSO, it is 1.071 for V17.
• For PSO, the maximum values of weights 1, 2, and 3 are 0.921479 (V25), 0.991277 (V3), and 0.986100 (V15), individually. Also,
the maximum cost selection is 1.952915 for V25. For QPSO, the maximum values of weights 1, 2, and 3 are 0.431289 (V22),
0.791274 (V3), and 0.783218 (V9), individually. Also, the maximum cost selection is 1.007293 for V15.

Data availability

Data will be made available on request.

References

Abdullah, W. A., Sahib, N. M., & Abass, J. M. (2019). Creation of optimal materialized views using bitmap index and firefly algorithm in data warehouse. In 2019 2nd
Scientific conference of computer sciences (SCCS) (pp. 171–176). IEEE.
Ang, K. M., Lim, W. H., Isa, N. A. M., Tiang, S. S., & Wong, C. H. (2020). A constrained multi-swarm particle swarm optimization without velocity for constrained
optimization problems. Expert Systems with Applications, 140, Article 112882.
Fadhil, Z. M. (2020). Human behavior based particle swarm optimization for materialized view selection in data warehousing environment. Periodicals of Engineering
and Natural Sciences, 8(4), 2367–2378.
Fouladi, S., Safaei, A. A., Mammone, N., Ghaderi, F., & Ebadi, M. J. (2022). Efficient deep neural networks for classification of Alzheimer’s disease and mild cognitive
impairment from scalp EEG recordings. Cognitive Computation, 1–22.
Gosain, A. (2016). Materialized cube selection using particle swarm optimization algorithm. Procedia Computer Science, 79, 2–7.
Gupta, S., Modgil, S., Bhattacharyya, S., & Bose, I. (2022). Artificial intelligence for decision support systems in the field of operations research: Review and future
scope of research. Annals of Operations Research, 308(1), 215–274.
Heydarpoor, F., Karbassi, S. M., Bidabadi, N., & Ebadi, M. J. (2020). Solving multi-objective functions for cancer treatment by using metaheuristic algorithms.
Algorithms, 21, 22.
Huang, Z. X., Savita, K. S., Dan-yi, L., & Omar, A. H. (2022). The impact of business intelligence on the marketing with emphasis on cooperative learning: Case-study
on the insurance companies. Information Processing & Management, 59(2), Article 102824.
Jaleel, R. A., & Abbas, T. M. (2020a). Design and implementation of efficient decision support system using data mart architecture. In 2020 International conference on
electrical, communication, and computer engineering (ICECCE) (pp. 1–6). IEEE.
Jaleel, R. A., & Abbas, T. M. (2020b). Materialized views quantum optimized picking for independent data marts quality. Iraqi Journal of Information and
Communications Technology (IJICT), 3(1), 26–39.
Kitsios, F., & Kamariotou, M. (2021). Artificial intelligence and business strategy towards digital transformation: A research agenda. Sustainability, 13(4), 2025.
Kumar, T. S. (2020). Data mining based marketing decision support system using hybrid machine learning algorithm. Journal of Artificial Intelligence, 2(03), 185–193.
Kumari, P., & Sahana, S. K. (2022). Swarm based hybrid ACO-PSO meta-heuristic (HAPM) for QoS multicast routing optimization in MANETs. Wireless Personal
Communications, 123(2), 1145–1167.
Lai, X., Hao, J. K., Fu, Z. H., & Yue, D. (2020). Diversity-preserving quantum particle swarm optimization for the multidimensional knapsack problem. Expert Systems
with Applications, 149, Article 113310.
Maggi, F. M., & Marrella, A. (2021). Preface to the special issue on artificial intelligence for business process management 2019. Journal on Data Semantics, 10(1), 1–2.

14
F. Wang et al. Information Processing and Management 60 (2023) 103319

Malekshah, S., Hovanessian, A., & Gharehpetian, G. B. (2016). Combined heat and power sizing in residential building using mixed integer nonlinear programming
optimization method. In 2016 24th Iranian conference on electrical engineering (ICEE) (pp. 1208–1213). IEEE.
Malekshah, S., Rasouli, A., Malekshah, Y., Ramezani, A., & Malekshah, A. (2022). Reliability-driven distribution power network dynamic reconfiguration in presence
of distributed generation by the deep reinforcement learning method. Alexandria Engineering Journal, 61(8), 6541–6556.
Mohseni, M., & Sohrabi, M. K. (2020). MVPP-based materialized view selection in data warehouses using simulated annealing. International Journal of Cooperative
Information Systems, 29(03), Article 2050001.
Pervaiz, S., Ul-Qayyum, Z., Bangyal, W. H., Gao, L., & Ahmad, J. (2021). A systematic literature review on particle swarm optimization techniques for medical diseases
detection. Computational and Mathematical Methods in Medicine, 2021.
Piotrowski, A. P., Napiorkowski, J. J., & Piotrowska, A. E. (2020). Population size in particle swarm optimization. Swarm and Evolutionary Computation, 58, Article
100718.
Sahana, S. K. (2022). Ba-PSO: A balanced PSO to solve multi-objective grid scheduling problem. Applied Intelligence, 52(4), 4015–4027.
Salah, M. A., & Hamad, MM. (2018). Materialized view optimal selection for data warehouse quality. International Journal of Engineering and Applied Sciences.
Seeling, M. X., Kreuter, T., Scavarda, L. F., Thomé, A. M. T., & Hellingrath, B. (2021). Global sales and operations planning: A multinational manufacturing company
perspective. PLoS One, 16(9), Article e0257572.
Srivastava, S., & Sahana, S. K. (2020). A survey on traffic optimization problem using biologically inspired techniques. Natural Computing, 19(4), 647–661.
Tu, S., Rehman, O. U., Rehman, S. U., Ullah, S., Waqas, M., & Zhu, R. (2020). A novel quantum inspired particle swarm optimization algorithm for electromagnetic
applications. IEEE Access, 8, 21909–21916.
Wu, X., & Liao, H. (2021). Customer-oriented product and service design by a novel quality function deployment framework with complex linguistic evaluations.
Information Processing & Management, 58(2), Article 102469.
Xin-gang, Z., Ji, L., Jin, M., & Ying, Z. (2020a). An improved quantum particle swarm optimization algorithm for environmental economic dispatch. Expert Systems
with Applications, 152, Article 113370.
Xin-gang, Z., Ze-qi, Z., Yi-min, X., & Jin, M. (2020b). Economic-environmental dispatch of microgrid based on improved quantum particle swarm optimization. Energy,
195, Article 117014.
Yang, J., Xiu, P., Sun, L., Ying, L., & Muthu, B. (2022). Social media data analytics for business decision making system to competitive analysis. Information Processing
& Management, 59(1), Article 102751.
Yao, D., Abulizi, A., & Hou, R. (2015). An improved algorithm of materialized view selection within the confinement of space. In 2015 IEEE fifth international
conference on big data and cloud computing (pp. 310–313). IEEE.
Zhu, A., & Hao, D. (2020). Research on decision support system of ship planning management based on data warehouse. In , 1651. Journal of physics: conference series.
IOP Publishing, Article 012082.

15