Current Research in Behavioral Sciences 2 (2021) 100051

Contents lists available at ScienceDirect

Current Research in Behavioral Sciences

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

An advanced intelligence system in customer online shopping behavior and

satisfaction analysis
Nazmun Nessa Moon, Iftakhar Mohammad Talha∗, Imrus Salehin
Department of Computer Science and Engineering, Daffodil International University, Dhaka, Bangladesh

a r t i c l e i n f o a b s t r a c t

Keywords: Online shopping or internet shopping is increasing day by day. With the advancement of modern technology, the
Online shopping online market is growing in a vast way. People nowadays prefer online shopping because it saves time, energy,
Behavioral impact and money. Because of the blessing of the internet that online shopping has made its debut which also affects the
Shopping impact
common citizens for online shopping. So, for the emerging growth of the online market, it is necessary to find
Customer satisfaction
out the behavior of online shopping and customer satisfaction. Safety, trust, product quality plays an important
role in customer satisfaction. In this study, we examined customer online shopping satisfaction its impact. The
quality of the product, the price of the product compared to the local market, the policy of return, timely delivery
of the product are also essential elements of online shopping. By analyzing all these factors we have tried to find
the customer behavior and satisfaction with online shopping. In our study, we used a machine learning method
to search for the result. We used 40 thousand data to find out the accuracy of our work and to analyze customer
shopping satisfaction. We use Naïve Bayes, Apiorir, Decision Tree, and Random Forest classification algorithms for
this analysis. We have got our best result by using Apiorir algorithm (88% accuracy) and Naïve Bayes algorithm
(87% accuracy). We have also focused on customer behavior and interest in online shopping. Our study can help
to develop the business intelligence and satisfaction enhancement about E-commerce.

1. Introduction will be very high and how they will get this product. Machine learning
methods provide powerful tools for extracting data from large and noisy
There are many small markets in the tech market. Technology in- databases. Machine learning provides an opportunity for researchers to
cludes everything in it and makes our lives more reliable and faster. better understand consumer behavior and improve the effectiveness of
Technology has a significant impact on business activities. Technical marketing activities. In the technology sector or for developing any sys-
infrastructure affects the culture, skills, and relationships of an orga- tem big data mining and machine learning rapidly expanding (Qiu, Wu,
nization. It also affects the protection of confidential information and & Ding, 2016). In the retail industry data management, examining the
business facilities. (Ngai, Hu, Wong, Chen, & Sun, 2011) (E.W.T. Ngai et impact of customer shopping behavior, the performance of that organi-
al.2010) In this article, the author represents the method, place, and aca- zation, decision-making strategy, operations across offline and online all
demic literature of the world. Apply FFDT by checking article browsing is being done by using big data mining and machine learning (Ying, Sin-
steps. One of the biggest benefits of technology companies is increasing dakis, Aggarwal, Chen, & Su, 2020). In our research, we tried to find out
productivity. (Moro, Rita, & Vala, 2016) (Moro et al. L. 2066) in their customer satisfaction on online shopping and changes in online shop-
article describes how to use data processing to predict the performance ping behavior. Through our work, we tried to build a structurally shaped
parameters of publications published on brands’ Facebook pages. Nowa- advance intelligent system (Salehin et al., 2020) that can find out cus-
days, online shopping is a fast-growing situation and the online market tomer shopping behavior and satisfaction rate in online shopping.
is a fast-growing marketplace. Progressive numbers of consumers shop In this study, our main contribution, (1) is primarily a focus on data
online to purchase their goods. Online services and the quality of that mining and machine learning approaches to find out customer online
products are very much important for online shopping. Online shopping shopping behavior changes and satisfaction. (2) Product quality case
environments are therefore playing an increasing role in the overall re- forecast. (3) Product price cost analysis. (4) Combination tables of differ-
lationship between marketers and consumers. Most people are afraid of ent products using data mining. (5) Give a visualization and a precision
online shopping because they think the product is not good or the price of the result.

∗
Corresponding author.
E-mail address: [email protected] (I.M. Talha).

https://doi.org/10.1016/j.crbeha.2021.100051
Received 22 May 2021; Received in revised form 2 June 2021; Accepted 1 August 2021
2666-5182/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)
N.N. Moon, I.M. Talha and I. Salehin Current Research in Behavioral Sciences 2 (2021) 100051

2. Related work Table 1

Data Collection.
Online Marketplace is one of the emerging marketplaces in the mod- Websites Frequency Percentage
ern world. During this epidemic situation, the online market has be-
daraz.com 60 14.53
come so popular. Many researchers have developed many models for
bikroy.com 56 13.56
online market analysis and monitoring. A theoretical model used to in- Others 137 33.17
vest in consumers has four effects: confidence in Internet shopping, a rokomari.com 64 15.5
major form of consumer e-commerce from business: the credibility of ajkerdeal.com 36 8.72
ekhanei.com 26 6.3
the Internet merchant, all of which are most important to consumers
othba.com 31 7.51
(Lee & Turban, 2001). The internet is a more helpful shopping medium amazon.com 104 23.12
for retail shoppers. In recent times, online stores offer more significant
services and save more time. Saving time is a major factor in people’s
tendency to buy on the Internet (Bhatnagar, Misra, & Rao, 2000). Lack
B under the condition in which minimum support and confidence are
of consumer confidence, distrust of the product is one of the major bar-
met (Apriori Algorithm In Data Mining 2021).
riers to online shopping. But there is no doubt about the relationship be-
tween knowledge and quality. Some argued for a negative relationship 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑎𝑛𝑠𝑎𝑐𝑡𝑖𝑜𝑛 𝑖𝑛 𝑤ℎ𝑖𝑐 𝐴 𝑎𝑝𝑝𝑒𝑎𝑟𝑠
𝑆𝑢𝑝𝑝𝑜𝑟𝑡(𝐴) = (1)
between knowledge and quality and others argued for a positive rela- 𝑇 𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑎𝑛𝑠𝑎𝑐𝑡𝑖𝑜𝑛𝑠
tionship. In this study, they discussed the relationship between knowl-
𝑆𝑢𝑝𝑝𝑜𝑟𝑡 (𝐴 ∪ 𝐵 )
edge and quality (Wang, Cheng, & Jiang, 2009). Understanding cus- 𝐶𝑜𝑛𝑓 𝑖𝑑𝑒𝑛𝑐𝑒 (𝐴 → 𝐵 ) = (2)
tomer satisfaction in online shopping is an essential part that plays an 𝑆𝑢𝑝𝑝𝑜𝑟𝑡(𝐴)
important role in increasing and decreasing the level of online shop-
4.2. Naive Bayes
ping (Karim, 2013). In this research, they used the Kotler and Killer
(2009) five-step purchasing process framework model to test customer
Naïve bayes algorithm is a classification algorithm which is used for
satisfaction when buying products online. In their survey, there were
solving classification tasks. It is a probabilistic machine learning algo-
932 participants for their survey. They have used the Technology Ac-
rithm based on the Naive Bayes theorem. It is a classification technique
ceptance Model (TAM) which is valid for predicting the acceptability
and it assumes that the presence of a particular feature in a class is unre-
of Internet shopping malls and the online and offline features have a
lated to the presence of any other feature [(6 Easy Steps to Learn Naive
positive effect on user acceptance (Ahn, Ryu, & Han, 2004). This pa-
Bayes Algorithm with codes in Python and R 2021) (Classification Meth-
per contains descriptions of many underlying structural and functional
ods in Machine Learning 2021)]. Bayes theorem provides a way of cal-
weaknesses related to Internet retailing. Various strategies designed to
culating posterior probability P(c|x) from P(c), P(x) and P(x|c). Look at
improve the effectiveness of retailers have been explored (Grewal, Iyer,
the equation below:
& Levy, 2002). In this study, they focus on big data mining techniques
in managing B2B customer relationships. They examine the effect of big 𝑃 (𝑥𝑐 )𝑃 (𝑐 )
𝑃 (𝑐𝑥) = (3)
data on customer relationship performance and sales growth, They have 𝑃 (𝑥 )
used multi-industry datasets from 417 B2B firms (Hallikainen, Savimäki,
& Laukkanen, 2019). Big data analysis, giving reshape in the business 𝑃 (𝑐𝑥) = 𝑃 (𝑥1𝑐 ) ∗ 𝑃 (𝑥2𝑐 ) … … .𝑃 (𝑥𝑛𝑐 ) ∗ 𝑃 (𝑐 ) (4)
industry. The fundamental key to marketing is people, product, place,
price, and promotion. Data mining provides a great impact on improv- Here, P(c|x) represents the posterior probability, P(c) is the prior prob-
ing these areas and big data analysis is an important resource for making ability and P(x|c) is the likelihood which is the probability of predictor
better decisions (Fan, Lau, & Zhao, 2015). given class, P(x) is the prior probability of predictor.

3. Proposed model 5. Data collection

This section discusses our rapid structure model and the develop- Analysis of various data is a necessary element for this study. Because
ment of its business algorithm. From Fig. 1, the analysis process will if marketers and website developers can identify patterns or trends in
start with a single frame and a multi-frame. Using Machine learning consumer search information, it will be more effective and easier to
and classification algorithms. This model trains its intelligent system monitor the behavior of consumer search changes.
through training datasets. After completion of the training, it will pro-
duce output results as Satisfied, Unsatisfied, Moderate, and Never with a 5.1. Data analysis and pre-processing
percentage measurement. Then we get the accuracy of the model using
the Classification algorithm. For this study, we tried to collect many types of data. We collect
review data, online survey data, customer comments from customers.
4. Research methodology We use Google forms to collect data for this analysis from online par-
ticipants, from websites, and even physically collected data. We have
In this section, we describe how the research conducted. To com- collected 40,000 thousand of data, including customer comments, cus-
plete our research, we have used apirori algorithm and Naive Bayes tomer reviews, feedback data, and survey data using Google Sheet. We
algorithm. There are six sorts of online customers. have use Python programming language for the analysis of our data.
After collecting all the data we use string lower() method to convert
4.1. Apiorir algorithm textual data into lowercase data. We also use the regex model for split-
ting the document into words. Then we remove all the stop words from
In data mining FPM (frequent pattern mining) algorithm is one of the our dataset. It helps to focus on the actual raw data and gives better
most used technique. This technique discover the relationship between accuracy.
different items in a dataset. All those relationships are represented in the Table 1 shows the different types of online shopping websites where
form of association rules. To find the data irregularities association rule we collect most of our data. About 33.17% are not using their mar-
mining is used. An association rule, A=> B, will be of the form, for a set ket for online shopping rather than they prefer physical shopping. Most
of transactions, some value of itemset A determines the values of itemset people are using daraz.com, bikroy.com, rokomari.com, amazon.com

2
N.N. Moon, I.M. Talha and I. Salehin Current Research in Behavioral Sciences 2 (2021) 100051

Fig. 1. Working Process Mode.

Fig. 2. Website analysis data.

Fig. 3. Male & Female participants.

(Fig. 2). 8.72% are using ajkerdeal, 7.51% are using othoba, 6.3% are From Fig. 4, we can see that age from 15 to 25 five years old people
using ekhanei.com for their shopping purposes. prefer online more than other people. 26 to 44 aged people are 33.66%
During the data collection period, we found that most of the par- and other ages are 8.23%.
ticipants were male. Men are more interested in online shopping than To complete our research, we need to know that the participants con-
women. From Fig. 3, we can see that 71.67% are the male customers tinuously purchase online or it is temporary. Fig. 5, shows that most of
and 28.33% female customers. the people around 52% people buy online but it not continuous, 31.23%

3
N.N. Moon, I.M. Talha and I. Salehin Current Research in Behavioral Sciences 2 (2021) 100051

Fig. 4. Age Group.

Fig. 5. Purchased history of customers.

Fig.: 6. what purchased from online.

people buy online continuously, and 16.46% are never shopped from the 6. Customer satisfaction analysis
online market (Erevelles, Fukawa, & Swayne, 2016).
To know the online shopping behavior, we collect what kinds of In this section, we have visualized our main satisfaction data graph
things people buy online. From Fig. 6, we can see that what type of prod- and its calculation. This all data variant we collect from our special data
ucts are bought from online most. 15.98% people are never shopped, bank and online best shopping sites as well. This data is highly protected
28.18% are shopped for Clothes, 18.4% are shopped for other things. data for security purposes.

4
N.N. Moon, I.M. Talha and I. Salehin Current Research in Behavioral Sciences 2 (2021) 100051

Fig. 7. Time Graph.

Fig. 8. Quality Graph.

6.1. Order delivery time Table 2

Algorithm Accuracy Rate.
For online shopping, this section is essential for purchasing any prod- Algorithm Accuracy
uct. So that we analyze this kind of data. In Fig. 7, we calculate the data
Decision Tree 82.69%
values of different organizations.
Random Forest 84.25%
After the analysis, we found that the on-time delivery rate in amazon Apirori 87.89%
is very high, where other online sites are not so fast enough to fulfill this Naïve Bayes 88.44%
category. Therefore, the data graph fluctuates all different companies.
Timely delivery of products makes the customer happier, more confi-
dent, and interested in online shopping. policy then the customer’s trust level automatically increases. In Fig. 9,
refund policy and customer’s feedback rate are displayed.
In this graph, Daraz gets the high pick to this refund policy from
6.2. Product quality
the customer feedback. They earn the best refund policy company and
customer support. However, other companies are below average in this
Ensure of product quality plays a vital role in online purchases.
sector so the graph line has decreased here as well.
Therefore, it is a major factor in online satisfaction prediction analy-
sis. Sometimes customers are not satisfied with their online purchase
products because of the product quality. In addition, some companies 7. Result
attempt fraud against their customers.
Fig. 8 is the quality graph of many different companies with customer To find out the result we have focused on main attributes such as
feedback. We calculated customer feedback and the main issue about customer Online Purchase and Customer online product purchase in-
product quality. In this graph, we see the product’s quality level. If the terest. Therefore, we collect many different types of data and make a
product quality is satisfactory then the selling rate will increase more data analysis result (Fig. 10) using different types of machine learning
than other companies. algorithms (Nazmun et al., 2021). After all the visualization, we make
a satisfactory decision about the online customer satisfaction rate.
In this analysis, we mainly focused on Naïve Bayes and Apirori algo-
6.3. Refund policy rithm. From those algorithms, we have got our best accuracy values are
88% and 87%. For better accuracy and more testing, we have also used
Product Refund Policy is helping a company reach the top Market Decision tree and Random forest algorithm. From those algorithms, we
place. The maximum time a customer focus on companies refund pol- have also got better accuracy that are 82% and 84% shown in Table 2.
icy. If any company assigned a money-return policy and product refund Selective algorithm train and test selective result as follow as:

5
N.N. Moon, I.M. Talha and I. Salehin Current Research in Behavioral Sciences 2 (2021) 100051

Fig. 9. Refund Policy Graph.

line buying satisfaction and their online shopping behavior (Kau, Tang,
& Ghose, 2003). The main point of this paper was to show the current
situation of internet shopping and customer satisfaction. The sensitive
impact of buyers is displayed in this analysis. Different kinds of algo-
rithms are used to analyze the customer shopping behavior and satis-
faction of online shopping.

Declaration of Competing Interest

The authors declare no conflict of interest.

References

Ngai, E.W.T., Hu, Y., Wong, Y.H., Chen, Y., Sun, X., 2011. The application of
data mining techniques in financial fraud detection: A classification framework
and an academic review of literature. Decis. Supp. Syst. 50 (3), 559–569.
Fig. 10. Result generate life cycle. doi:10.1016/j.dss.2010.08.006.
Moro, S., Rita, P., Vala, B., 2016. Predicting social media performance metrics and eval-
uation of the impact on brand building: a data mining approach. J. Bus. Res. 69 (9),
3341–3351. doi:10.1016/j.jbusres.2016.02.010.
For the satisfaction analysis, we work with some specific data vari- Qiu, J., Wu, Q., Ding, G., et al., 2016. A survey of machine learning for big data processing.
ations. Our main data types are order delivery time, authentic product, EURASIP J. Adv. Signal Process. (2016) 67. doi:10.1186/s13634-016-0355-x.
Product quality, customer support, and refund policy. We collected all Ying, S., Sindakis, S., Aggarwal, S., Chen, C., Su, J., 2020. Managing big data in the retail
industry of Singapore: Examining the impact on customer satisfaction and organiza-
data variants and apply for getting the final accuracy. After the data
tional performance. Eur. Manag. J. doi:10.1016/j.emj.2020.04.001.
level accuracy, we ensure the specific company or specific product pur- Salehin, Imrus, Talha, Iftakhar Mohammad, Moon, Nazmun Nessa, Saifuzzaman, Mohd.,
chase satisfaction. We also create a data bank for future prediction Nur, Fernaz Narin, Akter, Mariom, 2020a. Predicting the impact of the human brain
(Salehin et al., 2020) rate calculation. To use data science and customer and depression as a result of the use of technical devices with machine learning.
(In press) 2020 International Conference on Sustainable Engineering and Creative
purchase behavior (Talha et al., 2020) combination we get a formal de- Computing (ICSECC). President University 16-17 December(In press).
cision of customer online shopping satisfaction rate. Lee, M.K.O., Turban, E.L., 2001. A trust model for consumer internet shopping. Int. J.
Electron. Commerce 6, 75–91 2001.
Bhatnagar, A., Misra, S., & Rao, R. H. (2000). ‘On risk, convenience and internet shopping
Conclusion behavior’, association for computing machinery. Communication of the ACM; Nov
2000 43,ii,ABI/IN FORM Global pg.98.
Technology is making everything easier for us. Online shopping is Wang, C.C., Cheng, C.A., Jiang, J.C., 2009. ‘the impact of knowledge and trust on E-con-
sumer’s online shopping activities: an empirical study. J. Comput. 4 (1) January.
one of these things. Which has greatly reduced both our time and ef- Karim, R.A., 2013. Customer satisfaction in online shopping: a study into the reasons for
fort. During the pandemic situation, online shopping caught more at- motivation and inhibitions. IOSR J. Bus. Manag. 11 (Issue 6), 13–20 Jul-Aug.
tention. Internet shopping refers to systems in which customers pur- Ahn, T., Ryu, S., Han, I., 2004. The impact of the online and offline features on the user
acceptance of internet shopping malls. Electron. Commerce Res. Appl. 3, 405–420.
chase products through the online market. Using a personal computer
Grewal, ., Iyer, .R., Levy, M., 2002. Internet retailing: Enablers, Limiters and Market con-
or some other intelligent device. Online markets and websites are widely sequences. J. Bus. Res. 5743, 1–11.
developed nowadays and the dependency on web technology increases Hallikainen, H., Savimäki, E., Laukkanen, T., 2019. Fostering B2B Sales
with Customer Big Data Analytics. Industrial Marketing Management
mostly (Liu & Arnett, 2000). The quality and delivery time of the prod-
doi:10.1016/j.indmarman.2019.12.005.
ucts is the key to online shopping. Consumer intent, security, privacy, Fan, S., Lau, R.Y.K., Zhao, J.L., 2015. Demystifying big data analytics for business
trust all these dimensions have an adverse impact on online shopping intelligence through the lens of marketing mix. Big Data Res. 2 (1), 28–32.
(Ranganathan & Ganapathy, 2002). Building trust issues more secure doi:10.1016/j.bdr.2015.02.006.
Apriori Algorithm In Data Mining: Implementation with examples” [Online]. Avail-
payment gateways, and money-return guarantee policies also include able: https://www.softwaretestinghelp.com/apriori-algorithm/ Access Date: 20.02.
reasons for user satisfaction. All those factors are related to customer on- (2021).

6
N.N. Moon, I.M. Talha and I. Salehin Current Research in Behavioral Sciences 2 (2021) 100051

“6 Easy Steps to Learn Naive Bayes Algorithm with codes in Python and R” [Online]. Talha, I.M., Salehin, I., Debnath, S.C., Saifuzzaman, M., Moon, N.N., Nur, F.N., 2020.
Available: https://www.analyticsvidhya.com/blog/2017/09/naive-bayes-explained/ Human behaviour impact to use of smartphones with the python implementa-
Access Date: 15.03. (2021). tion using naive Bayesian. In: 2020 11th International Conference on Computing,
Classification Methods in Machine Learning [Online]. Available: https://medium.com/ Communication and Networking Technologies (ICCCNT), pp. 1–6. doi:10.1109/ICC-
@jorgesleonel/classification-methods-in-machine-learning-58ce63173db8 Access CNT49239.2020.9225620.
Date: 29.03. (2021). Liu, C., Arnett, K.P., 2000. Exploring the factors associated with website success in the
Erevelles, S., Fukawa, N., Swayne, L., 2016. Big Data consumer analyt- context of electronic commerce. Inf. Manag. 38 (2000), 23–33.
ics and the transformation of marketing. J. Bus. Res. 69 (2), 897–904. Ranganathan, C., Ganapathy, S., 2002. Key dimensions of business-to-consumer websites.
doi:10.1016/j.jbusres.2015.07.001. Inf. Manag. 38 (2002), 457–465.
Nazmun, Nessa Moon, Imrus, Salehin, Masuma, Parvin, Md. Mehedi, Hasan, Iftakhar, Mo- Kau, A.K., Tang, Y.E., Ghose, S., 2003. Typology of online shoppers. J. Consumer Mark.
hammad Talha, Sushanta, Chandra Debnath, Fernaz, Narin Nur, Mohd., Saifuzzaman, 20 (2), 139–156 2003.
2021. A natural language processing based advanced method of unnecessary video de-
tection and the framework model. on International Journal of Electrical and Computer
Engineering (IJECE), 11 DecemberIndonesia.
Salehin, I., Talha, I.M., Mehedi Hasan, M., Dip, S.T., Saifuzzaman, M., Moon, N.N.,
2020b. An artificial intelligence based rainfall prediction using LSTM and
neural network. In: 2020 IEEE International Women in Engineering (WIE)
Conference on Electrical and Computer Engineering. WIECON-ECE, pp. 5–8.
doi:10.1109/WIECON-ECE52138.2020.9398022.