Soft Computing (2023) 27:7627–7637

https://doi.org/10.1007/s00500-023-08147-3 (0123456789().,-volV)(0123456789().
,- volV)

FOCUS

Artificial intelligence and automatic recognition application in B2C

e-commerce platform consumer behavior recognition
Tian Xie1

Accepted: 30 March 2023 / Published online: 15 April 2023

 The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2023

Abstract
In recent years, the e-commerce industry has developed rapidly. This is one aspect of the wide application of science and
technology in business. Artificial intelligence technology is now inseparable from human society, and artificial technology
and human society are each other to promote progress. Voice is the carrier of language, and voice recognition is actually
the conversion of voice signals into symbols that can be recognized by the system. Accuracy can be greatly improved, but
also converts a variety of different languages into a universal symbol so that it can be recognized by a computer and its
meaning can be understood. The development of the e-commerce industry did not explode suddenly but went through a
long process. At the beginning of its development, it was mainly through the Internet to communicate with the e-commerce
industry under the universal application of artificial intelligence technology. E-commerce is also combined with new
technologies to continuously enrich the development model. Under this development trend, B2C e-commerce has also been
greatly promoted. However, there are still many B2C e-commerce which retain the traditional business model and cannot
make good use of big data. Therefore, this paper analyzes the current situation of B2C e-commerce platform, introduces
artificial intelligence technology into consumer behavior recognition and analyze the urgent needs of consumer behavior
recognition. For B2C e-commerce companies, in order to improve management efficiency and reduce operating costs, it is
necessary to classify users in a targeted manner and find the target users they need. Combine data mining algorithms to
build data models, this article effectively suggests a set of consumer behavior recognition models to help companies find
consumers quickly.

Keywords Artificial intelligence  Speech recognition  B2C e-commerce platform  Consumer behavior

1 Introduction double-edged sword for human society. While AI brings

technological advances and improves the convenience of
Artificial intelligence has integrated the knowledge of life, it also raises many ethical issues (Lazzeretti et al.
many disciplines, so it is very comprehensive and can be 2022). Therefore, when researching artificial intelligence
applied to various fields (Kokina and Davenport 2017). At issues, to maximize the effectiveness of artificial intelli-
present, the development of artificial intelligence has bro- gence, use artificial intelligence to promote the develop-
ken away from the stage of pure pursuit of breadth and is ment of human society and weaken the harm of artificial
gradually moving to a deeper level in various fields, which intelligence rather than using the powerful artificial intel-
makes artificial intelligence promote the production of ligence to wantonly invade others’ lives and bring threats
many new technologies and new technologies (Chatila to social development (Kulkarni et al. 2022). Cloud
et al. 2017). Artificial intelligence is not omnipotent and recognition technology is also a new scientific technology
perfect. The development of artificial intelligence is a in recent years. The current speech recognition technology
mainly includes a large number of vocabulary continuous
speech recognition systems and miniaturized and
& Tian Xie
[email protected] portable cloud speech recognition systems. The service
targets of these two different recognition systems are dif-
1
Henan Industry and Trade Vocational College, ferent: large vocabulary continuous speech recognition
Zhengzhou 450053, Henan, China

123
7628 T. Xie

systems are mainly used in the Internet, telephone networks kind of universal symbol can be accurately recognized by
or some large-scale voice query service systems, and these the computer and can ensure the accuracy of the informa-
systems are often complex in structure and diverse in tion (Alhawiti 2015). In the information age, science and
levels; and a miniaturized and convenient voice recognition technology have made people’s ways of communicating
system can be applied to mobile terminals (Johnson et al. more and more abundant; natural language is the most
2014; Juang and Rabiner 2005). Now, the ways people diversified delivery method. A convenient way of interac-
communicate with machines are becoming more and more tion in this context, the literature shows that the application
diversified, but how to choose the most convenient way of speech recognition is becoming more and more wide-
from them is a problem we need to consider (MacArthur spread, and it has now been widely used in mobile termi-
and Cavalier 2004). In order to solve this problem, speech nals (Panda 2017). B2C e-commerce is an emerging
recognition systems have begun to move from ordinary e-commerce operation method in recent years. With the
computer platforms to portable platforms such as mobile popularization of online shopping, more and more tradi-
terminals. At present, there are still some shortcomings in tional industries have begun to transform. The literature
the interaction between B2C e-commerce enterprises and discusses the current problems and deficiencies of B2C
customers. To solve these shortcomings, this paper pro- e-commerce companies, summarizes its development sta-
poses a rough set-based data mining method to identify tus, and conducts this research based on rough-level data
consumer behavior (Cao 2021). This method is realized mining methods (Gong-min 2010). Literature introduced
under the premise of exponential increase in data volume what is data mining technology and explained its related
and rapid development of data mining. In this way, we can theories, then described neural network and genetic algo-
help companies find potential customers and effectively rithm according to the k-means algorithm and verified the
establish consumer behavior recognition models. effect of the combined model through calculation. Finally,
it provides theoretical support for enterprises by analyzing
consumer behavior. Literature describes the recognition
2 Related work system in detail based on actual applications. Finally,
according to the purchase behavior of target users, data
According to the current status of artificial intelligence mining technology can be used to analyze data to provide
development, the development of artificial intelligence has theoretical support for enterprises (Hariguna 2020).
pressed the accelerator button for the development process
of human society. The rapid development of artificial
intelligence has brought great changes to human society. At 3 Distributed algorithm and robust speech
the same time, the application of artificial intelligence recognition
technology can effectively improve the efficiency of vari-
ous industries and reduce costs. Artificial intelligence is 3.1 Robust speech recognition
gradually becoming a key technology among countries
competing (Duvnjak et al. 2020). However, artificial The core problem of speech recognition is to solve the
intelligence is a science and technology. He will also bring problem of mismatch between model design and recogni-
certain harm to human society. The power of artificial tion environment and to improve the degree of matching.
intelligence can convey information as quickly as possible This article provides an in-depth analysis of why the match
through data mining and achieve data processing efficiency is not high. The accuracy of speech recognition technology
beyond human reach through machine learning (Tanoli extraction is promoted. Although the causes of the three
et al. 2021). If this ability is used effectively, it can types of noise are different, the principle of eliminating
accelerate the development of information and social pro- these three types of noise is noise robust technology.
gress, but if it is used recklessly, it will bring great adverse Through noise robust technology, noise can be effectively
effects to the society. People should fully avoid the risks reduced to match the training model.
faced by artificial intelligence while developing artificial The environment in which speech is converted from
intelligence so that the advantages outweigh the disad- sound to data is called the acoustic environment. In this
vantages (Zubatiuk and Isayev 2021). Speech is the carrier acoustic environment, there are two interference sources
of language, and speech recognition is actually to convert that affect speech recognition: additive noise and channel
speech signals into symbols that can be recognized by the noise. Additive noise is generally common noise that we
system. This not only improves the efficiency and accuracy can hear in daily life, while channel noise mainly refers to
of recognition but also converts a variety of different lan- noise that is difficult for people to hear, such as current
guages into a universal symbol so that it can be recognized tones or voice coding.
by a computer and its meaning can be understood. This

123
Artificial intelligence and automatic recognition application in B2C e-commerce platform consumer… 7629

Additive noise can be divided into two types: stationary Background channel
noise and non-stationary noise. As the name implies, noise noise

steady noise means that the frequency of the sound is

Ideal
stable and unchanging within a certain period of time. The Clean
+
Speech
+ communication
speech coding
channel
frequency of non-stationary noise changes regularly over
time.
The other Speech decoding
In order to measure the ‘‘how much’’ or ‘‘content’’ of the voice
noise in the speech signal with additive noise, the signal-to- Speech enhancement
noise ratio is introduced, which is defined as: processing
!
XL XL
2 2 Speech Human
SNR ¼ 101 log10 s ½n= d ½ n ð1Þ recognition perception
i¼1 i¼1

From the above formula, we can conclude that the larger Fig. 1 General process of speech enhancement application
the signal-to-noise ratio, the less the noise ‘‘content’’ in the
signal. In order to be able to calculate the noise content in However, in most cases, these two goals are often not
the speech signal more accurately, formula 4 is generally compatible.
rewritten as: The technology for performing speech enhancement
algorithms in the frequency domain is a short-term analysis
y½n ¼ s½n þ k  d½n; k [ 0 ð2Þ
technology. When the short-term analysis technology is
In the formula, k is the control factor, which can adjust applied, the noise in the speech signal will be processed in
the ‘‘content’’ of the noise in the noisy speech signal, and frames, and then, the noise will be transferred to the fre-
the relationship between it and the signal-to-noise ratio can quency domain through FFT transformation. The noise will
be obtained according to formula 4, namely: be processed in the frequency domain, and the estimated
! value of the noise will be increased so as to achieve the
XL XL
2
2 extraction of the pure signal frequency purpose, then per-
SNR ¼ 101 log10 s ½n= ð k  d ½ n Þ ð3Þ
i¼1 i¼1 form the inverse fourier transform to obtain the enhanced
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
! ffi speech signal, as shown in Fig. 2.
u L
u X X L
Suppose the noisy speech yt[n] of the t-th frame after
t s2 ½n= d2 ½n  10SNR=10 ð4Þ
i¼1 i¼1
windowing and framing is expressed as
y1 ½n ¼ s1 ½n þ d1 ½n; 0  n  N  1 ð5Þ
Therefore, when a pure speech signal, background noise
signal, and signal-to-noise ratio are given, we can calculate Among them, St[n] and dt[n] are the pure speech and
the noisy speech signal according to formula 5 and formula noise of the t-th frame, respectively, and N is the speech
7. frame length. Perform short-time Fourier transform on
Speech enhancement is the most common technique for Yt[n] to get
noise reduction. Speech enhancement can shield most of Yi ðwk Þ ¼ Si ðwk Þ þ Di ðwk Þ ð6Þ
the noise in the speech signal and extract important infor-
mation, which can effectively reduce noise and eliminate
noise interference. However, the change of noise frequency
makes it difficult for us to completely eliminate the inter-
ference of noise. In theory, we cannot extract completely
pure voice information from the voice signal. The way that
voice enhancement reduces the impact of noise is not to
eliminate noise directly but to reduce the impact of noise
by enhancing the quality of pure voice. Voice enhancement
can firstly improve voice quality, reduce the impact of
noise, and make voice information clearer, which increases
the listener’s acceptance of voice information to a certain
extent, which can be used for human hearing perception; in
addition, the increase in voice can also enable listeners to
listen. Understanding the content of voice, this effect is
more objective than the previous one, as shown in Fig. 1.
Fig. 2 Speech enhancement processing flowchart

123
7630 T. Xie

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
If the noise suppression function G (wk), also known as 1
G t ðw k Þ ¼ 1  ð13Þ
the gain function, has been obtained at this time, the esti- ut ðwk Þ
mated spectrum of the pure speech can be estimated by the
following formula: Since the power spectrum of speech is non-negative, the
_ posterior signal-to-noise ratio must satisfy Ut(wt) [ 1.
Si ðwk Þ ¼ Gi ðwk Þ  Yi ðwk Þ ð7Þ However, because the frequency of noise is difficult to
Spectral subtraction is one of the earliest methods used determine, there will be a certain deviation. In order to
in speech enhancement. The basic idea of this method is to reduce the influence of the deviation on the conclusion,
directly subtract the average frequency spectrum of noise some documents set a lower limit for it, and Eq. 16 can be
from the frequency spectrum of noisy speech signals. Find rewritten as:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
the power spectrum of the 9 square 1
Gt ðwk Þ ¼ max 1  ;a ð14Þ
jYi ðwk Þj2 ¼ jSi ðwk Þj2 þjDi ðwk Þj2 þSi ðwk Þ  Di ðwk Þ þ Si ðwk Þ ut ðwk Þ
 Di ðwk Þ
When VAD detects that the t-th frame is a non-speech
ð8Þ
frame, use the following formula to re-estimate the noise:
If it is assumed that the speech signal and the noise _ 2 _ 2
    2
signal are both zero mean, and they are not statistically Di ðwk Þ ¼ s  Di ðwk Þ þð1  sÞ  jYi ðwk Þj ð15Þ
correlated, that is to say, the cross term is 0, then Eq. 11
Smoothing between frequency components can effec-
can be approximated as:
tively suppress noise while ensuring the quality of voice
jYi ðwk Þj2 ¼ jSi ðwk Þj2 þjDi ðwk Þj2 ð9Þ information, and enhance the auditory effect. Similarly,
smoothing in time can also reduce distortion, namely:
In Eq. 12, we can estimate by taking the average of the
 2
previous frames of the noisy speech (assuming that the  
 Yi ðwk Þ 
previous frames of the noisy speech segment are all noise), ui ðwk Þ ¼ c  ut1 ðwk Þ þ ð1  cÞ   _  ; 0\c\1
namely: Di ðwk Þ
_ 2 1 MX1 ð16Þ
 
Di ðwk Þ ¼ jYi ðwk Þj2 ð10Þ
M i¼0 Smooth the previous multiple frames: The gain function
of Eq. 17 can also be changed to any order, namely:
Therefore, a more intuitive spectral subtraction can be
!!1=a
obtained according to Eq. 12 and Eq. 13, namely: 1
_ 2 Gi ðwk Þ ¼ max 1  a=2 ; a ð17Þ
_  
Si ðwk Þ2 ¼ jYi ðwk Þj2 þDi ðwk Þ ut
0  2 1
_  where s[n] is the estimated pure speech signal, and there
B  D ð w Þ 
k  C
 i
will be errors between it and the pure speech signal s[n] we
2B C
¼ jYi ðwk Þj B1  2C
ð11Þ
@ jYi ðwk Þj A expect:
e½n ¼ s½n  s^½n ð18Þ
If the posterior signal-to-noise ratio Ut(wt) related to Obviously, e[n] is a random variable, and the basic idea
frequency is defined as: of Wiener filter is to minimize the expected error square,
jYi ðwk Þj2 namely:
ut ðwk Þ ¼  _ 2 ð12Þ   
  E e2 ½n ¼ E ðs½n  s^½nÞ2 ð19Þ
Di ðwk Þ

However, the posterior signal-to-noise ratio does not Use the minimum mean square error criterion to solve
mean the signal-to-noise ratio. The posterior signal-to- Eq. 21 and Eq. 22 and transform to the frequency domain,
noise ratio is a special signal-to-noise ratio. It is not only and finally get
related to time but also to the frequency of the noise, so the jSi ðwk Þj2
gain function of the spectral subtraction can be obtained. Gi ðwk Þ ¼ ð20Þ
jDi ðwk Þj2 þjSi ðwk Þj2
Then the prior signal-to-noise ratio is:

123
Artificial intelligence and automatic recognition application in B2C e-commerce platform consumer… 7631

Fig. 3 Comparison of various features of pure speech and noisy speech (SNR = 0) (1)

jSi ðwk Þj2 is because the configuration of the system has been
ni ð w k Þ ¼ ð21Þ upgraded during the calculation process. The network
jDi ðwk Þj2
system shown in Fig. 5 is of three different types: small,
Figure 3 shows the MFCC of pure speech and noisy medium, and large. These three different types of network
speech. The second two-dimensional feature will change systems require different iteration times, which show that
with time. The image is shown in Fig. 3. In addition, we the iteration times will increase as the network scale
can also see that when pure speech is disturbed by noise, increases.
there is a change in the image. When processed by CMS, As can be seen in Fig. 6, the speed of the algorithm is
the mean value of the features of both pure speech and related to the network scale. As the network scale increa-
noise is 0, but the range of values is different. For example, ses, the speed of the algorithm is getting faster and faster.
in (c), the output features of pure speech, it can be seen that Therefore, the results calculated by algorithms in large-
their value ranges are roughly the same. scale networks are more realistic and reliable.
If the three different feature maps in Fig. 3 are drawn In this section, the dynamic changes of the number of
together, Fig. 4 can be obtained. From Fig. 4, we can see users in the environment will be simulated. In the simu-
that the features of pure speech and noisy speech show lation process, the unit time is cut into several identical
similarities after special processing. Therefore, it can be time slices, and the dynamic correlation algorithm is run in
inferred intuitively that CMVN is more robust than CMS the following two setting environments:
and MFCC. Set I : k ¼ 3 and u ¼ 6:

3.2 Distributed algorithm Set II : k ¼ 5 and u ¼ 3:

Figure 7 shows the operating results of two different
If a state is randomly selected and calculated where the settings under different network scales. It can be seen from
system is running, the convergence of the transmission Fig. 7 that the algorithm changes the total transmission rate
efficiency of the system under different network scales can of the system according to different APs.
be seen from Fig. 5. As the number of iterations increases,
the transmission rate of the system is also increasing. This

123
7632 T. Xie

Fig. 4 Comparison of various

features of pure speech and
noisy speech (SNR = 0) (2)

4 Online shopping consumer identification shoppers over the years so as to provide better services to
these consumers and more similar consumers.
4.1 Analysis of consumption patterns of online
shoppers 4.2 Data collation

The consumption patterns of online shoppers can be ana- The data in this article are all derived from the UCI data-
lyzed based on the consumption records of online shoppers base. The original data set is stored in this database. There
and related data. By analyzing these data, basic information are eight attributes related to online shopping consumers in
and characteristics of consumers can be obtained. After this all the original data sets. Due to space problems, these eight
information, the behavior patterns of consumers can be attributes will not be carried out in this article. To elabo-
analyzed and make inferences. According to the promotion rate, we extracted a total of 4422 pieces of valid data from
model of consumers, we can find that there are certain laws these data sets, and each piece of data extracted contains
to follow in the consumption behavior of online shoppers. important information about online shopping consumers.
According to the consumption characteristics of online Sampling survey is to select a part of the samples from
shoppers, it is possible to speculate on potential online all samples for research. From this part of the samples, the
shoppers to find potential customers for the company and characteristics of the overall sample can be roughly infer-
increase the company’s turnover. In other words, it is red. This method has high practicability and simple oper-
possible to dig out the products or services that online ation, which is more suitable for this kind of research
shoppers prefer from the consumption situation of online experiment with more overall samples.

123
Artificial intelligence and automatic recognition application in B2C e-commerce platform consumer… 7633

(1)12degrees

200
The optimal value

System convergence transfer rate

180
Algorithm 4.1
160
GAS
140
RAS
120
100
80
60
40
20
0
I II III IV V
Network size

(2)15degrees

The system converges the transmission

200
180 The optimal value
160 Algorithm 4.1
140 GAS
120 RAS

rate
100
80
60
40
20
0
I II III IV V
Network size

Fig. 6 Comparison of comparison algorithms under different network

scales

Fig. 5 System transmission rate under different network scales

The sample survey is a part of the whole. Although it

can roughly reflect the characteristics of the whole, it is still
a non-comprehensive survey. When investigating, it is
necessary to select an appropriate algorithm to calculate the
selected sample, explore the characteristics of the sample
through the data, and extend the characteristics to the
whole.
Although comprehensive features can provide a more
comprehensive overview of the characteristics of the
samples, they are not applicable in this article. The main
reasons for the method of sampling survey in this article
are:
(1) When conducting surveys by this method, all sam-
ples are randomly selected, so the uniformity of the
selected samples can be guaranteed and errors can be
reduced.
(2) The drawn samples can represent the whole to a
certain extent. Although there are slight differences Fig. 7 Dynamic algorithm performance analysis
from the whole, these differences will not affect the
final result. (3) The number and attributes of the selected survey
samples meet the requirements of this experiment,

123
7634 T. Xie

Table 1 Comparison table of quantitative attributes of bank data following is an application discussion of the classification
Attributes Attribute value Quantized value
results:

Job Admin, blue-collar, entrepreneur 1, 2, 3

1. Grouping information
Housemaid, management, retired 4, 5, 6 Quantity: 150; ratio: male: female = 1:1; average age: 29;
Self-employed, services, student 7, 8, 9
2. Analysis of online shoppers
Technician, unemployed, unknown 10, 11, 12
Marital Divorced, married, single, unknown 123, 4
Education Secondary, tertiary, 1, 2 (1) This group is more inclined to manage money
Primary, unknown 3, 4 than consumption;
Default No, yes, unknown 1, 2, 3 (2) The group trusts the loan and has already made
Housing No, yes, unknown 1, 2, 3 the loan, so it can provide the group with credit
Loan No, yes, unknown 1,2,3 products at a suitable price according to its real
Contact Cellular, telephone 1,2 situation;
Month Jan, feb, … nov, dec 1,2…11,12 (3) This group does not know much about financial
Poutcome Failure, nonexistent, success 1,2,3
management, but the income is more objective
y Yes, no 1,2
and trusts domestic products. There are very few
financial matters involved in life, and the enthu-
siasm for financial management is not high;
(4) The needs of this group for real life are much
higher than those for virtual life. Compared with
and the errors of the data are within the allowable
consumption and investment on the Internet,
range of the experiment, so the accuracy of the
they prefer to focus on their own business and
experimental results can be guaranteed.
family steadily. They are not interested in wealth
(4) The error of the sampling survey can be calculated,
management products. They prefer practical and
so the error of the sampling survey is small, and the
powerful products when shopping.
accuracy is high.
In summary, we can say that the sampling survey method is 3. Marketing strategy
the most suitable scientific method for this experiment.
In order for the data set to be used by the model, the (1) The main members of this group are women,
category attributes of the data set need to be quantified. The children, and the elderly; compared with the
quantitative comparison table is shown in Table 1. abovementioned types of people, this group of
The quantified sample data are shown in Table 2. people has less demand for life, and they rarely
come into contact with online shopping or
4.3 Data clustering financial management. They can conduct lec-
tures or distribute leaflets. Promote products to
We divided these 226 pieces of data into 3 groups with this group of people;
very significant behavioral characteristics. The center (2) Attract people’s interest through promotion and
points after clustering are shown in Table 3. Table 4 shows lottery so that people can learn about the
the clustering results of bank data. product;
According to the characteristics of the average index of (3) It is not only necessary to have a head office for
variables corresponding to each group combined with distribution but also to expand the scope of
business analysis, the characteristics of each category distribution through authorization, franchising,
obtained by subdivision can be described and named. The etc., and increase sales agency points to increase
advantage of naming is to establish an intuitive concept of turnover.
the group. Table 5 below explains the naming of the actual
group after this clustering.
Through clustering and classification, the different 4.4 Perform attribute reduction according
groups are divided. The main basis of the division is the to the generated categories and train
characteristics of each group, and targeted and personalized the neural network
marketing strategies and methods are proposed. This
method effectively transforms information into capital and The optimized BP neural network can realize the training
becomes the tangible wealth of the enterprise. The of classified data: input the relevant attributes of the online

123
Artificial intelligence and automatic recognition application in B2C e-commerce platform consumer… 7635

Table 2 Quantified sample table of bank data set

A B C D E F G H I J K

30 11 2 3 1 1787 1 1 1 19 10
59 2 2 1 1 0 2 1 3 5 5
39 8 2 1 1 9374 2 1 3 20 5
42 5 1 2 1 16 1 1 1 19 9
78 6 1 3 1 229 1 1 2 22 10
54 10 1 1 1 784 2 2 3 15 5
32 10 3 2 1 360 1 1 1 19 10
27 1 J 1 1 451 2 1 1 16 7
34 2 2 3 1 455 2 1 3 20 6
37 2 2 1 1 427 2 2 3 9 6
32 10 2 1 1 2693 2 1 3 21 5
27 10 3 1 1 769 2 1 3 9 5
… … … … … … … … … … …
53 8 1 1 1 4554 1 1 1 5 2
38 5 3 2 1 623 1 1 1 28 6
61 5 2 1 1 967 1 1 1 20 8
43 11 1 1 1 1577 2 1 1 19 9
42 4 2 3 1 1 2 1 2 28 7
36 5 1 2 1 0 1 1 1 28 8
31 10 3 2 1 1166 2 1 3 3 6
59 5 2 4 1 3534 1 1 1 21 9
29 2 2 3 1 200 2 1 3 30 5
33 10 2 1 1 4790 2 1 1 20 4
44 2 2 3 1 -237 2 1 2 10 7
43 2 2 1 1 335 2 1 3 6 5
42 5 2 2 1 603 2 1 1 5 9
32 5 3 2 1 577 1 1 1 7 8

shopping user, and then the online shopping user can be into the same Excel table. Relevant staff can formulate
classified and output the category of the online shopping personalized marketing methods for different types of users
user. The parameters of this process need to be saved for according to the content of the form so that marketing is
the next use. In addition, according to the classification more in line with the needs of consumers. Since the con-
results, the optimized BP neural network can be used to sumption pattern prediction model is essentially from the
predict the group of consumers and classify consumers. perspective of the nature of consumption, it can also
Through the analysis of the same category and different classify different online shopping consumers. This classi-
categories of consumers, consumer behavior information fication has been widely used in various fields, which can
can be obtained so as to provide companies with reliable effectively distinguish consumer behaviors and improve
information about consumers so that companies can enterprises. Table 6 shows Bank sample set after attribute
understand the main factors that affect consumer behavior reduction.
so as to increase consumers’ Loyalty may enable compa-
nies to change their business model to make it more in line
with consumer needs. In addition, by investigating the 5 Conclusion
businesses that online shopping consumers use when they
consume, it can help companies understand the business From the perspective of the enterprise, consumer behavior
preferences that affect online shopping consumers and recognition can effectively mine customer data, thereby
launch businesses that can better meet consumer needs. helping companies find potential customers. In practical
Therefore, it is necessary to classify users according to applications, we find that data mining will vary depending
their attributes and classify consumers of the same category on the recognition system. For the banking industry,

123
7636 T. Xie

Table 3 Cluster center of bank data set Table 5 Group introduction

Standardized attributes Clustering Grouping Name Description
1 2 3 1 Overall high-end This group of users has been paying
consumer group attention to or buying products
S1 0.37521 0.09122 0.4139 from the bank through the client,
s2 0.41137 0.21366 0.11942 with a high degree of education, no
S3 0.26472 0.22968 0.0986 loans, and zero high-quality users
s4 0.12585 0.25244 0.03848 2 Overall low-end This group of users has no chance of
consumer group having bank client behaviors, and
s5 0.69489 0.09428 0.09428
their education is generally low,
s6 0.05226 0.03595 0.01725 and their overall level is not high
s7 0.9895 0.30638 0.5572 3 Overall mid-end This group of users has had client-
s8 0.08724 0.02966 0.02139 consumer group side behaviors, but some people
s9 0c 5961 0.59688 0)6765 rejected the latest product. They
have a lot of concerns in their
S10 0.27047 0.2082 0.0945 consumption. Yes, their
S11 0.11892 0.38946 0.11187 consumption fluctuates greatly
s12 1.56118 0.09009 024,055
S13 0.29956 0.327] 6 0.12746
s14 0.37504 1.85546 0.37504
s15 0.34188 1.69143 0.34188 Table 6 Bank sample set after attribute reduction
S16 0.4171 2.06356 0.4171 s3 s5 s12 S14 S16 S17
S17 1.9944 0.25663 0.39504
2 1 79 -1 2 2
2 1 226 -1 2 2
Table 4 Clustering results of bank data 2 1 273 -1 2 2
2 1 371 -1 2 2
Category Number of clusters Percentage
2 1 16 -1 2 2
1 27 12% 3 1 317 -1 2 2
2 38 17% 1 1 8 -1 2 2
3 161 71% 3 1 297 144 2 1
2 1 322 -1 2 2
2 1 371 -1 2 2
whether a customer purchases a bank’s financial products 2 1 298 -1 2 2
involves data mining technology and consumer behavior 3 1 101 -1 2 2
recognition. In the current background of fierce interna-
tional competition, the key point for the country to develop
science and technology is to develop artificial intelligence
technology. This is the best era and the worst era. In this
Internet drives the common development of other indus-
era, we are faced with various development opportunities,
tries. B2C e-commerce is a new development model. As
but at the same time, we are also faced with many prob-
the number of online shoppers continues to increase, many
lems. How to effectively solve these problems is the key to
traditional industries are slowly undergoing transformation,
restricting development. Speech is the carrier of language.
and the Internet industry is one of the main goals of the
Speech recognition is to convert speech into symbols that
transformation. Currently, B2C e-commerce companies
can be recognized by the computer. This kind of symbol is
and customers are faced with problems such as lack of
a general-purpose symbol that can transmit and recognize
information and are facing many challenges. In order to
information in the computer, which can greatly improve
solve these problems, this paper proposes the application of
the recognition efficiency and classification efficiency of
rough-based data mining methods in consumer behavior
the computer. It can also effectively enhance human–
recognition. This method is carried out in the Internet
computer interaction. Through this symbol, the computer
where the amount of data has increased greatly, and data
can easily recognize voice information and convert the
mining technology needs to be used to identify consumers’
feedback information into voice to communicate with
consumption behavior. This method can help companies
people, so voice recognition technology has now been
find potential customers, increase their turnover,
widely used. The current era is the era of the Internet. The

123
Artificial intelligence and automatic recognition application in B2C e-commerce platform consumer… 7637

effectively establish a set of consumer behavior recognition Johnson M, Lapkin S, Long V, Sanchez P, Suominen H, Basilakis J,
models, and provide convenience for companies. Dawson L (2014) A systematic review of speech recognition
technology in health care. BMC Med Inform Decis Mak
14(1):1–14
Juang BH, Rabiner LR, (2005) Automatic speech recognition–a brief
Funding The authors have not disclosed any funding. history of the technology development,’’ Georgia Institute of
Technology. Atlanta Rutgers University and the University of
Data availability Data will be made available on request. California. Santa Barbara, 1, 67
Kokina J, Davenport TH (2017) The emergence of artificial intelli-
gence: how automation is changing auditing. J Emerg Technol
Declarations Acc 14(1):115–122
Kulkarni P, Mahadevappa M, Chilakamarri S (2022) The emergence
Conflict of interest The authors declare that they have no conflict of of artificial intelligence in cardiology: Current and future
interests. applications. Curr Cardiol Rev 18(3):46–52
Lazzeretti L, Innocenti N, Nannelli M, Oliva S (2022) The emergence
Ethical approval This article does not contain any studies with human of artificial intelligence in the regional sciences: a literature
participants performed by any of the authors. review. Eur Plan Stud 26:1–21
MacArthur CA, Cavalier AR (2004) Dictation and speech recognition
technology as test accommodations. Except Child 71(1):43–58
Panda SP, (2017) Automated speech recognition system in advance-
References ment of human-computer interaction, In: 2017 International
Conference on Computing Methodologies and Communication
Alhawiti KM (2015) Advances in artificial intelligence using speech (ICCMC), pp 302–306
recognition. Int J Comput Inform Eng 9(6):1432–1435 Tanoli Z, Vähä-Koskela M, Aittokallio T (2021) Artificial intelli-
Cao Y (2021) RETRACTED ARTICLE: Artificial intelligence-based gence, machine learning, and drug repurposing in cancer. Expert
plant environment detection in coastal areas and B2C e-com- Opin Drug Discov 16(9):977–989
merce network marketing. Arab J Geosci 14(11):1–17 Zubatiuk T, Isayev O (2021) Development of multimodal machine
Chatila R, Firth-Butterflied K, Havens JC, Karachalios K (2017) The learning potentials: toward a physics-aware artificial intelli-
IEEE global initiative for ethical considerations in artificial gence. Acc Chem Res 54(7):1575–1585
intelligence and autonomous systems [standards]. IEEE Robot
Autom Mag 24(1):110–110
Duvnjak K, Gregorić M, Gorše M (2020) Sustainable development– Publisher’s Note Springer Nature remains neutral with regard to
an artificial intelligence approach. Manag Res Pract 12(4):18–28 jurisdictional claims in published maps and institutional affiliations.
Gong-min ZHAO (2010) Research on customer loyalty of B2C
e-commerce. China USA Bus Rev 9(5):46 Springer Nature or its licensor (e.g. a society or other partner) holds
Hariguna T (2020) Implementation of search engine optimization exclusive rights to this article under a publishing agreement with the
(SEO) in e-commerce website using on page SEO and off page author(s) or other rightsholder(s); author self-archiving of the
SEO. Int J Adv Trends Comput Sci Eng 9(4):5481–5484 accepted manuscript version of this article is solely governed by the
terms of such publishing agreement and applicable law.

123