This chapter intends to probe into the predictability of consumer behavior classification (CBC) in online virtual stores under the trend of electronic commerce (e-commerce) and provide better consumer services (CS) for online shopping. First, the recurrent neural network (RNN) is expatiated and improved; thereupon, the bidirectional long short-term memory (BiLSTM) algorithm is designed and applied to the CBC; then, the support vector machine (SVM) and naive bayes classifier (NBC) are cited, and a CBC prediction model based on multi-class machine learning (ML) algorithms is implemented. Further, the proposed model is compared with other models from the perspectives of precision, accuracy, F1, and recall; the results signify that the proposed CBC prediction model has presented a 93.95% accuracy, which is at least 4.19% higher than that of other literature algorithms; besides, the performance analysis of network data transmission synchronization reveals that the proposed algorithm outperforms other algorithms with an overall transmission throughput around 1.
TopIntroduction
The widespread adoption of Internet technology has allowed more diversified consumption choices for consumers by creating a new consumer Consumption Pattern (CP) that couples the traditional offline spending mode with online purchases. In particular, Electronic-Commerce (E-commerce) has become one of the mainstream CP. Data statistics show that as of December 2018, Chinese netizens have passed 820 million, with 70% of them having at least one-time Online Shopping Experience (OSE). Thanks to e-commerce platforms, people can easily shop over the Internet after busy work, which greatly reduces people's consumption time cost (Lombart et al., 2020; Wu et al., 2019). Meanwhile, as the Fifth Generation (5G) mobile network and the big data analysis technology continues to be extended, more targeted services can be provided through Consumer Behavior (CB) Classification (CBC) that has become the focus of scientific research and commercial applications.
With the increasing rise of e-commerce, consumer browsing history-based big data acquisition is widely used in User Behavior Analytics (UBA), Emotion Analytics (EA), Marketing Analytics (MA), and Internal Operation Management (IOM), or more specifically, consumer content recommendation, accurate advertising, and Risk Control (RC) (Dzardanova et al., 2018). Internet companies, operators, and banks have strong demand for fine operation, especially, such Internet companies as Tmall, 360buy, and TikTok depend highly on Data Analytics (DA) system support to customize user recommendations. By effectively analyzing the CB data using big data analysis technology, merchandisers can recommend targeted commodities and services to users while ensuring a high User Experience (UE) and safety in shopping, payment, and other scenes (Park et al., 2018). The effective application of big data analysis and Artificial Intelligence (AI) technologies improves the operation efficiency and service quality of e-commerce enterprises, as well as UE in commodity purchase and order payment. As one of the AI algorithms, Machine Learning (ML) has been applied in many fields, such as Data Mining (DM), Text Classification (TC), and medical diagnosis. It has become the mainstream DM and Data Classification (DC) prediction method.
When consumers browse the online goods in virtual stores on e-commerce platforms, a series of CB data is acquired, which are then stored in the background servers of the e-commerce websites. For many Internet enterprises, an effective mechanism to mine voluminous CB data can provide targeted quality services to boost their market share. For example, issuing platform or merchant coupons, building commodity personalized recommendation systems, and pushing messages according to users' browsing data and purchase behaviors are effective CB DM methods to improve UE (Chang et al., 2019; Basalamah et al., 2020). Deep Learning (DL) algorithm is designed to autonomously extract the multi-level features from the massive amounts of raw data in an unsupervised state, and it can well abstract the overall picture of a user's information and provide effective support for further accurate and rapid analysis of interests, consumption habits, and other personalized information in CB (Paolanti et al., 2019). These new CB analysis methods can provide data decision-making basis for product sales and operation of online virtual stores, improve user stickiness and transaction rate of commercial applications, but also help to optimize product design, measure and improve online UE, and improve product competitiveness, which is of great significance to Internet enterprises.
To sum up, under the vigorous development of e-commerce today, providing a better UE in the virtual Internet space is of great practical value to social and economic development. Innovatively, the CB sequence is statistically analyzed from the time dimension; by introducing and improving the Recurrent Neural Network (RNN), the Bidirectional Long Short-Term Memory (BiLSTM) algorithm is designed and applied to the CBC. At the same time, Support Vector Machine (SVM) and Naive Bayes Classifier (NBC) are introduced. Finally, a CBC prediction model based on the fusion multi-class ML algorithms is implemented, and its performance is evaluated through case analysis, providing empirical guidance for CBC intellectualization in online virtual stores.