Study And Application Of Customers' Shopping Intention Prediction Model Based On Conversations With Customer Service Staff

With the development of the Internet,e-commerce is gradually replacing the traditional retail industry.Online shopping has become a part of public life,and the service center of merchants has begun to shift from goods to customers.How to use limited time to effectively communicate with customers,understand customer needs,and try to stimulate customers' shopping wishes is an important means for businesses to improve their competitiveness.In most online shopping scenarios,customers often want to get timely answers to product inquiries,otherwise they may turn to other businesses.Therefore,in the massive customer service session data,mining customers with potential shopping intentions(potential shopping customers)in real time to provide them with key services is very important for improving customer service efficiency and enterprise efficiency.At present,there are many studies to predict customers' willingness to shop based on their historical behavior(historical purchase records,browsing records,etc.),and few studies to achieve online prediction based on customer service sessions.Compared with the latter,the prediction model based on the customer's historical behavior has the following major deficiencies:(1)poor real-time performance;(2)it is difficult to make accurate judgments on the purchase wishes of new customers;(3)each customer service consultation of the old customer is not necessarily All have the willingness to shop;(4)Customers 'willingness to buy specific products often transfers;(5)Real-time prediction of customers' willingness to shop based on customer service sessions is one of the key functions of intelligent customer service.To this end,based on the real-time customer service session data of the e-commerce platform,this paper studies how to build an online prediction model of customer shopping willingness to realize the real-time mining of potential shopping customers.Its main work includes:(1)Based on the needs of a provincial telecommunications company WeChat Business Office to online predict customer shopping wishes,targeted research and analysis of relevant concepts,methods,algorithms,and technologies are conducted.(2)Based on the real-time customer service session data of the e-commerce platform,analyze its characteristics in detail,and study related pre-processing and labeling methods.(3)Based on deep learning technology,a shopping willingness prediction model based on Word2Vec,Transformer,BiLSTM,Attention,and CRF is given and tested and verified.(4)Combining with actual application scenarios and researching and applying advanced parallel computing architecture and technology for customer service session data flow,a scalable caching algorithm is designed to realize parallel processing of customer service session data and effectively improve processing efficiency.(5)In order to verify the research results of this paper and make it practical,a design scheme of an online customer shopping willingness prediction system based on customer service sessions is given,implemented and tested.