Research On E-Commerce Purchase Prediction Based On User Behavior Data

With the fast development of e-commerce and online shopping,people's daily transactions rely more on the Internet.Online shopping breaks the barriers of time and space,buy it also brings problems for sellers.A seller cannot judge a customers purchasing needs and preference to take specific marketing strategies like traditional offline shopping as he cannot see the buyer.But the way data is stored and used have dramatically change with the coming of big data era.Nowadays every transaction even every click buy the customer are stored inside server logs.These data make it possible for e-commerce sites to reproduce the decision-making process of every customer,using history data to predict further purchases becomes feasible.In this paper,we first review the literature of user purchase behavior prediction in the past,and thoroughly understand and analyze the basic way of purchase prediction for e-commerce platform.Then we summarized the methods data acquisition and how data are used in e-commerce sites marketing.Then analyzes the application of the existing big data marketing tools in enterprises taking JD as an example,and points out the blank in the application of big data marketing.Next,using real user behavior data as input data,we build a set of features for prediction.In turn,a double model is constructed to predict consumer demand and commodity preference,and the results of the dual model are fused to produce the final predictions.Using the independent as test data,the test results shows the fusion model in this paper have a great improvement compared with the single model.This paper builds a model using real data to prove the feasibility of using user behavior data to predict user purchases.The prediction results show the model have practical significance for the promotion of services and marketing for e-commerce sites.