Analysis And Prediction Of Purchase Conversion Rate Of Bookstore E-commerce Platform

With the increasingly fierce competition of e-commerce platforms,major e-commerce platforms need a precision marketing method to increase their profits,and research on user preferences and accurate predictions of book popularity are not only important for precision marketing on the platform Improved,and greatly improved the user’s browsing experience.Therefore,this thesis starts from the two perspectives of bookstore e-commerce platform users and books,and deeply studies the feature of bookstore e-commerce platforms to formulate bookstore users’ click-to-purchase conversion rate prediction methods and book popularity predictions.The contribution of this article can be summarized as follows:1.For the feature of users and book groups.First of all,we build a "buyer-bookactivity" BBA three-part diagram of the three elements of the e-commerce platform to describe the overall information space of the book e-commerce platform.And on the basis of BBA,a large-scale feature combination information mining method based on factorization machine is used to construct a complete feature space In particular,aiming at the concealment of rich book text information,this paper re decomposes the book text features,and uses FM combination to mine more abundant features,so as to form a more complete feature space.Secondly,in view of the complexity and diversity of the effective data feature space of BBA for "buyer-book-activities",considering that feature combination cannot fully capture the diversity of sequential features in e-commerce features,this thesis proposes to use the BBA2 vec method to use representation learning A more effective representation of the overall information space of the book e-commerce platform.Finally,a DNN prediction model for user click-to-purchase conversion rate is proposed.2.For the user’s individual time behavior sequence.Aiming at the relationship between the time sequence of users clicked to buy books and users’ knowledge of books,a book purchase prediction model based on deep knowledge tracking is constructed.Book merchandise is different from general merchandise,and users often purchase physical books for knowledge.And the user’s knowledge of a book often determines whether the user buys the book.Based on this idea,we use a recurrent neural network to model the user’s historical behavior,and use the idea of deep knowledge tracking to predict the user’s next purchase rate for a book.Finally,the validity and reliability of the model proposed in this thesis are verified through the experimental data set provided by Chongqing Xinhua Bookstore Group.Experiments show that the proposed model can effectively capture the internal relationship between bookstore features and predict the click conversion rate and purchase rate.