Study On Consumption Prediction And Psychology Modeling For E-commerce Users

Recent years have seen the rising of e-commerce platforms,which brings great convenience to our life.As a result,users are increasingly willing to shop online for items and services they need.While facilitating users to form the new consumption habit,e-commerce platforms store a large amount of user behavior log in their systems.These log has been widely used to profile users in recommender systems for improving the user experience.In order to effectively profile users in recommender systems,we first need to pre-dict the user consumption.Nowadays,e-commerce platforms usually predict it by an-alyzing or modeling users' historical purchase.The process of users' comparison and decision is taken little account of,which is inapplicable with the increasing number of products.To make more effective use of users' comparison and decision information in their behavior log,we propose a choice model based method.Our method integrates users' historical purchase with behavioral sequences and thus it can express user pref-erence and make more accurate predictions.One step further,users' behavior is determined by their psychology.Therefore,the study of consumer psychology makes great benefit to the understanding of user de-mands and providing intelligent services.Traditional consumer psychology researches are usually based on questionnaires,which is not only subjective but also time and la-bor consuming.With the era of big data coming,it becomes possible to model user psychology by using their behavior log.In this thesis,we make a study on user indeci-siveness psychology and propose to model it in a data-driven way.The main contents and contributions can be summarized as below:1)Research on predicting user consumption by choice model.Firstly,we estimate the optimum substitute in each consumption session by a utility function of users' be-havior sequences.Then we build a Latent Factor Based Choice Model(LF-CM)for the consumed items and substitutes.One step further,to make full use of the infor-mation implied in users' comparison and decision,we propose a Latent Factor and Sequence Based Choice Model(LFS-CM).It considers all the items in one session and integrates latent factors with utility function.Finally,we implement our models on a real-world dataset Rec-Tmall and our models are proved to be effective.2)Research on modeling indecisiveness of e-commerce users in a data driven way.Firstly,we propose a method to measure the observed indecisiveness Ds in each behavior session by some sequence characters.Then we propose an Indecisiveness Matrix Factorization(IMF)model to obtain users' and items' latent indecisiveness indexes.Besides,we explore some directions to use indecisiveness for better ser-vices.Finally,we also verify our model by using Rec-Tmall dataset.