Research On User Behavior Prediction Model And Its Application Based On Deep Walk And Ensemble Learning

In recent years,with the rapid development of e-commerce platforms such as Taobao,JD.com and Pinduoduo,online shopping has become an indispensable part of people's life.These e-commerce platforms store and retain hundreds of millions of purchase records through databases.How to use machine learning technology to predict the future purchase intention of consumers from these data is the key to the whole e-commerce platform using artificial intelligence and big data technology to provide personalized services to consumers.Therefore,this paper uses machine learning algorithm to build user behavior prediction model.This study is based on the bank APP data.In the previous article,the paper preprocessed the problems of duplication,anomalies and redundancy in the data set.From the pre-processed data set,statistics were extracted that reflected the behavior habits and preferences of consumers Information and activity information,in order to build a user portrait for the user,laid a good foundation for the prediction of the following model.Next,this paper proposes a user behavior prediction model based on Deep Walk.A new behavior sequence is obtained by randomly strolling from the social network graph structure of the user's purchase of goods,and then the Word2 vec model is used to obtain the upper and lower information of each behavior of the user.Training and learning in the model to further improve the learning ability of the model.Aiming at the problem of how to select a single model with large differences in ensemble learning,this paper proposes a selective ensemble learning method based on MIC and confusion matrix,through which the larger models with different differences are selected for Bagging fusion.Finally,the experimental results show that the model learning ability by constructing user portraits is better than the basic user behavior prediction model.After the Deepwalk technology is added to the user behavior prediction model,the expression ability of the model is further enhanced.The selective ensemble learning method based on MIC and confusion matrix can intuitively find a single learner with relatively small similarity,and finally carry out model fusion,the fused model has better learning ability than a single model.