Research On Mobile Network Buying Prediction Of One E-commerce Platform

With the development of the Internet and mobile terminals,online shopping is becoming more and more popular.Businesses or platforms can not face-to-face communication with users,which led to businesses can not better grasp the user's attitudes and needs,thus facilitating the transaction.However,with the development of science and technology,the behavior of users on the Internet will be recorded by memory.By mining these historical data,the prediction of purchasing decision can be realized.In this paper,from the perspective of data mining,we use the method of machine learning to obtain the influencing factors of a e-commerce user's purchasing decision and predict its future buying behavior。The results obtained in this paper can be applied to the recommender system of e-commerce platform,to achieve the accurate recommendation of users to meet the needs of users shopping and motivate purchase desire,improve the conversion rate and optimize the storage.In this paper,we use SQL language to pull historical data of a real user of an e-commerce business,construct derivative features based on business experience,and use Relief algorithm to extract the top 10 features with larger weights for feature selection.Based on this result,we use the machine learning classification algorithm-logistic regression,CART decision tree and naive Bayesian to construct forecasting model.Finally,it is concluded that for the factors affecting the user's purchasing decision-making,the sharing behavior in the course of shopping behavior has a significant impact on the category of the product,and the CART decision tree is more effective in predicting the accuracy.