Random Forest And Application In Consumer Behavior Of Mobile Users

With the rapid increase in the business type and user quantity of China Mobile,it is very important to scientifically divide the users and analyze the habits of consumer behavior,and to provide appropriate business products targeted.Thus,we propose a set of feature extraction method based on rough set for high-dimensional data and combine the random forest algorithm to construct the Chinese mobile user classification model.Compared with the classical random forest model,Which indicates that the prediction accuracy and classification accuracy of the model are greatly improved.At the same time to the actual user test,confirmed that this method is more suitable for China Mobile’s actual objective needs.Based on the theory of random forest,this paper analyzes and summarizes its definition,nature,classifier,generalization error and classification performance,and concretely demonstrates the convergence,generalization error and OOB estimation.Then,based on the data of Chinese mobile users,the feature extraction algorithm is used to reduce the dimension processing of high-dimensional data of mobile users.the user classification model is constructed according to the random forest algorithm,and the consumer behavior habits of all kinds of users are analyzed according to the classification results.The experimental results show that the classification accuracy of the model is 81.12%,Compared with the classic random forest model,the accuracy rate also increased by 3%-4%.Which proves the feasibility and scientificity of the model constructed in this paper.For the final classification results,study their habits of consumer behavior.High-value high-traffic users: mainly distributed in the county’s streets;an average of 3 months calls more than 50 yuan;call longer than 60 minutes;average 3 months flow is greater than 900 M.So in order to meet the dual needs of calls and traffic,for them to recommend the global package.High-value low-traffic users: mainly distributed in the county streets and nearby towns;monthly call costs more than 50 yuan;monthly calls longer than 60 minutes;the average flow of less than 100 M.Based on this,for its recommended data for Data card and Shenzhouxing card.Low value low traffic users: Since there is no strong demand for calls and traffic,they only want to be able to keep the phone and SMS,So as to recommend the public card.The whole article ideas for China Mobile’s business decision-making provides a new way of thinking.