Analysis And Research On Predicting User's APP Behavior Based On Bayesian Network And BP Neural Network

Mobile terminal applications show an explosive growth trend,and its storage capacity is increasing constantly.People download and install a large number of interesting applications.On the one hand,people need to organize their applications and files frequently,and need more time to find and choose the applications they want to use.On the other hand,the free mobile app revenue comes from ad mainly,but there are currently no advertising auctions that take into account the user's time involved in the application,although the longer the exposure,the greater the revenue.Knowing how the mobile app is used will help improve the quality of the device experience and provide mobile app developers with more insight.Based on the analysis of various factors of user context,this paper proposes an improved algorithm model based on Bayesian network and linear model to predict the next App that users to use only by using time and latest used App.Finally,using the MDC dataset evaluation,the proposed prediction model achieves an accuracy of 86%with the computational complexity reduced,which is higher than other algorithms.This paper proposes that ad exposure time should be included in the pricing of mobile ad,allowing publishers to adjust campaign engagement based on the effectiveness of their ads and providing a more effective marketplace for ad impressions,while reducing network utilization and device power consumption.In addition,this paper computes the variance analysis and information gain of various factors that affect the users'participation in the application time to finds out the main influencing factors,and proposes a BP neural network prediction model combined with genetic algorithm.Through data experimental verification and performance analysis,the error rate is within acceptable range.Finally,this paper puts forward an error return estimation model to estimate the difference in returns in the case of this prediction model.Because currently mobile users' engagement and exposure times conflict with the validity of current advertisements,if we know the user's participation in the app's session time and include it in the pricing of advertisements,it can benefit all parties of ad ecosystem.In summary,this paper analyzes and researches on the behavior of users using mobile App,and hopes to be able to bring convenience to users as well as contribute more effectively and rationally to the pricing of mobile ads.