Analysis Of Network Structure Of Mobile APP Conversion And Behavioral Prediction Model

With the rapid development of the Internet,people's behavior activities have extended from the physical space to the virtual space.This has also brought a series of problems,such as:network supply and demand conflicts,public opinion control,and network security.The development of information technology has made it possible for the data of people's behavior activities in virtual space,providing an opportunity to study the behavioral relationships and intrinsic laws of human beings in physical space and virtual space.In-depth understanding and prediction of user behavior in virtual space can help guide network infrastructure planning,network activity management,etc.,have important implications for building a stable and friendly network environment,and have references value for solving a series of virtual space problems.Mobile Internet is one of the main ways for users to behave in virtual space.APP switching is actually a change of users' behavior in virtual space.This paper studies the mobile phone APP conversion network and behavioral transformation network,analyzes the network structure characteristics and online time characteristics,and establishes the user behavior prediction model in virtual space.Specifically,the following research contents are included:(1)The Conversion of Mobile Internet APP and the Conversion of Online Behavior was analyzed based graph.Conversion network of mobile Internet surfing APP and conversion network of mobile Internet behavior have constructed,based on mobile Internet data.Then we analyze the structural characteristics of the network,and propose indicators for evaluating the relationship between behaviors.It is found that the distribution of in degrees,out degrees,and degrees obeys power-law distribution,and a large number of conversions occur between a small numbers of node APPs.The behavior of social communications is most central,and the centrality of WeChat and QQ nodes is obviously greater than that of other nodes.There are certain differences in the relationship between different behaviors under different time and space conditions.(2)An APP prediction model which takes into account online behaviors was proposed.First according to the currently used APP,the behavior of the user is obtained.The transition relationship between the behaviors in different time and space conditions is used to predict the behavior of the next moment.And,the behavior of the next moment is used to predict APP of the next moment.Finally,combined with the relationship between the APPs,gets the final forecast result.The top1,top5 accuracy of this model reached 24.66%and 60.72%,using the mobile phone data to conduct prediction experiments and analysis,better than the comparison model.(3)This paper proposed a method for predicting mobile surfing time.There is obvious periodicity in surfing time periods of 37.66%mobile phone users in experimental area by Fourier transformation and periodic test,which could help us understand the surfing characteristics of users.And the day is divided into low frequency,high frequency and transition period by the hierarchical clustering.This paper proposes a mixed Markov method(Lift-Markov method.LM),combining the traditional Markov model and association rule model,to predict the surfing time period of mobile phone users.The accuracy of LM method in the 10-minute and 60-minute interval scale can reach 92.17%and 79.75%,using the mobile phone data to conduct prediction experiments and analysis,better than the comparison model.