APP Usage Prediction And Application Based On User Behaviour Habits

In recent years,the popularity of smart phones has driven the explosive growth of mobile application.A large number of applications installed in the mobile phone not only increases the time and difficulty for user to find applications,but also the background application that is not closed in time occupies the memory resources of mobile phone.This problem can cause nonfluency and carton of mobile phone,which seriously affects the user experience.Based on the above problems,a mobile app optimization scheme based on user behavior habits is proposed.The specific research contents are as follows:1.The traditional prediction algorithms ignore the relationship between the app sequences,and does not take into account the user's preference for the app,so that the accuracy of the prediction algorithm is low.To solve this problem,an app prediction algorithm based on WPSPM(Weighted PrefixSpan Sequence Pattern Mining)algorithm is proposed.On one hand,in order to improve the pattern matching efficiency of prefixspan algorithm,the proposed algorithm integrates the AC automaton into the sequence search process,and proposes an improved prefixspan sequence pattern mining algorithm.On the other hand,the proposed algorithm add the user's preference for each app to the process of the prefixspan algorithm,and mines the pattern which is more line with user's behabiour habits.Then,the app sequence is used to predict the next app that the user will use.The simulation results show that the proposed algorithm can significantly improve the prediction accuracy based on the correlation algorithm.2.The traditional prediction algorithm only considers the next app that the user will use.However,if it is applied to the app smart cleaning applying scenario,it is very likely to clean up the background application that the user may launch in the future,which cause too many application restarts.To solve this problem,an app prediction algorithm based on the BNMC(Bayesian Network and Markov Chain)hybrid model is proposed.The algorithm uses markov algorithm to calculate the multi-step absolute transition probability matrix of each app,and then obtain the mean transition probability of switching to other app.Then,using the bayesian network model,the transition probability is integrate with app time period and usage position to predict the score that the application will be started soon.Finally,the proposed algorithm is verified on the android mobile terminal.The experimental results show that,compared with the traditional correlation algorithm,the proposed algorithm can intelligently clean the mobile phone background application,which effectively reduces the restart rate of the application.