Content Based Mobile Application Partitioning And Recommendation Method

With the rapid growth of internet,more and more people have their own smart phones.Mobile application is developing rapidly.According to statistics,the number of mobile applications in Google play and Apple APP store has exceeded 1500000,mini programs has exceeded 1000000 also.With so many mobile applicatios,it's hard for users to find suitable mobile application.Therefore,users need an effective mobile application recommendation system to help users make decisions when them need to find a new mobile application.Considering that the traditional mobile application recommendation methods used in mobile application stores generally provide recommendation methods such as rankings lists,classification lists,editing recommendations lists and so on.These methods can only recommend popular mobile applications or require manual selection.To some extent,we can not get better recommendation results.Therefore,this paper proposes a recommendation method based on mobile application partition.The method is divided into two parts:(1)We proposed a mobile application division method which based on the content of mobile application.Firstly,we use different text modeling algorithms to model the mobile application description text and obtain their document vectors which contain the useful information about their function.Then we use different classifying or clustering algorithms to divide mobile applications apart based on the documents vector.Thus,mobile applications are divided according to their functions,which can effectively reduce the time of searching mobile applications.(2)We proposed a mobile application recommendation method which based on the mobile application partition.Firstly,we model the user's query and the description text of mobile application at the same time,and find the most similar mobile application cluster based on the partition of mobile application.All the attributes of mobile applications in the cluster,such as classification,downloads,price,score are grouped together,and then the deep factorization machine algorithm is used to predict the recommended score of mobile applications.Finally,the top N mobile applications are recommended to users as recommendation results.