Popularity Sensitive APP Topic Recommendation Model

Everyday lots of apps are released for various purposes, we seems to have entered a mobile Internet era, which apps installed on our smart mobile terminal as the main selling point. With a great variety of apps appear on the intelligent terminal application store, we choose the apps which meet the needs of our lives has become increasingly difficult, so app recommendation is also becoming increasingly important. However most existing recommender systems make recommendations for app users rather than app developers. We observed that when developers decided to develop an app, they will first make sure some app attributes and feature information(such as App Category(Action Game, Music, Social etc), App Content Rating(Adult, Children etc), App OS Version(Android 5, IOS 8 etc) etc), and then they start to develop an app, at last they need to write a description, which plays a very important role in the choice making by app users, especially for those who are unfamiliar with the app.Inspired by this, we decided to make recommendations for app developers, helping them write an intriguing description. We present a novel model that recommends the topics related to app popularity and their probabilities for developers, which we name Popularity-Sensitive APP Topic Recommendation(PSATR). We use Labeled LDA to construct a representation of an app as a set of latent topics from app metadata and textual descriptions. We then construct a training set by collecting app pairs from same category, and calculate the corresponding difference feature vector for each pair of apps. At last we compute the relevancy between app popularity and each topic to generate popularity-sensitive ranked list of topics and their probabilities for each category.What is novel about our method is that recommender system should not only consider regular user, but app developers. And our model recommends the attractive topics and their probabilities to developers instead of Apps, which can be used to improve the quality of app description.