Combining Collaborative Filtering and Topic Modeling for More Accurate Android Mobile App Library Recommendation

The applying of third party libraries is an integral part of many mobile applications. With the rapid development of mobile technologies, there are many free third party libraries for developers to download and use. However, there are a large number of third party libraries which always iterate rapidly, it is hard for developers to find available libraries within them. Several previous studies have proposed approaches to recommend third party libraries, which works in the scenario where a developer knows some required libraries, and needs to find other relevant libraries with limited knowledge. In the paper, to further improve the performance of app library recommendation, we propose an approach which combines collaborative filtering and topic modeling techniques. In the collaborative filtering component, given a new app, our approach recommends libraries by using its similar apps. In the topic modelling component, our approach first extracts the topics from the textual description of mobile apps, and given a new app, our approach recommends libraries based on the libraries used by the apps which has similar topic distributions. We perform experiments on a set of 1,013 apps, and the results show that our approach improves the state-of-the-art by a substantial margin.