基于HITS算法的安卓应用程序分类技术研究

he rapid development of Android operating system and the large increase of Android applications have brought the problem of automatic classification to application markets. The existing automatic classification methods of android applications focus mainly on expanding the scope of features, which ignores the internal relationship between them and limits the improvement of accuracy. This paper proposes a solution to this problem, and analyze the call relationships to set different weights for APIs. This paper designs a new screening method for API features and make the most of the static feature information to build a machine learning model of application classification. In order to select the key APIs in each category, this paper proposes to rank the correlation between API and application\'s functionality by hyperlink induced topic search (HITS) algorithm. Through this application classification solution, this paper analyzed 12791 applications from Google Play Store, and finally achieved an 86.6% classification accuracy, which is about 7% higher than the performance of existing methods.