Hits:

Indexed by:会议论文

Date of Publication:2015-06-10

Included Journals:EI、CPCI-S、SCIE、Scopus

Volume:9101

Page Number:415-424

Key Words:Mobile applications; Recommendation system; Clustering

Abstract:Due to the huge and still rapidly growing number of mobile applications, it becomes necessary to provide users an application recommending service. In this work we present a recommendation system that recommends new applications for users according to their outdated applications. The recommender assumes that each owned application has complex information containing both descriptions and API information. The proposed approach mines application descriptions from publicly available online specifications and identifies APIs from the downloaded APK(Android PacKage) files. Text mining and incremental diffusive clustering(IDC) algorithm are utilized to generate common features. And APIs are extracted by disassembly technology. Then the complex information of applications can be represented by the features and APIs. In the processing of recommending, the k-Nearest-Neighbor algorithm based on the self-adaptive similarity(SS-KNN) is adopted to generate candidate sets of applications, and then the coverage-weighted similarity is utilized to select the final recommendations from the candidates. Extensive experiments are conducted on different application categories and the experimental results illustrate the effectiveness and efficiency of the approach.