Diversifying personalized mobile multimedia application recommendations through the Latent Dirichlet Allocation and clustering optimization

The rapid development of mobile multimedia applications explosively increased the availability of the number of applications in the apps market. Among the crowd of mobile multimedia applications with diverse functions, identifying the appropriate application is a significant challenge. Hence, it is essential for the app stores to recommend the desired applications to the users in the surge of applications. Several research works lack to consider the interactions among the contextual information of applications such as application category and features in different aspects instead of user preferences. Thus, it is of significance to developing a practical approach that provides high-quality application recommendations for users according to personal preferences. This paper presents the DIversifying Personalized Mobile Multimedia Application Recommendation (DIPMMAR) by fusing the user ratings, review texts, application description, and application popularity. Initially, the DIPMMAR approach analyzes the user reviews and application descriptions by applying the Latent Dirichlet Allocation (LDA) based topic model. It employs the Principle Component Analysis (PCA) principal components are the continuous solutions to the discrete cluster membership indicators for K-means clustering among all the extracted features of the applications and retains the optimal latent features alone. Further, the DIPMMAR approach computes the user-specific local popularity score on applications and exploits the application-specific global popularity score to generate the top-N personalized recommendation. Moreover, by exploring the mobile application categories and sub-categories, the DIPMMAR approach ensures the relevance and diversified applications in the recommendation list. The experiments on the real-world mobile app store dataset demonstrate the accuracy of the personalized recommendation.