Experimental Analysis of the Effects of Social Relations on Mobile Application Recommendation

In this paper, we empirically analyze the effects of social relations on the recommendation of mobile applications in a community of students at a university. We identify three social relations by questionnaires and two relations by students properties, and examine their effects from a wide variety of perspectives in the framework of top-N recommendation by user and item based collaborative filtering with two re-ranking mechanisms. In the analysis, we assess the difference of the effects by the origin and strength of social relations as well as by the methods of collaborative filtering and re-ranking mechanisms. As a result of the analysis, we confirm that appropriate social relations can significantly improve the performance of recommendation, in terms of increasing diversity and novelty with keeping high accuracy, especially for the late adopters.