Reseach On App Recommendation Method Based On Implicit Feedback Data

With the rapid development of mobile Internet,users usually download Apps from mobile application markets.On the one hand,in order to increase its influence,mobile application markets face the situation of recommending App to users and keeping them in the markets.On the other hand,for a user,with millions of Apps from each application market,how to choose the Apps that they really like is an important problem.This paper proposes two App recommendation methods based on implicit feedback data,including meta-path-based App recommendation approach based on heterogeneous network analysis and an approach called RQE-EDKL(Recommendation Quality Evaluation based on Empirical Distribution and KL Divergence).The former approach constructs a user-App model based on heterogeneous network and then proposed the meta-path-based App recommendation approach to find App that users may be interested in.And QE-EDKL firstly makes use of historical user-item data to produce the historical usage probability distribution of items at different quantities.Secondly,we calculated the KL divergence based on the distributions of the historical usage probability and the usage probability of different recommendations.Thirdly,the recommendation with the minimum KL divergence is regarded as with the best quality and is recommended to the user.By using the App dataset from Talking Data,we compare the proposed meta-path-based algorithm with baseline algorithm.Experiment results demonstrate that meta-path-based algorithm has produced better performances than the baseline algorithms on MAP and MRR measures.RQE-EDKL can effectively improve the quality of recommended results of collaborative filtering significantly on both accuracy(MAP?MRR)and diversity(NOV?ILD).