APP Recommendation Based On Distributed SVD And Databased Social Influence

With the rapid development of mobile internet,the variety and quantity of mobile phone applications(APP)are increasing rapidly.As of March 2017,Android and Apple application market have about 5 million APPs.APP recommendation is playing an important role to enhance user experience and raise revenue.Existing recommendation strategies are mainly based on user's individual information while their social relations are often neglected.However,it is an intuitive knowledge that users tend to be affected by their friends' recommendation in the choice of APPs.Therefore,it is worth investigating how social influence can be employed for APP recommendation.In this paper,personalized recommendation algorithm and social influence are deeply studied,and we propose a novel APP recommendation method,which is called SVD-WSI for short.The core of this algorithm is recommending APPs with higher predicted usage to users,mainly includes two parts:the first part according to the user's personal information,which is distributed SVD algorithm;the second part according to social influence referring to D-SI(Databased Social Influence)algorithm.The experimental results are composed of three parts.First,in the personalized recommendation algorithm,based on MovieLens public data,we verify that the accuracy of the distributed SVD algorithm is higher than userCF,itemCF and SVD++algorithm.Second,results based on the real world datasets from Tencent APP Store demonstrate that our proposed method with social influence can achieve a better recommendation results.And last,based on the same datasets in second part,our proposed influence computation model outperform than benchmark WC in influence maximization problem—selecting the most influential users.