A Point-of-Interest Recommendation Model Based On Effective Path Coverage

The rapid development of the internet has brought a huge amount of user data,how to mining valuable information accurately and efficiently from these massive data to improve the users' service quality has become one of the key point of today's research.in different platforms,because the recommended project and the recommended user groups are different,the corresponding personalized recommendation method is also different.In local based social networks,point-of-interest(POI)has location feature,when recommending POIs for users,it not only to consider users' interest,but also to consider the spatial factors,which lead to the traditional method can't be directly applied.In the latest research on POIs recommendation,it not only adds the spatial factors,but also further subdivided the time factors.Then to provide user with POIs in different time periods,the optimization method also enhances the diversity of recommendation results,it avoids the POIs tends to homogeneity,and provides user with POIs that are both interesting and diversified.However,in the above POI recommend method and it's improvement algorithm,firstly,when the user visits POIs,the position of the user is moved,while the recommended POIs are all generated for the current position of the user,after the user visit to a POI,previously recommended POIs may have lost the recommended value because of the changes of the user's location now.In addition,users usually do not continuously access a set of unrelated POIs,it only considers the diversity of POIs in the improved recommendation algorithm,but the association between these different categories of POI is not taken into account,the recommended POIs may have not association among them,it is easy to lead to some of the invalid recommendation in result,user can only pick some of the recommended POIs from result to visit,which greatly reduces the efficiency of the recommendation.For the above two problems,this paper proposes a point-of-interest recommendation model based on efficient path coverage.The model firstly classifies the POIs,and then adds the weight to the POIs in the clusters according to the distances between the clusters and user,the larger the distance,the larger the weights of POIs.The purpose of adding weight is to make the recommended POIs clustered in the nearest cluster,even when the location of user has changed after accessing to some of POIs,the rest of the recommended POIs are still close to the user,it reduce the risk of other recommended POIs failure.Secondly,the model analyzes the historical access data of all users,andmining the association between different categories of POIs,we can compose the effective path from the related POIs and then recommend to users,this will reduce the invalid recommended radio.we propose the point-of-interest recommendation model based on effective path coverage,we developed a dynamic programming algorithm and further optimization to approximately solve it.The experimental is based on two real-world data sets shows that the proposed algorithm outperforming two other state-of-the-art methods in terms of diversity and precision.