Mobile News Recommendation With Location Constrain

With the rapid development of communication technology and the increasingly of the functions of mobile smart terminals,Location Based Services and mobile news recommendations are gradually merged to form a location-based mobile news recommendation service.It associates the user's location with reading preferences that can describe the user's reading interest more accurately.Since the data sparsity and “cold start” of new locations in location-based mobile news recommendation systems,it often leads to poor recommendation effect.In response to these problems,this dissertation makes the following improvements:1.Introduce location-context information: user's reading interest is closely related to the location.Introducing the location context can describe the user's reading interest in detail,thereby providing more accurate recommendations and improving the user's reading experience.For this purpose,extracting the position information when user reading,and establishing a reading interest model of the corresponding position by cluster analysis.2.Propose a three-dimensional similarity algorithm: To solve the problem of "cold start" and "sparse",proposing a similarity algorithm,namely three-dimensional similarity algorithm.The algorithm can make full use of the user's information,improve the accuracy of the similar user's calculation and solve the problem of sparseness of the scoring data.When user in a new position,When the user is in a new location,the three-dimensional similarity algorithm can be reduced to a two-dimensional similarity algorithm,which is trajectory-attribute,to solve the problem of "cold start" of a new location t.3.Establishing the DLR(Diversity news Location-Oriented Recommendation model)model.Establishing long-term reading interest model and diversity reading interest model for users based on content-based recommendation and collaborative filtering recommendation which based on three-dimensional similarity algorithm respectively.Then,integrating two models.The model meets the personalized needs of users while increasing the diversity of recommendation results.