In recent years,with the civil use of positioning technology and the popularity of smart devices,location-based social networks which combine location information with ordinary social networks have got greatly development.This paper aims at researching the Point-of-Interest recommendation algorithm in this kind of networks.Point-of-Interest recommendation could help people find interesting locations.Traditional recommendation methods only consider the effects Spatial clustering phenomenon may have on Point-of-Interest recommendation,or just focus the recommendation on a fixed period of time,without taking account of the attenuation phenomena of time and space.In order to further improve the performance of the recommendation algorithm,this paper mainly does the following research:Firstly,in terms of statistical significance,The shorter the time interval or spatial distance is,the higher similarity Point-of-Interest will have and thus the greater weight it will take in the recommendation algorithm.Therefore,on the basis of the best traditional collaborative filtering recommendation algorithm based on Point-of-Interest,it introduces time attenuation function or space attenuation function as weights to solve the similarity of Point-of-Interest and applies the improved algorithm to Point-of-Interest recommendation.The test results on real datasets show that the collaborative filtering Contextual-Aware recommendation algorithm based on Point-of-Interest which introduces time attenuation function or space attenuation function performs better than traditional recommendation algorithms.Secondly,for considering attenuation characteristics the similarity of Point-ofInterest may have both in time and space,this paper combines the two improved Point-of-Interest recommendation algorithms with the method of linear combination and designs the collaborative filtering Contextual-Aware blending recommendation algorithm based on Point-of-Interest which introduces both time attenuation function and space attenuation function.The test results on real datasets show that the blending Point-of-Interest recommendation algorithm has much better effects than introducing time attenuation function or space attenuation function separately. |