Research And Implementation Of Recommendation Algorithm Based On Time And Space Factors

With the development of technology,news information can be accessed more conveniently and quickly.However,in this era of information explosion,how to efficiently obtain news of their own interest has become a research hotspot.The timeliness of news is very important,which prompts the media to spread the news in a short time and push it to the appropriate user community.Therefore,the recommendation algorithm for the news category came into being.The current recommendation algorithm is very comprehensive,but there are advantages and disadvantages in the way the cold start problem is handled.And most of the recommended algorithms at this stage still have shortcomings in considering the changes in user interest.Therefore,this paper proposes two improved recommendation algorithms based on the spatiotemporal factors,and implements the news recommendation system by using Android.The main research work can be listed as follows:(1)In view of the fact that the current recommendation algorithm fails to deal with the cold start problem of new users and the sparseness of user sign-in data,this paper proposes a POI recommendation algorithm that combines expert trust.First,the user sign-in data in a certain time and space range is selected according to the user sign-in information to obtain a set of candidate expert users.Then,combined with the influence of the number of sign-in and the scope of the sign-in of the candidate expert user,the candidate expert user set is filtered to obtain an expert user,and the Top-N recommendation list is obtained according to the interest point of the expert user and recommended to the new user.For users with sparse data,the Top-N recommendation list can be optimized by combining the geographic location information of the remaining sign-in data of the user and calculating the geographical distribution impact by the kernel function to obtain a final recommendation list.Experiments show that the algorithm can effectively alleviate the cold start problem,and the accuracy rate is also significantly improved.(2)The existing tag-based personalized recommendation algorithm does not consider the time factor,especially the short-term interest of short-term interest on the recommendation result when constructing the user interest model.This paper proposes a personalized recommendation that considers the user's recent interest changes algorithm.First,based on the data of the user and the item,the label weights of the user and the item are respectively calculated.Then,combined with the forgetting curve and recent user interest changes,the user tag weights are updated,then the similarity is calculated,then the k-nearest neighbors are searched,the sparse matrix is filled,and the top-N recommendation list is output.In comparison,the algorithm has a certain improvement effect which increases the interpretability while improving the accuracy of the recommendation results.(3)This paper designs and implements the news recommendation system through the fusion of the two improved algorithms and Android.The system uses the user's geographical location information to recommend new users,which alleviates the cold start problem;it uses the user's collection,comment,sharing and deletion of news to analyze the user's interest changes to achieve more accurate news recommendation.The system not only improves the user experience when browsing the news,but also has high economic value after application.Through the above research,our paper realizes the optimization of recommendation algorithm by using space-time factors,alleviates the cold start problem,improves the recommendation effect,and applies it to the news recommendation system,which has contributed to the future news recommendation category.