| With the rapid development of information technology,human beings have entered the network information society.People enjoy the convenience of information at the same time,have to face a lot of information caused by serious information overload problem.In order to solve the problem better,the recommendation system came into being.It collects the user’s historical behavior,dynamically perceives the user’s interest or demand,and realizes the individual resource recommendation,thus effectively solves the information overload problem.Collaborative filtering algorithm is one of the most successful recommendation algorithms in recommender systems,which is based on the user’s group intelligence,which has been widely used in various fields because of its good recommendation effect.However,with the continuous change of application scenarios,and a rapid increase in the number of users and items of the existing collaborative filtering algorithms are also faced with many problems,based on the collaborative filtering algorithm as the main research object,mainly focus on the face of the user interest drift,sparsity and scalability issues,and research the combination of theory and practice,obtained the following research results:Aiming at the problem of user interest drift,the time-weighted collaborative filtering algorithm is studied,and the existing methods are pointed out.This paper proposes a new time-weighted function by using the temperature change function in Newton’s law of cooling to represent the general process of user’s interest change,and combines the information quantity,item attribute and user rating.On this basis,The results show that the proposed method can effectively alleviate the influence of user interest drift on the recommendation result under the premise of guaranteeing the recommended quality.The results show that the proposed method can improve the accuracy of the proposed method.Aiming at sparseness and expansibility,the clustering idea in machine learning is introduced into collaborative filtering algorithm.Firstly,we use the time-weighted function proposed above to preprocess the user’s score,and then use the improved K-means algorithm to complete the clustering of the items.Secondly,taking into account the high sparseness of the user-object scoring matrix,And then use the attribute of the item to cluster the items.Thirdly,for the problem that the accuracy of the single clustering model is poor,the results of the above two kinds of clustering are combined,and a collaborative filtering algorithm based on fusion clustering is proposed This algorithm preserves the advantages of clustering method for off-line modeling and expansibility.Compared with the single clustering model,this method can improve the recommendation accuracy on the basis of alleviating the problem of sparseness and expansibility degree.A film recommendation system based on B/S architecture is designed and implemented for the problem of movie information load.The system mainly includes off-line computing module,online recommendation module,information feedback module,information management module.The system comprehensively implements the improved algorithm proposed in this paper,which can provide high quality film recommendation for users. |
