Research On Mean Value Of Collaborative Recommendation Algorithm Based On Clustering And Project’s Properties

In the last twenty or thirty years, with the rapid development of Internet technology and computer technology to promote the rapid progress of the society, it is not only makes the society more diverse and colorful, but also makes the communication and access to information faster and more convenient. With the accumulation of time,the information saved on the internet increased into a geometric, information overloaded leads us to find the useful and interesting content is anxious and helpless.Collaborative filtering recommendation technology has been widely used in the field of personalized recommendation because of its simple and efficient advantages.To solve the problems of collaborative filtering recommendation, this paper propose a collaborative filtering model based on clustering and project mean.By using the improved clustering algorithm, the whole scoring matrix is clustered to reduce the time of calculating the similarity.The model calculates the feature similarity of items by the time attribute and genre attribute to solve the problem of the cold start of the collaborative filtering recommendation.This paper designs the experiments and verifies the proposed model on the universal data set movielens.First, design experiments to find the best time weight factor α and type weighting factor β;Second, through five sets of training and test sets to experiment to find the best neighbor number K of collaborative filtering recommendation, laying the groundwork for subsequent comparison experiment;Third, experiment to find the best model combination factor γ;Finally, through the design of the contrast experiment,validated the model of this paper is better than the traditional collaborative filtering recommendation model based on item, and it solved the cold start, scalability and other issues of the collaborative filtering.