Research On Personalized Recommendation System Based On Collaborative Filtering

With the rapid development of computer networks,the era of big data has arrived.E-commerce and other fields are full of massive data.It has become increasingly difficult for users to select their favorite items from a large number of items.Personalized recommendation system can quickly recommend people's favorite items.Collaborative filtering recommendation algorithm is a hotspot of personalized recommendation algorithm,which has been deeply studied and discussed by scholars at home and abroad.The user-based collaborative filtering recommendation algorithm calculates the similarity to select neighbors,and performs the neighborhood recommendation.Aiming at the problems that the traditional user-based collaborative filtering recommendation algorithm ignores the user's interest partitioning and the recommendation recall rate is low,and ratings predicted by the traditional scoring method has large error,the collaborative filtering recommendation algorithm based on interest partitioning and support vector regression is proposed in this paper(IPSVR UCF).Firstly,IPSVR_UCF algorithm clusters items using K-Means algorithm.Different item sets represent specific user interest point,and then the user interest is partitioned.On the one hand,constructs the feature vector of the item according to the partitioned interest points,and the IPSVR_UCF algorithm calculates user's preference for tbe item by using the ratings of item.Then,according to the items'features and the user's preference for the items,IPSVR UCF algorithm calculates the user s interest vector.On the other hand,by combining the items' features and their ratings,IPSVR UCF algorithm trains the support vector regression model and will use the model for predicting ratings.Finally,the concept of user maturity is proposed.Items can be recommended by combing collaborative filtering which is based on interest similarity and ratings prediction method based on support vector regression.The traditional collaborative filtering recommendation algorithm exists the problem of low recommendation accuracy caused by ignoring trust between users.A improved collaborative filtering recommendation algorithm based on user trust is proposed in this paper(Trust UCF).Firstly,the Trust UCF algorithm mines user interest features by the Latent Factor Model,and then calculates the user interest similarity.Then,the calculation method of familiarity degree is proposed.The user trust degree can be established by combing user familiarity with the neighbor and user interest similarity.Finally,the user similarity weighted by the trust degree as the final similarity,and the user-based collaborative filtering recommendation will be performed.Finally,the experiment is carried out by comparing with the traditional user-based collaborative filtering algorithm on the MovieLens dataset.It can be concluded that IPSVR_UCF can effectively improve the recall rate and reduce the ratings prediction error,and Trust_UCF can effectively improve the accuracy of the recommendation.Figure[26]table[11]reference[51]...