Research On Collaborative Filtering Recommendation Algorithms Based On Similarity Fusion

In the age of information explosion,there is a huge amount of information,and there are also a lot of complicated content,which makes users sometimes unable to find interesting information in a short time.Therefore,personalized recommendation systems are widely used in various fields,such as social news,e-commerce,video media,etc.The Recommendation algorithm,which is an indispensable part of the recommendation system.And there are a wide variety of recommended algorithms.Slope One algorithm is a item-based collaborative filtering algorithm.It is classic and practical,with simple and efficient calculation process.It has been widely used with high prediction accuracy in the case of a small amount of data.However,the similarity between users is not measured,and the similarity between projects is not taken into account,which has an impact on the prediction accuracy and execution efficiency,resulting in no obvious advantages of the algorithm.Around the above problems,this paper obtains an improved weighted Slope One algorithm by improving the calculation method of weights.The main research work is as follows:(1)This paper introduces the basic theory of collaborative filtering algorithm,elaborates the research status,achievements and existing problems at home and abroad,elaborates the prediction method and calculation steps of Slope One algorithm,and proposes corresponding improvement strategies for existing defects.(2)This paper carries out an in-depth study on the problem that Slope One algorithm fails to consider the similarity.First,the trust Weighted method and Jaccard coefficient Weighted method are adopted to measure the similarity among users,and the Weighted Slope One algorithm that integrates user similarity is proposed.Second,Pearson correlation coefficient method is adopted to calculate the similarity among items,and an algorithm that integrates item similarity is presented as Weighted Slope One.Finally,two hybrid weighted recommendation algorithms are obtained by fusing the two improved methods.(3)In this paper,the experimental verification of the improved algorithm is based on the Movielens dataset and the comparison based on the Epinions dataset.It can be found that the improved algorithm is better than the original Slope One algorithm,which effectively benefit the improvement of recommendation Quality.