Research On Hybrid Algorithm Of Slope One Based On Predicating And Filling Missing-Data By Iterated

With the rapid expansion of the Internet in recent years,the volume of data has grown exponentially and recommendation systems have emerged.The corresponding recommendation system appeared in each different field,which greatly alleviated the difficulty of filtering useful data from a large amount of data by merchants and users.Slope One is popular because of its easy to understand principle,clear flow and high accuracy,which is different from other algorithms,which uses the average score difference between users to predict items without scores.Nowadays,there are numerous optimization papers on Slope One and the optimization directions are various.For example,the popular algorithm of introducing item or user similarity weighted Slope One has a higher accuracy than the weighted Slope One because of the introduction of similarity as a weight.However,the algorithm still has the problems of data sparsity,limitation of single algorithm and unpredictability in some cases.Therefore,based on the above defects,this paper proposes two improvements to the weighted Slope One of project or user similarity.The main contents are as follows:First,The similarity weighted Slope One algorithm needs to calculate the similarity due to the introduction of similarity,so it has the same problem of data sparsity as Collaborative Filtering.Deal with data data sparseness,a Slope One Based on Predicating and Filling Missing-Data by iterated is proposed.This algorithm fills users-score matrix by iterated,and for each iteration,scoring matrix calculated similarity matrix is used to predict the next filling.after several iterations improve the accuracy of the similarity matrix,thus improving the accuracy of thealgorithm.Second,Aiming at the shortcomings of the first improved algorithm,The algorithm is integrated with the collaborative filtering algorithm to solve the problem that the original two items of Slope One have no common score and cannot be predicted,and the limitation of a single algorithm is improved.At last,in order to verify the effectiveness of Slope One hybrid algorithm based on iterative prediction of missing value filling by using MovieLens.Finally,the proposed algorithm in this paper is more effective.