Research On Collaborative Filtering Recommendation Algorithm Based On Spectral Clustering SM Algorithm

With the development and innovation of 5G technology,the amount of network resources data has also shown explosive growth.The development and supply of intelligent terminal products have created a good opportunity for the development of M-Electronic Commerce.The core of the recommendation system is the recommendation algorithm.The mobile recommendation system needs to produce timely recommendation results to consumers,and people have higher requirements for the effectiveness of recommendations.For the problems of collaborative filtering recommendation algorithm,people take corresponding measures to improve.This article focuses on the solution to sparsity and scalability problems in the recommendation algorithm.Some scholars proposed a recommendation algorithm based on user k-means clustering and preference.However,when k-means algorithm deals with non-convex functions,it is easy to fall into a local optimal solution,and the clusters formed by k-means algorithm are mostly spherical.Spectral clustering can cluster data and get the optimal solution in any shape dataset space.Therefore,the proposed collaborative filtering algorithm based on user spectral clustering is first based on the user-item scoring matrix R(m,n).The user clusters first to generate a number of user-center clusters;according to the results of the clustering,new similarity attributes are used to search for the user's nearest neighbor.The deficiencies of this paper is that there is no further research in the face of the cold-start problem in the collaborative filtering algorithm;in the experiment of verifying the algorithm,we assumed that the similarity threshold exists unitary and does not further proceed under different similarity thresholds.In terms of the number of clusters,although the concept of clustering profile coefficients has been introduced,if the number of other clusters is selected,this paper does not use scientific methods for verification.This article uses the Movielens data set for proof.Through verification of the MAE value,it is finally proved that the proposed collaborative filtering algorithm based on spectral clustering SM algorithm has better recommendation effect than the traditional collaborative filtering recommendation algorithm in mobile e-commerce system.