Research On Recommendation Algorithm Based On Sparse Matrix

In this era when Internet technology is leading the trend,the amount of data and information in cyberspace is too large to be counted.People urgently need to "retrieve" the required content from the vast amount of information data.The recommendation system has become a sword of "information retrieval" by linking items with information.In practical applications,e-commerce is taken as an example.On the one hand,high-quality recommendation system can help e-commerce platform to increase sales volume,on the other hand,it can also improve customer satisfaction with e-commerce platform.Based on the research of collaborative filtering recommendation algorithm,this paper proposes improvement measures for the deficiencies in this algorithm.Firstly,according to the principle that similar users and similar project scores are also similar,a similar data filling method is proposed to solve the sparsity of data.On this basis,in order to alleviate the problem of poor scalability of personalized recommendation system,the k-means clustering method is used to find the clustering clusters of target users so as to narrow the scope for target users to find the nearest neighbor users,which can reduce the computational time complexity and improve the online stability of the recommendation system.Finally,a hybrid similarity calculation method considering user's item type preference and user's own attributes is proposed for the similarity calculation method commonly used in traditional personalized recommendation systems,which can only calculate the similarity of user scores.Through the improvement of the above method,the experimental analysis is carried out on the Moivelen-Latest-Samll data set,and the optimal K value and the nearest neighbor number are 21 and 8.Then the improved recommendation algorithm is compared with the traditional recommendation algorithm by using MSE as the evaluation index,and the MSE average values are 0.54 and 0.57 respectively,thusproving that the recommendation model established by the method can effectively improve the quality of the recommendation system.