Top-N Recommendation Algorithm Based On Sparse Linear Method

The real-time updating of network information brings great convenience to people's lives.At the same time,the problem of information overload brought about by a large amount of network information also brings certain problems to people's lives and work.The emergence of the recommendation system alleviates this problem to some extent,alleviates the problem of information overload in today's society,and reserves a lot of time for people's lives and work.In the research and development process of recommendation systems in recent years,the Top-N recommendation problem has gradually become a key part of the research in the recommendation system,and a variety of efficient recommendation algorithms have been continuously proposed.Currently,in the Top-N recommendation algorithm,the sparse linear method has obvious advantages compared with other methods.However,it has been found through studies that sparse linear methods may not well capture the differences in preferences among different users,resulting in recommendation results.Accuracy tends to an average.For the problem that the recommendation accuracy in Top-N recommendation is not high,this paper chooses the sparse linear method as the theoretical basis,and proposes a personalized combination method based on sparse linear method.First of all,for the problem of different preferences among different users,the idea of using a multivariate local model is used to divide all users into multiple user subsets and use a sparse linear method to construct multiple local models for multiple user subsets to capture different users.The difference in preferences.Then,the global model and the multivariate local model trained by the traditional sparse linear method are combined in a personalized way.The multivariate local model is responsible for ensuring that the combined model can effectively capture the differences in preferences among different users.The global model can guarantee the combination.The accuracy of the model.The individualized weight controls the proportion of the multi-local model and the global model in the combined model,and updating the personalized weight can effectively find a suitable user subset for the user.To a certain extent,the problem that Top-N recommendation cannot capture the difference in preferences among differentusers and the accuracy of recommendation results is not high.By selecting multiple types of experimental data and designing experiments,the advantages of the proposed combined method compared with the single method are first validated,and then the methods proposed in this paper are compared with the current methods popular in Top-N recommendations.Experimental results show that the proposed method is optimized in both HR and ARHR.