Research On Collaborative Filtering Recommendation Algorithm Based On Information Entropy

With the rapid development of Internet technology and the advent of big data and cloud computing era,the emergence of a large amount of resources makes the phenomenon of "information load" more and more aggravating,which makes it difficult for consumers to screen out valuable resources and at the same time,a large amount of resources in the network become long tail data because of low utilization.Therefore,how to help users find the information they need quickly and efficiently from vast amounts of data and improve the user experience has also been a hot spot in the field of Internet research in recent years.And the birth of the recommendation system exactly helps users improve the problem.In order to adapt to the resource recommendation in different scenarios,the recommendation technologies are also under constant improvement and updating,among which the collaborative filtering recommendation is the most widely used and most successful one by far.But there are still some limitations collaborative filtering also has some shortcomings:data sparsity,cold start and so on.Based on the traditional recommendation of collaborative filtering,this paper analyzes and studies the existing problems deeply.The main research work of this paper is as follows:(1)The low recommendation accuracy in traditional collaborative filtering recommendation is mainly caused by severe sparsity of data.In the traditional user similarity measure,only the number of common scoring among users is considered,and the difference between specific scores is ignored,resulting in unsatisfactory recommendation effect.To solve this problem,this paper introduces the information entropy in information theory and considers the influence of scoring differences between users.A similarity calculation method based on information entropy is proposed by calculating the entropy value of user difference.(2)In order to improve the accuracy of the nearest neighbor set,this paper comprehensively considers the influence of the user's common scoring item number and score values,and linearly weights the similarity measurement method based on information entropy and the Pearson similarity calculation method in the traditional collaborative filtering algorithm,and a new method called weighted similarity is proposed.(3)Aiming at the problem of ignoring the user's interest in traditional similarity calculations and based on the weighted similarity,considering the change trend of the user's score,the user's interest degree is introduced,and a similarity calculation method for the fusion of information entropy and interest is proposed to improve the recommendation quality.Finally,in order to verify whether the proposed similarity calculation method is effective,the algorithm is applied to the MovieLens dataset to observe the experimental results and multiple groups of contrast experiments with the traditional collaborative filtering algorithm are set up.The results show that the recommendation quality is better than the traditional collaborative filtering recommendation by considering the common scores,the score values and the trend of the scores,which proves the effectiveness and feasibility of the methods.