The Collaborative Filtering Recommender Algorithm With User's Similarity And Trust

With the rapid development of network information technology and e-commerce,network data expand constantly.User is hard to find the information which he needs quickly and efficiently,it raises the problem of information overloading.The mass of electronic commerce product made it difficult for a user to quickly lock and found himself really interested in what kind of goods,for companies selling on the site,of course,also want to be able to accurately recommend their products to increase sales and increase revenue.E-commerce recommendation system is to solve this problem.Collaborative filtering recommendation is widely used in recommender systems,it is well known,and the effect is satisfactory,however,the traditional collaborative filtering recommendation algorithm exists data sparse,cold start and scalability and other issues.In this dissertation,the problem of data sparsity in collaborative filtering algorithm is discussed,not only consider the user similarity in the traditional collaborative filtering recommendation,but also consider the factors of trust.A recommendation algorithm is proposed which considers both the similarity and the user trust,finally to reduce the sparsity,and further improve the accuracy of recommendation information.Research work in this paper as follows:(1)This dissertation introduces the composition and core of the recommendation system,introduces the recommendation algorithm,and the steps and problems of collaborative filtering recommendation algorithm.Consider trust to further illustrate the knowledge of trust.(2)Because of the low recommendation accuracy caused by the sparsity of the score data,this dissertation proposes a collaborative filtering recommendation algorithm based on user similarity and trust.On the basis of user similarity,trust relationship is introduced,which includes direct trust and indirect trust.Indirect trust is the transmission of direct trust.After the similarity and the degree of trust,the correlation degree is formed,and the whole is used as the recommendation weight.Select the nearest neighbor collection of the recommended user to recommend.(3)The design of the algorithm programming test to achieve the desired results.Based on the results of the experimental data analysis,we get the way of the similarity and trust of this paper,which is better than the traditional recommendation in the performance of MAE.In this dissertation,the algorithm achieves the expected purpose,can provide users with more accurate recommendation.(4)A movie recommendation system is designed to test the effect of the algorithm.The actual effect achieved the expect.