The Improvement And Research On Collaborative Filtering Algorithm In E-commerce Recommendation Systems

In recent years,with the rapid development of Internet,the electronic commercekeeps a sustained increasing trend, the information are in expansion even overload,so the electronic commerce is facing an enormous challenge. In order to solve thisproblem, the electronic commerce recommendation system emerge as the timesrequire, the electronic commerce recommendation system can help users quickly topositioning commodities which they love. The collaborative filtering technology isthe most rapid developed technology, but the traditional collaborative filteringalgorithms have its own limitations, it exists "sparsity" problem,"cold start" and"scalability" problem, in view of the lack of collaborative filtering algorithmcommonly used in electronic commerce recommendation system, this dissertationputs forward an kind of improved combination of collaborative filteringrecommendation algorithm.This dissertation introduces the clustering analysis firstly, in the clusteranalysis model, just searching in the target search space which by clustering, ratherthan in the whole data space, thus reducing the searching range, improves theefficiency of search, effectively improves the real-time response speed of the system,this is the reason we introduce clustering analysis.This dissertation introduces the trust model, making the similarity matrix whichis produced by user-item rating matrix and the trust matrix which is produced byuser-user trust score matrix together using the regression analysis method, as thenew standards which search the nearest neighbor, improve the recommendationaccuracy, so as to provide better recommendation results for target users.This dissertation is based on the traditional collaborative filteringrecommendation algorithm, introducing the clustering analysis and the trust model tobecomes a new combination of collaborative filtering recommendation algorithm, the recommendation algorithm can improve or overcome the problem of thetraditional filtering technology, so as to provide the accuracy of the system. thisdissertation designs the algorithm and do the contrast experiment, the experimentshows, the algorithm can reduce the average absolute deviation between the userscore and actual prediction of user ratings, improve the quality of recommendationsystem.