Research On Algorithms Of Recommendation In Personalized Recommendation Systems

Nowadays the scale of e-commerce sites is increasing,therefore,leading to the kinds of goods online becoming more various.The emergence of recommendation systems can help users make choices in face of so much information.However,the current recommendation algorithm is still in development and its accuracy can not meet the needs of users.As a result,there’s a large amount of research on recommendation algorithms.In order to provide users with more convenient and more accurate recommendation service according to the user’s particular preferences,more and more experts and scholars devote their time to the research on recommendation systems.Association rules and collaborative filtering technology are two important recommendation algorithms.However,the accuracy of these algorithms needs to be improved because of the heterogeneity of items distribution and the sparsity of data.The limitation of association rules and collaborative filtering algorithm is improved in this paper and then a new algorithm is proposed,which is applied in the personalized web mall recommendation systems.The work of the dissertation is partly supported by the National Natural Science Foundation of China(No.61172072,61271308),Beijing Natural Science Foundation(No.4112045),and Research Fund for the Doctoral Program of Higher Education of China(No.20100009110002).The main research contents of the dissertation are as follow:(1)The association rule algorithm is modified according to the defects of Apriori algorithm and Partition algorithm.Aiming at the items with strong regularity in the database,a weighted formula is designed to give items the corresponding weight in order to increase their support degree.(2)In order to solve the problem of data sparsity in collaborative filtering algorithm,the implicit information is obtained through the records in the database and MapReduce technology is used to optimize the sparse matrix in the personalized web mall recommendation systems.(3)A multiple K-means collaborative filtering algorithm based on Binary Cycle Index is proposed in this paper to find the nearest neighbor sets for the target user for whom the recommendation service is provided.The proposed algorithm is then applied in the web mall recommendation systems to optimize the process and models of recommendation.(4)A recommendation system prototype combining the improved weighted association rules with the optimized collaborative filtering algorithm is designed in this paper.And the performance of the recommendation system prototype is then analyzed through experiments.The results of experiments show that the accuracy of the modified association rule is improved;Using MapReduce technology to get users’ implicit information can make up for the defects of that the web mall only use users’ explicit information;The improved collaborative filtering algorithm can improve the efficiency and accuracy of discovering nearest neighbor sets;The performance of the recommendation system prototype integrating the two algorithms is good.