| With the increasing scale of e-commerce websites,the problem of information overload is becoming more and more serious.One of the most potential ways to solve this problem is the personalized recommendation system.However,the proliferation of users and commodities has caused many problems such as data sparsity,cold start,diversity and precision dilemma.Cross-domain recommendation has been a hot topic,which aims to utilize knowledge from related domains containing rich data to perform or improve recommendation in the target domain.In this paper,considering the cross-domain recommendation technology,we propose a new way to solve the problems existing in the e-commerce recommendation system by applying the transfer learning to the recommendation of the e-commerce.The main work of this paper is as follow:(1)In view of the inaccurate calculation of user similarity calculation in the case of extreme sparse data,this paper proposes a cross e-commerce solution based on user similarity migration.In the field of user overlapping scenarios,this method transfers the shared user similarity model calculated by the relatively rich source domain to the target domain.The experimental results show that this method can effectively transfer the user similarity and obtain better recommendation performance.(2)In view of the actual scene of e-commerce sites existing data protection and difficult to obtain overlapped users,this paper discusses how to obtain and transfer the source domain knowledge effectively,and proposes a cross e-commerce recommendation method based on clustering feature transfer.In this method,the user group's rating patterns of the commodity group is transferred to the sparse target domain so that it can help target domain reconstruct the user-item matrix and then produce the recommended list.The experimental results show that this method has a wide applicability,improved the recommendation accuracy,and do not lose the diversity of its recommendations. |
PDF Full Text Request