| The demand of consumer loan products is increasing day by day in our country.Among the many consumer loan products,it is a great challenge to recommend satisfactory products for users.At present,collaborative filtering is a widely used recommendation algorithm among all kinds of mainstream recommendation algorithms.However,traditional collaborative filtering has the problem of user data sparseness,which leads to the reduction of the precision of user similarity calculation.In response to the above problems,this thesis proposes consumptive loan products recommendation model based on the Density Peaks Clustering Algorithm(DPCA)based on the mobile operator network data supported by the Ministry of Education-China Mobile Research Fund Project.The main research contents and contributions are as follows:(1)A consumptive loan product recommendation model based on DPCA clustering is proposed and implemented.In the process of user classification of collaborative filtering recommendation model,DPCA clustering is introduced to extract consumer behavior characteristics from user mobile data,which includes two types of features: shopping tendency and purchasing power of users,based on the above features.Through the DPCA clustering method,the users with similar demand for consumer loan products are clustered into the same class,which simplifies the process of finding the nearest neighbor and improves the precision of the similarity calculation of the users,and the users with similar demand for consumer loan products are clustered into the same class,so as to improve the accuracy of user similarity calculation.Thus,the accuracy of the recommended results is improved.(2)Design and implement recommendation model of consumptive loan products based on improved DPCA clustering.The traditional DPCA clustering can not deal with mixed attribute data sets,it is difficult to distinguish the center of cluster from the noise point,the uncertainty of truncation distance selection and the precision of single cluster are low and so on.In order to improve the accuracy of user classification,DPCA clustering is improved to solve the above problems.The distance measurement method between mixed attribute objects based on information entropy and the cluster center recognition method based on twice residual analysis are proposed.GA-based optimal truncation distance acquisition method and a clustering integration method based on minimum spanning tree.Thus,the accuracy of the recommendation model is further improved.Experiments were carried out on the data sets and public data sets provided by the partner companies.The evaluation indicators were the mean absolute deviation(MAE)value,the ARI index and the FMI index.The experimental results show the validity of the model..In summary,aiming at the shortcomings of collaborative filtering algorithm and the shortage of DPCA clustering algorithm,this paper proposes a solution respectively,and proves that the proposed scheme can effectively improve the accuracy of recommendation results,and has certain reference value in the research of recommendation field. |
