Research On Personalized Recommendation Method And Application Based On Professional Document Management System

With the acceleration of global information construction,the enterprise’s professional document management system will accumulate a large number of electronic documents in daily operation.The increasing number of documents in the system makes users unable to get the required documents quickly,which seriously affects the office efficiency of enterprises.In the face of the massive document resources in the system,the personalized document recommendation subsystem for different users can efficiently help users find the resources they are interested in.The emphasis of this paper is to further improve and optimize the recommendation algorithm in the personalized recommendation subsystem,and to prove the effectiveness of the improved algorithm by comparative analysis of experimental evaluation data.At the same time,the article also carries on the detailed demand analysis,the logical structure and the main database design,as well as the overall and each function module design research to the personalized document recommendation subsystem which exists in the professional document management system.For the original collaborative tag recommendation algorithm,the profile information of users is not taken into account,and it can not be personalized for specific users,and the accuracy is also poor.Based on the standard collaborative tag recommendation algorithm,the system further improves the accuracy of personalized document recommendation results for specific users by incorporating graph clustering method and trust information.This paper mainly improves the original algorithm in the following two aspects:(1)In the process of classifying a large number of users and document tags,by introducing the most dense subgraph clustering method,all tags are generated into several tag clusters,so that the tags in the cluster are similar to each other and different tag clusters are different from each other.The difference improves the processing efficiency of the original algorithm for large data sets and the mining ability of unknown recommendation objects.(2)In this paper,the user’s trust information for different tag clusters is introduced into the original algorithm.Based on the original algorithm,the user’s trust information and its interest in document resources are linearly combined to generate a personalized score for each document.Not only can the system make personalized document recommendations for specific users,but it also improves the accuracy of the recommendation results.Under two different data sets,by comparing and evaluating the experimental data of HTGCPR algorithm with KMCPR and TCFR algorithms,the results show that the application of personalized combined recommendation method meets the user’s requirements for the real-time and accuracy of document recommendation,and improves the user experience to a certain extent.