Research On Library Personalized Recommendation Method Based On Collaborative Filtering

In such an era of “information explosion”,Internet provides people with large quantity of information resources,among which plenty of valuable knowledge is involved.Nonetheless,more than enjoying the advantages of such information,people are at a loss.We call it as “information overload” or “information loss”.Thus it becomes an urgent demand how to make the users have a quick access to the required information from the mass.In good time comes up personalized recommendation system.Personalized recommendation system is an intelligent recommendation system which is able to provide personalized service according to users’ interest.It can recommend only the right objects for users by filtering out redundant data on the basis of some algorithm,which greatly reduces the cost of searching resources.As a matter of fact,personalized recommendation has become one of the most powerful tools to solve the problem of information overload.Collaborative filtering is one of the most core techniques of the Recommendation system as well as a technology most widely and successfully used.Different from many traditional algorithms,Collaborative filtering does not need to consider the content of the items so that it is relatively easy to implement and has been widely adapted in many large websites.In recent years,researches on the recommendation system is not confined to the aspects of algorithms,but many research hotspots in applications are included,such as Electronic Commerce,libraries,etc.and especially university libraries.With the target of collaborative filtering algorithm and university library,this essay aims at the solution to some problems encountered in application of collaborative filtering algorithm,such as cold start and low user satisfaction.In terms of collaborative filtering algorithm in recommendation system,the researches on theories and application are as follow:In this paper,we conducted comprehensive study about domestic and international research in the field of collaborative filtering,introduced the working process and basic types of collaborative filtering,and summarized up the basic concepts and the key issues about collaborative filtering.Aimed at the massive problems,a calculation of similarity based on user attribute and features of items is proposed.This will make full use of the books in the university library and the user’s own inherent characteristics so as to avoid the problems of data sparse and cold start.Detailed analyzing the relative issues in traditional clustering algorithms,we specifically came up with an improved algorithm which can automatically generate K initial centers distributing relatively uniformly.On this basis,the concept of “match tree” was put forward creatively which could further improve the accuracy in recommendation.In the terms of the sparse users rating data problem,clustering algorithms based on items are combined with improved relevant similarity calculation methods to take the place of the traditional project-related clustering algorithms in neighbors searching,so as to avoid the cold start problem,solve the problem of new users and new projects,and improve the recommendation accuracy and users satisfaction.According to the above research,this paper presents the concept of user attributes similarity and active similarity in libraries,integrated into various algorithms,eventually forming an hybrid collaborative filtering recommendation algorithm.The results show that the improved algorithm can effectively improve the recommendation accuracy,and relieved the cold start problem to a certain extent.