Research And Design Of Personalized Service Model In University Digital Library

In the Digital Library contains the mass information, in such a vast resources ofeasy to fall into the "information trek". To solve the problem,"information trek"personalized service arises at the historic moment, but the present digital libraryindividualized service most cannot satisfy user variety and individuality demand, thelack of initiative push type personalized service. According to the personal interest,will the user really interested in information offered to him, can truly "each required"is the core of individualized service system.Through the domestic and international personalized service key technologyresearch, analyzes the mainstream cooperative filter technology principle,characteristics and method, it is concluded that the digital library after field, thecollaborative filtering mechanism, though that can optimize personalized service. Butalong with the digital library is widely used, the user number and quantity is increaseddramatically, users score data sparse problem became apparent, difficult to determineorthodox similar user group, so the quality of collaborative filtering sharply. Intraditional collaborative filtering algorithm does not suitable to describe the user’sinterest, the integrity of the defects is put forward based on digital library usersmodeling method, the classification of the user’s interests into to subdivisisionselecting projects, and introducing Chinese project time and interest weights of thethreshold of concepts to research the user in different subdivisision interest categoryover time show interest differences and change, realize the user’s interest preferenceof quantitative measure. Proposed based on Chinese classification and clusteringcooperative filter recommendation algorithm are proposed. Based on the userbackground characteristics in initial clustering, on the basis of classification based onseveral projects-a user subdivisision concept, classification matrix calculatedrespectively similar to user groups. Thereby narrowed the scope, neighbours reducesthe time complexity and between two users may because "local point" is similar andbecome a subdivisision classification interests neighbour to produce usefulrecommendations.Aiming at the improved algorithm proposed in the library has selectedcorresponding data experiment, experiment shows that the improved algorithm thantraditional collaborative filtering algorithm in recommending precision was obviouslyimproved.