Tthe Ppeerrsonalized Recommendation System In Digital Library

With the rapid development of the library informatization, thetraditional information retrieval is difficult to meet the needs of thelibrary users. Because the form of keyword combination has difficultyreflectingthe user needs, the personalizedrecommendationservices havebeen widely provided. In the recommendation service, the valuablehidden information can be founded with user information, bookinformation,browsinghistory, etc andwill be recommendedto the user.Compared with the main recommendationalgorithms which arewidely researched and applied, association rules are proposed in thispaper and improved with consideringthe library situation. In the libraryrecommendationsystem, association rules do not require the ratings ofthe books, the characterizations of the resources and other additionalinformation. However, it also has shortcoming.Firstly, the new storagebooks are hardly to be recommended; Secondly, the high-view booksfrequently appear in the recommendation information but have lowrecommendationvalue.Lastly,associationrules havea poor performanceon findingthe relationshipamongsimilarbooks. The concept of multipleminimum supports, MMS, is added intoApriori algorithm in this paper. Meanwhile,two new factors, the bookstoragetime andthe book view,are introducedintothe formulaof MMS.Moreover,We shoulddouser clustering,the algorithmbased on k-means,with considering user characteristic prior to association rules, whichmakes users acquire recommendation information from the associationrulesgeneratedin the similarusers. In consideringthe libraryusers, usercharacteristicincludes grade, major and interest which are non-numeric,which makes the mixed-attribute distance function be applied in theclustering. Based on the above improvements, the new algorithm caneffectivelysolvethe issueswhichare difficultforApriori.Basedon thisimprovedalgorithm,a new recommendationsystemis designedandimplementedfor libraryusers. Experimentalresultsshowthat the improved algorithm can effectively improve the deficiency.Meanwhile,it not only increases the utilization of library resources, butalso improves the qualityof the personalizedrecommendationservice.