| The thesis aims at improving the application level of library data from low level on-line query to a higher level of decision support and analytical predication via employing data mining technology. By collecting, analyzing, summarizing and reasoning service operational data recorded in the automation system with the help of macro, medium and micro level methods, the system can find out the association rules, generalized knowledge and future trend of data. The knowledge mined could be used to direct library and improve its service level, thus ameliorating the management level and reader satisfaction.The thesis systematically studies and analyzes several classic data mining algorithms, especially for meeting the need of mining the library circulation and reducing calculation on irrelavent items, the thesis focused on Apriori algorithm and improved its performance.Based on the improved Apriori algorithm, we designed and implemented a "personalized library collection recommendation system". The system uses categorization and cluster mining technology to find out groups of readers that have similar background and interest based on reader's identity, company and the kind and number of books they borrowed . In addition, the system uses association rule mining technology to find out the most interested books in a group to assist the librarians in book recommendation service. The experimental results show that the designed mining scheme is reasonable to distinguish group first and then make recommendations based on reader's interests. The system's response is quick and the books recommended normally can meet reader's demand. |
