Design And Implementation Of Personalized Documents Recommender System

With the rapid development of the internet technology and the popularization of the applications, the internet is entering the Recommendation Era from the Search Era. Nowadays, the websites take the initiative to provide the information the users are interested in without the users inputting the keywords. Since the recommender system can improve their satisfaction with and loyalty to the websites, designing the personalized recommender system has important practical significance.Based on the results of the analyses of the common personalized recommender systems and the characteristics of the enterprises’office platform, a personalized document recommender system is designed in the thesis. The research mainly includes the following aspects:1) On the basis of analyzing the records of the users’behavioral data, this dissertation extracts the users’behavioral data from the records and abstracts them to the users’behavioral eigenvectors to build the user interest model and the user behavior model. By virtue of these two models, the interest of the users can be studied and the correlative relationships between the users and between documents can be figured out.2) The user similarities, document similarities and content similarities can be calculated based on the collaborative filtering idea whereby the nearest-neighbor list can be obtained. After elements on the nearest-neighbor list are filtered and ranked, the final document list for recommendation can be got and showed to the users, which has three methods of recommendation, namely "the recommended contents","the relative contents" and "what others have searched".3) Considering that common recommender systems hardly attend to the users’authority while the users’authority of the enterprises’office platform is very complicated, this research has figured out the recommended proposals based on the users’ authority. In addition, this study employs the off-line and the incremental calculation methods together to process the users’log files and calculate the similarities to improve the operation efficiency.After analyzing the users’needs and studying the working process of personalized recommender systems, this thesis has designed a personalized document recommender system with high accuracy and satisfaction from the users. This system has proved to be satisfactory after a series of function tests.