Research On Application Of Web Log Mining In Web Recommendation Service

With the rapid development of Internet, People are facing the situation of "information overload" and "resource lost" when obtaining a lot of information. The contradiction between diversity of network information and specialization of user requirements gradually becomes a problem that plagues website and users. To resolve this problem, Web Recommendation Service, that is Personalized Information Services, was proposed. Based on User Interest Model, Web Recommendation Service could predict the information, which users are interested in. After considerable seriously studying and researching on principle and procedure of Web Log Mining, this paper designed a website recommendation service system based on Web Log Mining, which is relying on a special software platform—Yeemu Accelerator. With a series of treatment and analysis on web log files, the system collects web pages after web users surf the internet. User Interest Model is established via cluster analysis. Based on this model, the system provides personalized information service to web users.During data preprocessing for web log mining, the traditional procedure of data preprocessing is improved in the paper. The improvement not only reduces the difficulty and workload of data preprocessing, but also ensures the quality of data preprocessing. Because of their own characteristic of dividing clustering algorithm and hierarchical clustering algorithm, the algorithm combining split clustering algorithm and k-means clustering algorithm is contrived in the course of cluster analysis, which is high efficient and Suitable for dealing with large amounts of data. According to the respective characteristics, different recommendation algorithms for personal navigation and personal page are put forward in the process of designing personal recommendation service. The experiment results show high practicality of algorithm and strategy in the paper.