Research On Methods Of Mining User Navigation Patterns And Application

With the rapid development in WWW, web server on internet has accumulated a large number of web logs. Web usage mining is one of the hot current studies and belongs to a multi-interdisciplinary area of research, which involves database technology, artificial intelligence, neural network, pattern recognition, statistics, fuzzy sets, rough sets and many other disciplines. After mining the web logs, we can discover some deep-level knowledge and rules-the navigation behavior and interests of web users. The knowledge and rules can be applied to many fields such as web personalization, web sytem improvement and business intelligence etc.This dissertation mainly focuses on the research on methods of mining user navigation patterns and application. Major works are listed as follows:1. Accordding to the four steps of mining user navigation patterns system: data preparation, interest navigation patters discovery, navigation patters clustering and application, this dissertation firstly collects and sums up the clasical and latest researches at home and abroad.2. To better mine the user usage patterns, the user path selection interest measure and the page navigation interest measure are proposed on the analysis of web usage imformation.According to the conception, an algorithm is proposed based on the path selection interest matrix and user interest navigation matrix.Firstly, two URL-URL matrices are set up from web logs according to navigation paths and user navigation interest by page. Then navigation interest sub-paths could be discovered from the computation of the two matrices. Finally, all the sub-paths were combined.3. Different web pages may have different interest. The interest of each web page is evaluated by managers and comprehensive fuzzy evaluation methods are used to evaluate the interest measure of each web page. Besides, the time duration on a wep page is an important feature in analyzing users'navigation behavior. In order to neglect the tiny difference on time durations, time durations are characterized as fuzzy linguistic variable. Then a new fuzzy web-mining algorithm is proposed, which can process to mine fuzzy navigation interest patterns with linguistic variables from web logs effectively. The fuzzy navigation interest patterns are more natural and more similar to people reasoning.4. An effective approach for clustering web users is proposed based on a function of the longest common subsequence of their clickstreams that takes into account both the navigation interest and the selection interest on each page. At the same time ,In order to reduce the dimension of sessions and improve the efficiency of the algorithm , the sessions are then generalized according to the page hierarchy.5. A fuzzy rough approximation-based approach is applied to cluster user navigation patterns from web logs. In this approach each user navigation pattern is denoted by a fuzzy vector representing visited web pages and time durations during a surfing, in which only the linguistic term with the maximum cardinality for a page is chosen to describe the characteristic of the web page. Then a new distance is defined to measure similarity between two sessions. Finally rough approximation approach adopted to cluster user navigation patterns.6. By analyzing the user navigation patterns and frequent navigation paths, an algorithm is proposed to get the major navigation sub website structure and it can be used to construct wap website for the mobile phone visitors, also it can offer business application to web managers.7. To provide the personalized recommendation according to web users'interest, the dissertation presents a recommendation approach based on the clustering analysis and the weighted association rules. In this approach, we extend the traditional association rule problem by allowing a weight to be associated with each item in a transaction to reflect the user interest. We assign a significant weight to each page based on the time spent by user on each page and visiting frequency of each page, taking into account the degree of interest. Then the weighted association mining technique is applied to mine rule in clusters to provide real time personalized recommendation for users.Finally, this dissertation summarized the author'works and discussed the future works.