Research On Frequent Web Access Pattern Discovery Based On Homogeneous Semantic Log Knowledgebase

Frequent Web access pattern discovery is the key element of Web usage mining and can be used to find interesting user access information from Web data.It is helpful for companies to improve the design of existing sites and servers and enhance the quality of user services.Semantic Web enables current Web semantics which can be understand and explain by computers,so it can effectively improve the efficiency of Web usage mining. The combination of ontologies and rules can make best use of the advantages and bypass the disadvantages,and is critical to improve the efficiency of Web usage mining.In this thesis we start from the combination of log ontologies and datalog rules,and then focus on the discovery of frequent Web access patterns.The research works are as follows:1.Improving the formal description of log ontology by defining it as a six tuple.The relationship of domain field are indicated by application rules.The hierarchical formal description of ontology are also improved in this thesis.Compared to the origin definition,our improvement can not only avoid complex relationship in the field of log ontology but also meet the requirements of Web usage mining.2.Building a homogeneous log knowledge base on the basis of intergration of log ontology knowledge base and Datalog rule base. On the basis of existing research we improve the approach for translating logic description based knowledge base into disjunctive Datalog rule.The knowledge base after translated can be integrated by existing DL-safe Web rules to construct a homogeneous log knowledge base.3.Improving an algrithem for mining frequent Web access patterns and association rules on the basis of the algorithm FARMER. The algorithm is constructed by the trie frequent pattern tree, and we can discover the frequent pattens and expand the tree nodes by pattern validation and support calculation.4. Designing and implementing a frequent Web access pattern discovery system. Integrated with the previous research results, we develop a system which is used for frequent Web access pattern mining. The system can import ontologies and rules,with the function of constructing homogeneous log knowledge base by translating log ontology into disjunctive rules.It also implements the algorithm proposed in the thesis which based on the depth-first pattern mining method. The simulation results verify the feasibility of the theory.