ABSTRACT

With the explosive growth of WWW, the web mining techniques are densely concentrated to discover the relevant behaviors of the web user from the web log data. In fact the pattern discovery techniques generate many hundreds, often thousands, of patterns, that are unwanted, unexpected, disputable and unbelievable in nature. The success of representing the real knowledge out of such patterns is highly reliant on the pattern analysis stage in investigating the web user usage behavior. To retain most genuine and interesting patterns it is necessary to filter out unqualified patterns and use more sophisticated visualization techniques to present the knowledge of web user usage effectively. The authors in the present paper propose an Analysis and Knowledge Representation System (AKRS) that equally concentrates on both knowledge identification and representation. The key measures are combinedly used for the knowledge identification as a three phase filtering system, to determine the interestingness of patterns in the proposed AKRS. Initially, the objective measures applied on the patterns discovered by pattern discovery techniques to filter out the patterns that do not meet statistical strengths with the frame work of interest factor. Later, subjective measures are applied to identify the patterns that are of most genuine interestingness based on web knowledge. Finally, the heuristic measures evaluate the semantics of patterns based on both user specific objectives and utility of mined patterns. The measures of AKRS efficiently determine the correlation among the most interesting patterns. In addition, to meet the challenges in knowledge representation, like identifying relevant information, finding the depth of information and achieving the visualization competency, the proposed AKRS also designates the recent knowledge visualization techniques like multidimensional and specialized hierarchical. The AKRS amplifies the truthfulness and rate of success in representing final knowledge of web user behavior.