Session identification based on linked referrers and web log indexing

摘要

Web usage mining has been widely adopted in various fields such as optimizing site structure, user-behavior analysis, personalized web services and system performance tuning. Although much research has been done against web log mining algorithms and log pre-processing techniques, the study of efficient retrieval of the structured contents for web usage mining is seldom reported. In this paper, we first show that people are much more interested in discovering user navigation patterns based on various path-sources. Then, we present a novel session identification algorithm Referrer Link based on discovering linked referrers to serve source-oriented path mining. Next, an efficient web log indexing and path extracting technique is introduced to provide structured web log data for general purpose log mining. The experimental results has shown that the accuracy of the mining results conducted against the sessions discovered by the proposed Referrer Link algorithm is 10% higher in average compared with Time-out approach.