An Approach of Community Detecting Based on Block Level Link Analysis

With the explosive increment of the information in the Web, how to efficiently understand collective behaviors and emerging phenomenon in the WWW is becoming a serious problem. Web community, which can be recognized as a set of web units, usually the pages, based on a common topic, is a significant structure in the Web. An efficient way to detect a community corresponding to a specific topic can greatly help users to obtain useful information. However, existing community detecting algorithms are all based on pages as units, the topic drift phenomenon becomes the inherent problem of each existing community detecting algorithm since the Web pages always have multiple topics in content. This paper puts forward a new community detecting algorithm based on blocks as units, which combines the page dividing algorithm with the primary community detecting algorithm, overcoming the defect in the existing community detecting algorithms. The communities obtained by the new algorithm have much more definite topic in content, which is more meaningful for further data processing such as information extraction and so on.