Clustering algorithms for intelligent web

摘要

Detecting communities of users and data in the web are important issues as the social web evolves and new information is generated everyday. In this paper, we discuss six different clustering algorithms that are related to the intelligent web. These algorithms help us to identify communities of interest in the web, which is very necessary to perform certain actions on specific group such as targeted advertisement. We consider the following algorithms: single-link algorithm, average-link algorithm, minimum-spanning-tree single-link algorithm, K-means algorithm, ROCK algorithm and DBSCAN algorithm. These algorithms are categorised into three groups: hierarchical, partitional and density-based algorithms. We show how each algorithm works and discuss potential advantages and shortcomings. We then compare these algorithms against each other and discuss their ability to accurately identify communities of interest based on social web data which are large datasets with high dimensionality. Finally, we illustrate and discuss our findings through a case study, which involves clustering in online social networks.