Application of Ant-based Template Matching for Web Documents Categorization

The self-organization behavior exhibited by ants may be modeled to solve real world clustering problems. The general idea of artificial ants walking around in search space to pick up, or drop an item based upon some probability measure has been examined to cluster a large number of World Wide Web (WWW) documents. However, this idea is extended with the direct application of template matching with a Gaussian Probability Surface (GPS) to constrain the formation of the clusters in pre-defined areas of workspace with these multi-agents in this paper. Some comparisons between the clustering performance of supervised ants using GPS against the typical ants clustering algorithm are shown. Their performance are evaluated on the same dataset consisting of a collection of multi-class web documents. Finally, the paper concludes with some recommendations for further investigation. Povzetek: Tehnike kolonij mravelj so bile uporabljene za kategorizacijo internetnih dokumentov.