Probabilistic Ant based Clustering for Distributed Databases

Abstract

In this paper we present PACE - a probabilistic ant based clustering algorithm for distributed databases. This algorithm is based on the well-known swarm based approach to clustering. Its characteristic feature is the formation of numerous zones in various distributed sites based on the user query to the distributed database. Keywords, extracted out of the query, are used to assign a range of values according to their corresponding probability of occurrence or hit ratio at each site. An ant odor identification model is used as a preceding step to the colony building and formation of clusters inside the zones. Reordering or sorting of the heap trees formed by the ants to enable agglomeration of only the most probable data forms the crux of this algorithm. Experimental results are reported showing the comparison of PACE with other existing clustering algorithms