Agent-Based Non-distributed and Distributed Clustering

摘要

The paper deals with the non-distributed and distributed clustering and proposes an agent-based approach to solving the clustering problem instances. The approach is an implementation of the specialized A-Team architecture called JABAT. The paper includes an overview of JABAT and the description of the agent-based algorithms solving the non-distributed and distributed clustering problems. To evaluate the approach the computational experiment involving several well known benchmark instances has been carried out. The results obtained by JABAT-based algorithms are compared with the results produced by the non-distributed and distributed k-means algorithm. It has been shown that the proposed approach produces, as a rule, better results and has the advantage of being scalable, mobile and parallel.