ABSTRACT

Density based clustering basically operates by associating related items contained in the sample space. The association is performed by maintaining maximum inter class similarity and minimum intra class similarity. However, the major downside of such approach is that it is time consuming in case of huge datasets. This paper proposes a metaheuristic based density clustering technique that utilizes a modified Particle Swarm Optimization (PSO) for fast and efficient neighbor selection. In this work, the PSO is integrated with simulated annealing to perform faster node selection and the distribution of catfish particles in the search space helps to avoid local optima to the maximum extent. Experiments were conducted with real-time spatial datasets and it was identified that the proposed clustering technique performs effectively in terms of both time and efficiency.