Analysis of the Usage of Chaotic Theory in Data Clustering Using Particle Swarm Optimization

Clustering has been a popular topic of research for many years. In fact, the first clustering methods were statistical methods which were used prior to computer age. Clustering plays an important role in lots of sciences such as medicine, astronomy, economy, etc. In recent years, application of artificial intelligence methods in clustering gets more attention. Particle Swarm Optimization is one of the optimization algorithms which proposed in 1995 and in spite of being relatively new, has gained popularity. Using PSO algorithms in clustering started from 2003 and from that time, different methods has been proposed. A clustering method should partition the data set so that the most related data items are placed in the same group. In this paper, we tried to make a novel attempt to apply PSO and chaotic theory to clustering problem. In our proposed solution, the fitness function is changed and then, using chaotic map functions, the initialization method of PSO algorithm has been altered. Then, we tried to improve the results using a specified mutation technique called chaotic mutation. Finally, we applied inertia weights to improve our results to a new level. The empirical results shows using the suggested fitness function alongside chaotic and inertia weights improve the results hugely.