IJCSNS International Journal of Computer Science and Network Security, VOL.9 No.10, October 2009 132 Manuscript received October 5, 2009 Manuscript revised October 20, 2009 Comparative Study of Particle Swarm Optimization based Unsupervised Clustering Techniques Mr.V.K.Panchal 1 , Mr. Harish Kundra *2 , Ms. Jagdeep Kaur #3 1 Scientist ‘G’, Defence Research Organization, Delhi, India *2 HOD (Computer Sc.& Engg.) , Rayat College Of Engg. & IT, Ropar, Punjab, India #3 Lecturer (Computer Sc.& Engg.), Rayat College Of Engg. & IT, Ropar, Punjab, India Summary In order to overcome the shortcomings of traditional clustering algorithms such as local optima and sensitivity to initialization, a new Optimization technique, Particle Swarm Optimization is used in association with Unsupervised Clustering techniques in this paper. This new algorithm uses the capacity of global search in PSO algorithm and solves the problems associated with traditional clustering techniques. This merge avoids the local optima problem and increases the convergence speed. Parameters, time, distance and mean, are used to compare PSO based Fuzzy C-Means, PSO based Gustafson’s-Kessel, PSO based Fuzzy K-Means with extragrades and PSO based K-Means are suitably plotted. Thus, Performance evaluation of Particle Swarm Optimization based Clustering techniques is achieved. Results of this PSO based clustering algorithm is used for remote image classification. Finally, accuracy of this image is computed along with its Kappa Coefficient. Key words: Particle Swarm Optimization(PSO), Fuzzy C-Means Clustering (FCM), K-Means Clustering (K-Means), Swarm Clustering, Gustaffsons-Kessel Clustering (GK), Unsupervised Clustering, Remote Sensing, Image Clustering, Image Classification. 1. Introduction Image clustering [8] can be defined as the identification of natural groups within a multispectral data set. The algorithm that performs clustering functions to partition a set of objects (pixels) into relatively homogenous subsets based on inter-object similarities with little or no overlap. In general, clustering methods can be categorized by principle (objective function, graph theoretical, hierarchical) or by model type (deterministic, statistical, heuristic, fuzzy). In the traditional clustering algorithm, the samples are classified in the unique cluster, which is all known as a hard division. However, there is not definite boundary in most things. The concept of fuzzy clustering applies to the essence of most things, and reflects the reality of objects better. Clustering algorithms are usually applied to feature space, and as such they do not use any spatial information (e.g. the relative location of the patterns in the feature space). One major limitation of many classical clustering algorithms is that they assume that the number of clusters is known. However, in practice, the number of clusters may not be known. This problem is sometimes called unsupervised clustering. Unsupervised prototype-based clustering aims at determining the correct number of clusters, C, without any prior knowledge about it. 2. Data Mining Data Mining is an analytic process designed to explore large amounts of data in search of consistent patterns and/or systematic relationships between variables, and then to validate the findings by applying the detected patterns to new subsets of data. In order to achieve this, data mining uses computational techniques from statistics, machine learning and pattern recognition. This paper focuses on available data mining techniques for unsupervised clustering of remote images. 3. Objectives The objective of this paper is to enhance the quality of the satellite image by placing the pixel into it’s most appropriate land cover, for this we need to- • To develop an efficient clustering algorithm based on PSO. • Help researchers in comparing different clustering algorithms and generate benchmarks. • To develop an efficient clustering algorithm that can find the “optimum” number of clusters in a data set • To show that PSO can bring out results with in reduced iterations.