2009 International Conference of Soft Computing and Pattern Recognition
Cat Swarm Optimization for Clustering
Budi Santosa
Mirsa Kencana Ningrum
Department of Industrial Engineering Institut Teknologi Sepuluh Nopember (ITS) Surabaya, Indonesia
[email protected]
Department of Industrial Engineering Institut Teknologi Sepuluh Nopember (ITS) Surabaya, Indonesia
[email protected] ACO and PSO algorithm have already been developed to be applied in clustering case. Meanwhile, CSO algorithm as a new algorithm, up to this research is being proposed, has only been applied in function optimization problem. This research mainly focuses on developing CSO for solving clustering problem. CSO algorithm has proven to be effective in solving the optimization problem. It comes up with a better result than PSO [2].
Abstract—Cat Swarm Optimization (CSO) is one of the new heuristic optimization algorithm which based on swarm intelligence. Previous research shows that this algorithm has better performance compared to the other heuristic optimization algorithms: Particle Swarm Optimization (PSO) and weighted-PSO in the cases of function minimization. In this research a new CSO algorithm for clustering problem is proposed. The new CSO clustering algorithm was tested on four different datasets. The modification is made on the CSO formula to obtain better results. Then, the accuracy level of poposed algorith was compared to those of K-means and PSO clustering. The modification of CSO formula can improve the performance of CSO Clustering. The comparison indicates that CSO clustering can be considered as a sufficiently accurate clustering method.
II.
Chu et al. [3] divided CSO algorithm into two submodels based on two of the major behavioral traits of cats. These are termed ”seeking mode” and ”tracing mode”. In CSO, we first decide how many cats we would like to use in the iteration, then we apply the cats into CSO to solve the problems. Every cat has its own position composed of D dimensions, velocities for each dimension, a fitness value, which represents the accommodation of the cat to the fitness function, and a flag to identify whether the cat is in seeking mode or tracing mode. The final solution would be the best position of one of the cats. The CSO keeps the best solution until it reaches the end of the iterations.
Keywords- Swarm Intelligence, Cat Swarm Optimization, clustering, k-means clustering, Particle Swarm Optimization
I.
INTRODUCTION
De Castro [6] noted that in the last two decades, many advances on the computer sciences have been based on the observation and emulation of processes of the natural world, or so-called bioinspired informatics. Still, de Castro [8] believed that bioinspired informatics are traced to the development of perceptrons and artificial life, which tried to reproduce the mental processes of the brain and abiogenesis, respectively, in a computer environment. Moreover, Chu and Tsai [2], whom called the computational artificial life as computational intelligence, noted that computational intelligence is a hot research topic and many related algorithms have been proposed in recent years. These algorithms generally were being proposed to solve the optimization problems. Some of these optimization algorithms were developed based on swarm intelligence by simulating the intelligent behavior of animals, like Ant Colony Optimization (ACO) which imitate the behavior of ants, Particle Swarm Optimization (PSO) which imitate the behavior of birds, and the recent finding, Cat Swarm Optimization (CSO) which imitate the behavior of cats. Along with the demanding problems from many fields in need of a better and faster solution, these algorithms have begun to develop so that they can be used in many fields, with many problems. Among of them are the Shortest Path problem, Travelling Salesman Problem (TSP), Vehicle routing Problem (VRP), scheduling, and data mining, especially clustering problem. 978-0-7695-3879-2/09 $26.00 © 2009 IEEE DOI 10.1109/SoCPaR.2009.23
CAT SWARM OPTIMIZATION (CSO)
A. Seeking Mode This sub model is used to model the cat during a period of resting but being alert- looking around its environment for its next move. Seeking mode has four essential factors, which are designed as follows: seeking memory pool (SMP), seeking range of the selected dimension (SRD), counts of dimension to change (CDC) and self position consideration (SPC). Seeking mode according to Chu et al. [3] is described below. Step 1: Make j copies of the present position of cat k, where j = SMP. If the value of SPC is true, let j = (SMP − 1), then retain the present position as one of the candidates. Step 2: For each copy, according to CDC, randomly plus or minus SRD percents the present values and replace the old ones. Step 3: Calculate the fitness values (FS) of all candidate points. Step 4: If all FS are not exactly equal, calculate the selecting probability of each candidate point by equation (1), otherwise set all the selecting probability of each candidate point be 1. Step 5: Randomly pick the point to move to from the candidate points, and replace the position of cat k. 54
Authorized licensed use limited to: University of Oklahoma Libraries. Downloaded on May 17,2010 at 23:01:41 UTC from IEEE Xplore. Restrictions apply.
Pi=
|SSEi -SSE max| ,0
