Biologically Inspired Techniques for Data Mining: A Brief Overview of Particle Swarm Optimization for KDD

Swarm Intelligence

Swarm Intelligence, inspired by the biological behavior of animals, birds, and fish, is an innovative intelligent optimization technique (Abraham et al., 2006) (Engelbrecht, 2006). SI techniques are based on the collective behavior of swarms of bees, fish schools, and colonies of insects while searching for food, communicating with each other and socializing in their colonies. The SI models are based on self-organization, decentralization, communication, and cooperation between the individuals within the team. The individual interaction is very simple but emerges as a complex global behavior, which is the core of swarm intelligence (Bonabeau & Meyer, 2001). Although swarm intelligence based techniques have primarily been used and found very efficient in traditional optimization problems, a huge growth in these techniques has been observed in other areas of research. These application areas vary from optimizing the solution for planning, scheduling, resource management, and network optimization problems. Data mining is one of the contemporary areas of application, where these techniques have been found to be efficient for clustering, classification, feature selection and outlier detection. The use of swarm intelligence has been extended from conventional optimization problems to optimization-based data mining.

A number of SI based techniques with many variants have been proposed in the last decade and the number of new techniques is growing. Among different SI techniques, Ant Colony Optimization (ACO) and Particle Swarm Optimization (PSO) are the two main techniques, which are widely used for solving discrete and continuous optimization problems. In the next sections we will discuss the foundation PSO followed by its use in KDD.