Decision function estimation using intelligent gravitational search algorithm

Abstract

There are various kinds of supervised classification techniques such as Bayesian classifier, k nearest neighbor, neural network and rule based classifiers. A kind of supervised classifier, estimates the necessary decision hyperplanes for separating the feature space to distinct regions for recognizing unknown put patterns. In this paper a novel swarm intelligence based classifier is described for decision function estimation without requirement to priory knowledge. The utilized swarm intelligence technique is gravitational search algorithm (GSA) which has been recently reported. The proposed method is called intelligent GSA based classifier (IGSA-classifier). At first, a fuzzy system is designed for intelligently updating the effective parameters of GSA. Those are gravitational coefficient and the number of effective objects, two important parameters which play major roles on search process of GSA. Then the designed intelligent GSA is employed to construct a novel decision function estimation algorithm from feature space. Extensive experimental results on different benchmarks and a practical pattern recognition problem with nonlinear, overlapping class boundaries and different feature space dimensions are provided to show the capability of the proposed method. The comparative results show that the performance of the proposed classifier is comparable to or better than the performance of other swarm intelligence based and evolutionary classifiers.