Pattern classification with genetic algorithms

A method is proposed for finding decision boundaries, approximated by piecewise linear segments, for the classification of patterns in R 2, using an elitist model of a genetic algorithm. It involves the generation and placement of a set of lines (represented by strings) in the feature space that yields minimum misclassification. The effectiveness of the algorithm is demonstrated, for different parameter values, on both artificial data and speech data having non-linear class boundaries. Its comparison with the k-NN classifier is also made.