An optimized LVQ algorithm for multi-interval discretization of continuous values

Abstract

Discretization of continuous valued features is an important problem to consider during classification learning. There already exist a number of successful discretization techniques based on LVQ algorithm. In this paper, we have approached the problem of discretization from a different angle, and have proposed an algorithm based on optimization of Learning Vector Quantization (LVQ) with Genetic Algorithm (GA). LVQ has been employed to function as a classification algorithm and discretization is performed using this classification nature of LVQ algorithms. We have modeled a GA based algorithm, which enhances the accuracy of the classifier.