DISCRETIZATION OF CONTINUOUS ATTRIBUTES USING GENETIC AND ENTROPY BASED CONCEPT LEARNER

Abstract

Genetic Algorithm (GA) based concept learner is widely used in supervised learning system for attribute based spaces. The conventional GA based operators can not directly deal with continuous attributes. We have used two separate approaches one is pure Genetic Algorithm and another is Entropy based approach coupled with Genetic Algorithm for converting continuous attributes into discrete ones. Later on these converted discrete attributes have been used in traditional GA based concept learner to enhance its performance. In our experiments with benchmark data set it has been revealed that statistically significant improvement has been achieved with the proposed technique.