Data Classification Using Variation of Genetic Programming Fitness Function

Abstract

Genetic Programming (GP) is a technique that deals with evolving computer programs using biologically inspired methods. GP is a set of instruction and a fitness function to evaluate the best solution. The objective of GP is to find a computer program capable of solving a predefined problem. GP has capability to select the useful features for the new generation and discard the unwanted features during evolution. In this paper, GP is used for real world classification problems. Five real world problems are used to evaluate the GP performance. In this paper, Gaussian Distribution Criteria, Standard Accuracy Method, Average Class Accuracy Method and Artificial Neural Networks (ANN) are used for the evaluation of fitness function for binary classification problems. A number of experiments are carried out to evaluate and compare the results obtained from GP. Results prove that GP (ANN) provide a better accuracy as compared to others methods.