Automated design of genetic programming classification algorithms using a genetic algorithm

Abstract

There is a large scale initiative by the machine learning community to automate the design of machine learning techniques to remove reliance on the human expert, providing out of the box software that can be used by novices. In this study the automated design of genetic programming classification algorithms is proposed. A number of design decisions have to be considered by algorithm designers during the design process and this is usually a time consuming task. Our automated design approach uses a genetic algorithm to automatically configure a genetic programming classification algorithm. The genetic algorithm determines parameter values and sets the flow control for the classification algorithm. The proposed system is tested on real world problems and the results indicate that induced classifiers perform better than manually designed classifiers.