Genetic Algorithm Based Learning Using Feature Construction

Genetic algorithms (GAs) are excellent for learning concepts that span complex space, especially those with a large number of local optima. Learning algorithms, in general, perform well on data that has been pre-processed to reduce complexity. Several studies have documented their effectiveness on raw as well as pre-processed data using feature selection, etc. Unlike other learning algorithms (e.g., those in feedforward neural networks), GAs are not particularly effective in reducing data complexity while learning difficult concepts. Feature construction has been shown to reduce complexity of space spanned by input data. In this paper, we present an algorithm for enhancing the learning process of a GA through the use of feature construction as a pre-processing step. We also apply the same procedure on two other learning methods, namely C4.5 and Lazy Learner, and show improvement in performance.