Class Prediction Based on Gene Expression: Applying Neural Networks via a Genetic Algorithm Wrapper

This project focuses on applying neural networks to the classification of biological state based on gene expression data. In order to take advantage of the non-linear classification abilities of neural networks, a genetic algorithm is employed as a “wrapper” feature selector. Results indicate that the genetic algorithm effectively identifies features that allow successful neural network training. In addition, it is shown that ensembles created by combining neural networks from multiple runs of the genetic algorithm consistently outperform single networks.