Feature Selection Using Genetic Algorithm for Classification of Clinical Data

In this paper, we have proposed a Genetic Algorithm based feature selection approach for clinical decision support system, which is designed to assist physicians with decision making tasks, as to discriminate healthy people from those with Parkinson's disease. We have compared the performance of Genetic Algorithm with two feature ranking algorithms namely Chi-Square algorithm and Information Gain. The genetic algorithm that we propose is wrapper based scheme where the fitness of an individual is determined based on the ability of the selected features to classify the training dataset. To measure the performance of the feature selection algorithms, two different types of standard classification algorithms were implemented namely Bayesian Classifier and K-Nearest Neighbor (K-NN) Classifier. We determine which feature selection algorithm is best suited for clinical datasets under consideration. Experiments show that Genetic Algorithm would be the best choice for feature selection in Parkinson's clinical dataset.