Hybrid Genetic Algorithm with SVM for Medical Data Classification

In medical data classification system, several parameters can affect its performance, notably, the quality of the features which poses problems in real applications. Some of the attributes are redundant while others are irrelevant, or are even unnecessary to the classification problem. Feature selection plays a crucial role in medical data analysis by identifying and removing irrelevant features from the training data. In this work, a feature subset selection method is proposed using hybridization of a genetic algorithm with a simulated annealing meta-heuristic and combined with SVM classifier. It tries to reduce the initial size of data and to select a set of relevant features to enhance the accuracy and speed of classification system. For evaluation, the proposed method is applied to eleven public medical datasets and then compared to two other methods of feature selection applied on the same datasets. Experimental results have shown that the proposed method with optimized SVM parameters gives competitive results and finds good quality solutions with small size.

I. Introduction

Clinical information systems store a large amount of information in medical databases. The manual processing of these data for the purpose of diagnosis becomes very difficult. This is giving rise to a growing interest in the development of automated assessment methods for disease monitoring. Classification is one of the techniques of data mining which involves extracting a general rule or classification procedure from a set of learning examples. The classification of medical data consists of building predictive models from medical datasets for the purpose of improving the quality of health care [1].