Feature Selection Method using Genetic Algorithm for Medical Dataset

There is a massive amount of high dimensional data that is pervasive in the healthcare domain. Interpreting these data continues as a challenging problem and it is an active research area due to their nature of high dimensional and low sample size. These problems produce a significant challenge to the existing classification methods in achieving high accuracy. Therefore, a compelling feature selection method is important in this case to improve the correctly classify different diseases and consequently lead to help medical practitioners. The methodology for this paper is adapted from KDD method. In this work, a wrapper-based feature selection using the Genetic Algorithm (GA) is proposed and the classifier is based on Support Vector Machine (SVM). The proposed algorithms was tested on five medical datasets naming the Breast Cancer, Parkinson’s, Heart Disease, Statlog (Heart), and Hepatitis. The results obtained from this work, which apply GA as feature selection yielded competitive results on most of the datasets. The accuracies of the said datasets are as follows: Breast Cancer - 72.71%, Parkinson’s - 88.36%, Heart Disease - 86.73%, Statlog (Heart) - 85.48 %, and Hepatitis - 76.95%. This prediction method with GA as feature selection will help medical practitioners to make better diagnose with patient’s disease.