Experimental Comparison of Metaheuristics for Feature Selection in Machine Learning in the Medical Context

We explore in this paper the use of metaheuristics to select features from a dataset in order to improve the prediction performance of models build with different machine learning methods. To this end, we compare the performances of 5 learning methods: Logistic Regression (LR), K-Nearest Neighbors (KNN), Gaussian Naive Bayes (GNB), Support Vector Machine (SVM) and Random Forest (RF) on 4 heterogeneous datasets in the number of data and features, for different feature selection methods (metaheuristics or statistical filters). The results obtained show that feature selection by improving a metaheuristic derived from the genetic algorithm leads to much better performances no matter the learning method used compared to without feature selection on the same dataset.