Performance Analysis of Machine Learning Algorithms for Medical Datasets

Abstract

AbstractMedical applications using machine learning have gained importance; therefore, research on the applicability and adequacy of algorithms has also gained momentum. Investigating the performance of the algorithms is crucial for the real-life implementation of classification systems as diagnostic or decision support systems. This paper analyzes the performance of six machine learning algorithms, Decision Tree, Random Forest, Gradient Boosting, Extreme Gradient Boosting, Support Vector Machine, and Logistic Regression, on three medical datasets to demonstrate the weaknesses and strengths of the algorithms. Three experiments in two stages, hyperparameter tuning and performance evaluation are performed. Binary classification and multi-class evaluation are considered in order to analyze the performance of the models with varied challenges. Gradient Boosting and Logistic Regression achieved superior results in binary classification experiments, while Random Forest produced superior results in the multi-class experiment. The results suggested that different machine learning algorithms might obtain unstable performances or stable and consistent results.KeywordsGradient boostingLogistic regressionRandom forestMedical datasetPerformance evaluation