A Comparative Study of Machine Learning Algorithms for Early-Stage and Rare Disease Diagnosis Using Highly Imbalanced Datasets

Abstract

The implementation of accurate, reliable, and interpretable diagnostic tools is a pressing need in modern healthcare. This study addresses this challenge through a comprehensive comparative analysis of machine learning algorithms—including Support Vector Machines (SVM), Random Forest, Neural Networks, and Logistic Regression—for medical diagnosis prediction. Based on systematic experimentation across multiple datasets, our results demonstrate the superior performance of ensemble methods, with Random Forest excelling in accuracy, sensitivity, and specificity. The findings provide valuable insights for healthcare practitioners and contribute significantly to the advancement of clinical decision support systems.