Code and Workflow for OSA Severity Prediction

Abstract

This repository contains the complete, reproducible machine learning workflow for the above-titled manuscript. The workflow includes data preprocessing (Min-Max normalization, SMOTE balancing, 75/25 stratified train-test split), model training (Random Forest, XGBoost, Neural Network, SVM, Logistic Regression), and statistical evaluation with bootstrapped 95% confidence intervals. Contents: 1_preprocessing.ows – Orange workflow for data preprocessing 2_model_training.ows – Orange workflow for model training and evaluation bootstrap_analysis.py – Python script for confidence interval calculation weka_smote_log.txt – WEKA SMOTE execution log README.md – Detailed reproduction instructions