Supervised Learning

ML Pipeline · Chest X-Ray Classification

Overview

End-to-end supervised learning pipeline for multi-class classification of chest X-ray images into COVID-19, Pneumonia, and Normal categories. Eight algorithms evaluated across classical ML, ensemble, and deep learning paradigms.

Models Evaluated

Model	Accuracy	AUC-ROC	Type
YOLOv8-cls	0.9767	0.9980	Deep Learning
ResNet50 (Transfer)	0.9746	0.9973	Deep Learning
MLP	0.9682	0.9930	Neural Network
SVM (RBF)	0.9640	0.9955	Classical — Best
KNN (k=7)	0.9301	0.9796	Classical
Random Forest	0.8475	0.9614	Ensemble
Naive Bayes	0.6377	0.8237	Probabilistic

Dataset

• 3,141 images · 224×224 RGB · Classes: COVID / Pneumonia / Normal
• Split: 70% train / 15% val / 15% test (stratified)
• Class imbalance handled via class_weight='balanced'
• PCA: 300 components · 95% explained variance

Tech Stack

Python 3.11 · scikit-learn · TensorFlow 2.x · PyTorch · OpenCV · NumPy · Pandas · Matplotlib

---

Maestría en Inteligencia Artificial · Universidad Politécnica Metropolitana de Hidalgo