heart-disease-ml-classifier

A PyTorch model designed to predict the risk of heart disease based on a combination of symptoms, lifestyle factors, and medical history from 70,000+ data samples. The model achieves approximately 99.27% accuracy on test data.

Download the model here

Kaggle Dataset link

Overview

This project uses PyTorch to build a neural network classifier for heart disease risk prediction. The model analyzes several medical predictor variables to determine if a patient is at risk of heart disease.

Dataset

The dataset contains medical predictor variables from the heart_disease_risk_dataset_earlymed.csv file.

It contains 18 medical predictors of heart disease:

1. Chest Pain: Presence of chest pain (Yes/No)
2. Shortness of Breath: Difficulty breathing (Yes/No)
3. Fatigue: Feeling of tiredness (Yes/No)
4. Palpitations: Irregular heartbeat sensations (Yes/No)
5. Dizziness: Feeling lightheaded (Yes/No)
6. Swelling: Edema in extremities (Yes/No)
7. Pain Arms Jaw Back: Pain radiating to arms/jaw/back (Yes/No)
8. Cold Sweats Nausea: Presence of cold sweats or nausea (Yes/No)
9. High BP: High blood pressure diagnosis (Yes/No)
10. High Cholesterol: High cholesterol diagnosis (Yes/No)
11. Diabetes: Presence of diabetes (Yes/No)
12. Smoking: Current smoking status (Yes/No)
13. Obesity: Obesity status (Yes/No)
14. Sedentary Lifestyle: Physical inactivity (Yes/No)
15. Family History: Family history of heart disease (Yes/No)
16. Chronic Stress: Ongoing stress condition (Yes/No)
17. Gender: Patient's gender (Male/Female)
18. Age: Age of patient in years

Output variable:

• Risk: Risk of Heart Disease (low/high)

Requirements

• Python 3.8+
• PyTorch
• pandas
• scikit-learn
• matplotlib

Usage

1. Clone the repository
2. Create a virtual environment py -m venv .venv and activate it .venv/Scripts/activate
3. Install dependencies: pip install -r requirements.txt
4. Run the model: python main.py

Model Architecture

• Input layer: 18 features
• Hidden layer 1: 64 neurons with ReLU activation
• Hidden layer 2: 28 neurons with ReLU activation
• Output layer: 2 neurons (Binary classification)
• Optimization: Adam optimizer with learning rate 0.005
• Loss function: Cross Entropy Loss

Results

The model is trained for 1000 epochs and the training progress is visualized through a loss plot that is automatically generated and saved as 'loss_plot.png'. The model achieves approximately 99.27% accuracy on the test set, with results being reproducible using a fixed random seed (392).

Model Persistence

The trained model is saved to 'heart_disease_classifier_model.pth' for later use.

License

MIT License