Heart Disease Classification Models

This repository demonstrates the implementation and evaluation of two machine learning models—Logistic Regression and Random Forest Classifier for predicting heart disease. It includes data preprocessing, exploratory data analysis (EDA), model training, testing, and statistical evaluation of model performance using permutation tests.

Features

• Data Preprocessing: Handling numerical and categorical features using scaling and one-hot encoding.
• Exploratory Data Analysis (EDA):
  • Scatterplot Matrix
  • Correlation Matrix Heatmap
  • Parallel Coordinates Plot
• Machine Learning Models:
  • Logistic Regression: A baseline linear model.
  • Random Forest Classifier: A non-linear ensemble model.
• Model Evaluation:
  • Accuracy scores for training and testing datasets.
  • Statistical significance test (Permutation Test) to compare model performance.
• Visualization:
  • Distribution of feature relationships.
  • Histogram of accuracy differences from the permutation test.

Results Summary

• Logistic Regression:
  • Training Accuracy: 87%
  • Testing Accuracy: 88%
• Random Forest:
  • Training Accuracy: 93%
  • Testing Accuracy: 90%
• Statistical Significance:
  • Observed Accuracy Difference: 0.0195
  • P-value: 0.7467
  • Conclusion: No statistically significant difference between the models.

Requirements

• Python 3.8+
• Required libraries (install via pip):
  • pandas
  • numpy
  • matplotlib
  • seaborn
  • scikit-learn

Installation

1. Clone the repository:

   git clone https://github.com/ravindusenavirathna/Heart-Disease-Classification-Models.git
   cd Heart-Disease-Classification-Models