Heart Attack Prediction Model

Overview

This repository contains a machine learning project that predicts the likelihood of a heart attack based on a dataset of 170,501 rows and 25 features. The current model achieves an accuracy of 75%, with ongoing improvements through feature engineering and scaling.

Features

• Dataset Size: 170,501 rows and 25 columns.
• Model Accuracy: 75%.
• Techniques Used:
  • Feature Engineering: Enhancing feature selection and transformation.
  • Scaling: Standardizing feature values for better model performance.

Objectives

1. Improve the model's accuracy and robustness.
2. Optimize feature selection and scaling techniques.
3. Provide a user-friendly interface and detailed documentation.

Project Structure

HeartAttackPrediction/
│
├── data/                   # Dataset files
├── notebooks/              # Jupyter notebooks for data analysis and modeling
├── Heart_Attack_test1/     # Source code for data processing and model training
├── Heart_Attack_70/        # Saved trained models
├── README.md               # Project documentation
└── requirements.txt        # Python dependencies

Feature Engineering and Scaling

1. Current Focus:
   • Identifying redundant or irrelevant features.
   • Transforming features (e.g., normalization, log transformation).
   • Encoding categorical variables.
2. Scaling:
   • StandardScaler for numerical features.
   • RobustScaler to handle outliers.

Future Goals

• Improve accuracy to 90% or higher.
• Experiment with ensemble methods (e.g., Random Forest, Gradient Boosting).
• Deploy the model using a web app (e.g., Flask, FastAPI).
• Conduct hyperparameter tuning for further optimization.

Installation

1. Clone this repository:

   git clone https://github.com/yourusername/HeartAttackPrediction.git
   cd HeartAttackPrediction

2. Install dependencies:

   pip install -r requirements.txt

3. Run the model:

   python src/train_model.py

Usage

1. Add your dataset to the data/ directory.
2. Use the provided Jupyter notebooks for data analysis and feature engineering.
3. Train the model with the command above or customize the pipeline as needed.

Acknowledgments

• Dataset Source: [Include dataset link/source if applicable].
• Contributions: Feel free to contribute via pull requests or open issues for feedback and suggestions.

License

This project is licensed under the MIT License.