Early Heart Disease Prediction

Introduction

This project aims to predict the likelihood of early heart disease using machine learning techniques. The dataset used for this project is obtained from Kaggle and is locally named "dataset". It contains over 70,000 rows and 21 features. The target variable used for predicting is "HeartDiseaseorAttack".

Dataset URL: https://www.kaggle.com/datasets/alexteboul/diabetes-health-indicators-dataset

Data Preprocessing

Before building the machine learning models, the dataset underwent preprocessing steps including:

• Class Balancing
• Handling missing values
• Feature scaling

Machine Learning Models

Two machine learning algorithms were used for prediction:

1. Random Forest Classifier
2. Gradient Boosting Classifier

Model Evaluation

The models were evaluated using the following metrics:

• Accuracy
• Precision
• Recall
• F1 Score

Results

• Random Forest Model:

  • Accuracy: 91.67%
  • Precision: 0.92
  • Recall: 0.91
  • F1 Score: .915

• Gradient Boosting Model:

  • Accuracy: 90.52%
  • Precision: 0.91
  • Recall: 0.90
  • F1 Score: .905

Conclusion

Both Random Forest and Gradient Boosting models showed promising results in predicting early heart disease. Further optimization and fine-tuning of hyperparameters could potentially improve the model performance.

Future Work

Future work may include:

• Exploring additional machine learning algorithms
• Feature engineering to create new informative features
• Hyperparameter tuning for improved model performance