🫀 Smart Heart Disease Detection System

A machine learning project that predicts the likelihood of heart disease in a patient based on clinical features. Multiple classification algorithms are trained, evaluated, and compared to find the best-performing model.

---

📌 Problem Statement

Given a set of medical features, predict whether a person has heart disease:

• 1 → Person has heart disease
• 0 → Person does not have heart disease

---

📂 Dataset

• Source: Kaggle — Heart Disease UCI
• File: heart.csv
• Features used for modelling: cp, trestbps, chol, fbs, restecg, thalach, exang

---

🛠️ Tech Stack

Category	Tools
Language	Python
Data Analysis	Pandas, NumPy
Visualization	Matplotlib, Seaborn
Machine Learning	scikit-learn
Model Export	Joblib
Notebook	Jupyter Notebook

---

📊 Exploratory Data Analysis (EDA)

The following visualizations were used to understand the data:

• Count Plot — class distribution of target variable
• Density Plot — feature distributions
• Histogram — spread of each feature
• Box Plot — outlier detection
• Correlation Heatmap — feature relationships
• KDE Plot — kernel density estimation
• Pair Plot — multi-feature relationships
• Distance Plot — age and trestbps distributions

---

⚙️ Feature Engineering

• Identified categorical vs continuous features based on unique value count
• Applied One-Hot Encoding on categorical columns using pd.get_dummies()
• Selected 7 key features for model training

---

🤖 Models Trained

Model	Accuracy
Logistic Regression	83.61%
Naive Bayes	80.33%
Ensemble (LR + NB)	81.97%
Random Forest	75.41%
Support Vector Machine (SVC)	68.85%
Decision Tree	67.21%
K-Nearest Neighbors	65.57%

✅ Best Model: Logistic Regression — 83.61% accuracy

---

📈 Model Evaluation

• Train/Test Split: 80% / 20% (random_state=42)
• Evaluation Metric: Accuracy Score
• Confusion matrices plotted for all 7 models
• Bar chart comparison of all model accuracies
• Best K value for KNN identified as 6, 14, or 18 via score plot

---

💾 Model Export

The best model (Logistic Regression) is exported using joblib:

import joblib
joblib.dump(logreg, "hdp_model.pkl")

---

🚀 How to Run

1. Clone the repository

git clone https://github.com/hema123-4/smart-heart-disease-detection.git
cd smart-heart-disease-detection

2. Install dependencies

pip install pandas numpy matplotlib seaborn scikit-learn joblib

3. Download the dataset from Kaggle and place heart.csv in the project folder

4. Run the notebook

jupyter notebook Heart_Disease_Prediction.ipynb

---

📁 Project Structure

smart-heart-disease-detection/
│
├── Heart_Disease_Prediction.ipynb   # Main notebook
├── hdp_model.pkl                    # Exported Logistic Regression model
├── heart.csv                        # Dataset (download from Kaggle)
└── README.md                        # Project documentation

---

👩‍💻 Author

Hemalatha Muchumarri
AI Engineer | B.Tech CSE – AI & ML, Alliance University