Diabetes Prediction System

🚀 Project Overview

📊 Dataset

• Source: Pima Indians Diabetes Dataset
• Attributes: Pregnancies, Glucose, BloodPressure, SkinThickness, Insulin, BMI, DiabetesPedigreeFunction, Age
• Target: Outcome (0: No Diabetes, 1: Diabetes)

🛠️ Data Pipeline

1. Cleaning: Handled invalid zero values in features like Glucose and Insulin by replacing them with column medians.
2. EDA: Analyzed distributions, correlations, and class balance.
3. Preprocessing: Standard scaling and class imbalance handling using SMOTE.
4. Modeling: Compared Logistic Regression, KNN, Random Forest, SVM, and XGBoost.
5. Optimization: Hyperparameter tuning of Random Forest using GridSearchCV to maximize Recall.

📈 Model Performance

• Accuracy: > 80% (Expected)
• Recall: > 75% (Expected)
• ROC-AUC: > 0.85 (Expected)

💻 How to Run Locally

1. Clone the repository

2. Install dependencies

3. Run the analysis (Optional)

4. Run the Flask Web App

📂 Project Structure

Diabetes-Prediction/
│
├── data/
│   └── diabetes.csv
│
├── notebooks/
│   └── diabetes_analysis.ipynb
│
├── models/
│   ├── model.pkl
│   └── scaler.pkl
│
├── app.py
├── requirements.txt
├── README.md

🔮 Future Improvements

• Integrate Deep Learning models (ANNs).
• Add more patient demographics for better accuracy.
• Deploy the Streamlit app to Streamlit Cloud or Heroku.
• Implement real-time monitoring of model drift.