🩺 Diabetes Prediction App

A machine learning web application that predicts the likelihood of diabetes in a patient based on clinical diagnostic measurements. Built using a Random Forest Classifier trained on the PIMA Indian Diabetes Dataset from Kaggle, and deployed live via Streamlit.

---

🚀 Live Demo

🔗 Click here to try the app

---

🖼️ App Screenshot

---

📌 Project Overview

This project aims to assist in the early detection of diabetes by taking key medical parameters as input and predicting whether the patient is diabetic or non-diabetic. Two classification models were evaluated — Logistic Regression and Random Forest — with Random Forest selected as the final model due to its superior accuracy and recall on the diabetic class.

---

📊 Dataset

• Source: PIMA Indian Diabetes Dataset – Kaggle
• Records: 768 patient entries
• Features: 8 clinical input features + 1 target variable
• Target: Outcome — 1 (Diabetic), 0 (Non-Diabetic)

Feature	Description
Pregnancies	Number of times pregnant
Glucose	Plasma glucose concentration (2-hour oral glucose tolerance test)
BloodPressure	Diastolic blood pressure (mm Hg)
SkinThickness	Triceps skinfold thickness (mm)
Insulin	2-hour serum insulin (µU/ml)
BMI	Body Mass Index (weight in kg / height in m²)
DiabetesPedigreeFunction	Diabetes pedigree function (genetic risk score)
Age	Age in years

---

🧠 Model Details

Model	Accuracy	Notes
Logistic Regression	71.4%	Lower recall on diabetic class
Random Forest	77.3%	Selected — better F1 and recall

Final Model: RandomForestClassifier(class_weight='balanced', random_state=42)

• class_weight='balanced' used to handle class imbalance in the dataset
• Features scaled using StandardScaler before training
• Model and scaler saved together as diabetes_model_final.pkl using joblib

Random Forest — Classification Report (Test Set):

              precision    recall    f1-score   support
           0       0.80      0.87      0.83       100
           1       0.71      0.59      0.65        54
    accuracy                           0.77       154

---

🛠️ Tech Stack

Layer	Technology
Language	Python 3.x
ML Libraries	scikit-learn, NumPy, pandas
Visualization	Matplotlib, Seaborn
Model Serialization	joblib
Web App	Streamlit
Deployment	Streamlit Cloud

---

📁 Project Structure

diabetes-prediction-app/
│
├── app.py                        # Streamlit web application
├── diabetes_model_final.pkl      # Trained Random Forest model + scaler
├── diabetes.csv                  # PIMA Indian Diabetes dataset
├── Diabetes_detection.ipynb      # EDA, training, and model evaluation notebook
├── requirements.txt              # Python dependencies
└── README.md                     # Project documentation

---

⚙️ Run Locally

1. Clone the repository

   git clone https://github.com/<your-username>/diabetes-prediction-app.git
   cd diabetes-prediction-app

2. Install dependencies

   pip install -r requirements.txt

3. Run the Streamlit app

   streamlit run app.py

4. Open your browser at http://localhost:8501

---

📈 Workflow

Raw Data (CSV)
     │
     ▼
Exploratory Data Analysis (EDA)
     │
     ▼
Data Preprocessing (StandardScaler)
     │
     ▼
Model Training (Logistic Regression + Random Forest)
     │
     ▼
Model Evaluation & Selection (Random Forest — 77.3%)
     │
     ▼
Model Serialization (joblib → .pkl)
     │
     ▼
Streamlit Web App → Deployed on Streamlit Cloud

---

⚠️ Disclaimer

This application is built for educational and demonstration purposes only. It is not a substitute for professional medical advice, diagnosis, or treatment. Always consult a qualified healthcare provider for medical decisions.

---

👤 Author

Navneet

• 📧 [navneetnitin.ece87@gmail.com]
• 💼 [https://www.linkedin.com/in/navneet-nitin/]
• 🐙 [https://github.com/navneet-nitin]

---

📄 License

This project is licensed under the MIT License.