🩺 Diabetes Prediction: Research & Deployment System

📌 Overview

This repository serves a dual purpose, bridging the gap between academic research replication and practical software deployment:

1. Research Phase (.ipynb): A strict replication of a specific medical study on diabetes prediction using Random Forest, SMOTE, and Global Scaling.
2. Deployment Phase (.py): A modular web application built with Streamlit that serves the trained model to end-users in a user-friendly interface.

⚠️ NOTE: This is the Academic Replication version which intentionally contains data leakage to reproduce a paper's results. For the industry-standard, robust version (Leakage-Free & Fine-Tuned), please visit: 👉 Diabetes Prediction Fine-Tuned Project

---

⚠️ Academic Methodology Disclaimer

Please Read Before Reviewing the Research Notebook:

To strictly adhere to the cited reference paper's methodology and reproduce their reported metrics, the Jupyter Notebook (ml-prediction-diabetic-code.ipynb) follows a specific preprocessing workflow:

• Global Scaling: MinMaxScaler is applied to the entire dataset before splitting.
• Global SMOTE: Oversampling is applied to the entire dataset before splitting.

Methodology Note: I am fully aware that applying these techniques before the Train-Test split introduces Data Leakage and is not standard industry practice. However, this was done intentionally to reproduce the exact results reported in the academic paper.

---

📂 Project Architecture

This project moves beyond a simple notebook by implementing a Modular Architecture for deployment. The logic is separated into distinct responsibilities:

├── app/                                   # 💻 APPLICATION SOURCE CODE
│   ├── app.py                             # Main Streamlit application
│   ├── model.py                           # Backend logic & inference
│   └── preprocess.py                      # Utils for input formatting
├── assets/                                # 🖼️ STATIC ASSETS
│   └── diabetes_app_ui.png                # App Screenshot
├── models/                                # 📦 ARTIFACTS (Serialized Objects)
│   ├── scaler.joblib                      # Saved MinMaxScaler
│   └── model_rf.joblib                    # Saved Random Forest Model
├── notebooks/                             # 🔬 RESEARCH & EXPERIMENTATION
│   └── ml-prediction-diabetic-code.ipynb  # Replicated Research Study
└── requirements.txt                       # Dependency list

🛠️ Tech Stack

• Frontend Framework: Streamlit
• Machine Learning Core: Scikit-Learn (Random Forest Classifier)
• Data Handling: Pandas, NumPy, Joblib
• Imbalanced Data: Imbalanced-learn (SMOTE)
• Environment: Python 3.9+

🚀 How to Run Locally

1. Clone the Repository

git clone https://github.com/viochris/Diabetes-prediction-project.git
cd Diabetes-prediction-project

2. Install Dependencies

Make sure you have the required libraries installed:

pip install -r requirements.txt

3. Run the Streamlit App

Execute the main application file from the root directory:

streamlit run app/app.py

4. Access the Dashboard

Open your browser and navigate to: http://localhost:8501

📊 Research Results

Based on the replication study conducted in the notebook (ml-prediction-diabetic-code.ipynb) using Random Forest with Global SMOTE:

• Algorithm: Random Forest Classifier
• Accuracy: ~83.67%
• Precision: ~84%
• Recall: ~84%

(Metrics are inflated due to the intentional data leakage required for paper replication)

📷 Screenshots

Web Application Interface

User-friendly interface built with Streamlit allowing for real-time patient data input and instant prediction:

---

Author: Silvio Christian, Joe "Bridging the gap between Academic Research and Practical Deployment."