Motaseam-Yousef/diabetes

Diabetes Prediction App

Overview

This Diabetes Prediction App is built using Streamlit to predict whether a person is likely to develop diabetes based on their health parameters. The app takes various inputs, such as pregnancies, glucose levels, blood pressure, BMI, and more, and uses a pre-trained machine learning model to provide a prediction.

Features

• User-friendly interface: Input fields are neatly organized into two columns for better readability.
• Interactive prediction: Users can input values and receive a prediction by clicking the "Predict" button.
• Custom styling: The app uses custom CSS for a clean and modern look.
• Diabetes Prediction Model: The app leverages a pre-trained ensemble model (loaded from a file) to make predictions based on input parameters.

Setup Instructions

Prerequisites

Before running the application, ensure you have the following:

• Python 3.x installed.
• Necessary dependencies installed from requirements.txt.

Folder Structure

Make sure your project has the following structure:

project_root/
│
├── model/
│   ├── scaler.pkl
│   └── ensemble_model.pkl
│
├── app.py
├── requirements.txt

• scaler.pkl: The scaler used for normalizing input data.
• ensemble_model.pkl: The trained ensemble model for diabetes prediction.
• app.py: The main script that runs the Streamlit app.
• requirements.txt: The file containing the necessary Python libraries.

Installing Dependencies

To install the required dependencies, run the following command in your project directory:

pip install -r requirements.txt

Running the App

Once dependencies are installed, you can run the app by executing the following command in the terminal:

streamlit run app.py

The app will open in your browser, and you can input health parameters to predict the likelihood of diabetes.

How the App Works

1. Loading Models: The app loads the pre-trained scaler and ensemble model from the model/ directory using Joblib.
2. User Input: The app collects the following user inputs:
   • Pregnancies
   • Glucose
   • Blood Pressure
   • BMI
   • Skin Thickness
   • Insulin
   • Diabetes Pedigree Function
   • Age
3. Scaling Input Data: The inputs are scaled using the loaded scaler.pkl model.
4. Prediction: The scaled data is passed to the loaded ensemble_model.pkl, which predicts whether the person is likely to have diabetes (binary output: 0 or 1).
5. Displaying Results: Based on the prediction, the app displays:
   • An error message if the prediction is positive (likely to have diabetes).
   • A success message if the prediction is negative (unlikely to have diabetes).