Parkinson Disease Prediction using Machine Learning

This project implements a machine learning model for the prediction of Parkinson's disease. It uses ensemble learning methods and various feature selection techniques to predict Parkinson's disease based on a set of relevant features. The application includes a web interface built with React and Tailwind for users to input data, receive predictions, and view the prediction history.

Table of Contents

1. Project Overview
2. Methodology
3. Model Training and Evaluation
4. Web Interface
5. Technologies Used
6. Installation and Usage
7. Contributing
8. License

Project Overview

This project focuses on predicting Parkinson’s disease using machine learning. The model is trained using a dataset that includes features such as UPDRS, Tremor, Functional Assessment, MoCA, and Rigidity. The primary goal is to develop an accurate model to help in early detection of Parkinson’s disease. A web-based interface is integrated for real-time predictions.

Methodology

1. Data Collection & Exploratory Data Analysis (EDA)

• Dataset with Parkinson’s-related features is gathered.
• EDA is performed to understand data distribution and anomalies.

2. Data Pre-Processing

• Handling missing values through imputation.
• Removing irrelevant features for efficiency.
• Detecting and handling outliers using IQR.
• Normalizing numerical data using StandardScaler.
• Encoding categorical variables.

3. Feature Selection

• Univariate Feature Selection
• Mutual Information
• Lasso (L1 Regularization)
• Correlation-based Feature Selection

4. Train-Test Data Split

• Data split into training and testing sets (80-20, 70-30, 60-40).

5. Model Training using Ensemble Learning Methods

• Random Forest, Gradient Boosting, AdaBoost, Stacking, Bagging, Voting.

6. Model Evaluation

• Accuracy, Precision, Recall, F1-score.
• Confusion Matrix, AUC-ROC Curve.

7. Best Model Selection & Deployment

• AdaBoost model achieved 86.67% accuracy and 91.59% AUC-ROC Score.
• Deployed through a web interface.

Model Training and Evaluation

The top five features found to have the highest correlation with Parkinson’s disease:

• UPDRS (Unified Parkinson’s Disease Rating Scale)
• Tremor
• Functional Assessment
• MoCA (Montreal Cognitive Assessment)
• Rigidity

Web Interface

Features

1. Sign Up / Sign In Pages: User authentication.
2. Home Page:
   • Input values for the top five features.
   • Displays prediction results with probability scores.
   • View past predictions.
3. History Page:
   • Shows previous predictions with timestamps.

Technologies Used

• Machine Learning: Python, Scikit-learn, Pandas, NumPy, Matplotlib, Seaborn.
• Backend: Flask (ML model integration), Node.js, Express.js, JWT, CORS.
• Frontend: React.js, Tailwind CSS.
• Database: MongoDB.

Installation and Usage

Prerequisites

• Python 3.x
• Node.js & npm
• MongoDB (local/cloud)
• pip

Clone the Repository

git clone https://github.com/yourusername/parkinson-disease-prediction.git
cd parkinson-disease-prediction

Backend Setup (Flask API)

cd backend/model-src
pip install 
python app.py

Backend Setup (Node.js & Express)

cd backend
npm install
npm run dev

Frontend Setup

cd frontend
npm install
npm run dev

Now, access the application at http://localhost:5173.