sabique@portfolio:~$ cat projects/neuroguard.md

NeuroGuard

ML web app to predict stroke risk with 78.8% accuracy and 76.9% recall using SMOTE-enhanced Logistic Regression. Built with Flask API, Gemini 2.0, and a Vite-powered React frontend.

PythonFlaskscikit-learnReactGemini 2.0Vite
View Source CodeLive Demo

NeuroGuard

A machine learning web application that predicts stroke risk using advanced data preprocessing and machine learning techniques.

Overview

NeuroGuard is a comprehensive stroke risk prediction system that combines machine learning with an intuitive web interface. The application achieves impressive performance metrics with 78.8% accuracy and 76.9% recall, making it a reliable tool for preliminary stroke risk assessment.

Key Features

Machine Learning Pipeline

  • •SMOTE-Enhanced Dataset: Utilized Synthetic Minority Oversampling Technique to balance the dataset
  • •Logistic Regression Model: Implemented with careful feature engineering and hyperparameter tuning
  • •High Performance: Achieved 78.8% accuracy and 76.9% recall on test data
  • •Feature Engineering: Comprehensive preprocessing pipeline for optimal model performance

Backend Architecture

  • •Flask API: Robust REST API built with Flask
  • •Gemini 2.0 Integration: Advanced AI capabilities for enhanced predictions
  • •Data Validation: Comprehensive input validation and error handling
  • •Scalable Design: Modular architecture for easy maintenance and updates

Frontend Experience

  • •React + Vite: Modern, fast, and responsive user interface
  • •Interactive Forms: User-friendly input forms for medical data
  • •Real-time Predictions: Instant risk assessment with detailed explanations
  • •Responsive Design: Optimized for desktop and mobile devices

Performance Metrics

  • •Accuracy: 78.8%
  • •Recall: 76.9%
  • •Precision: 82.1%
  • •F1-Score: 79.4%
  • •AUC-ROC: 0.85

Technology Stack

Backend

  • •Python: Core programming language
  • •Flask: Web framework for API development
  • •scikit-learn: Machine learning library
  • •pandas: Data manipulation and analysis
  • •numpy: Numerical computing
  • •Gemini 2.0: AI integration for enhanced predictions

Frontend

  • •React: JavaScript library for building user interfaces
  • •Vite: Build tool for fast development
  • •Tailwind CSS: Utility-first CSS framework
  • •Chart.js: Data visualization library

Deployment

  • •Vercel: Frontend hosting and deployment
  • •GitHub Actions: CI/CD pipeline
  • •Docker: Containerization for consistent environments