← Back to Projects Deployed System

MediScan – Health Risk Guidance System

A machine learning-based healthcare system that predicts diseases from user symptoms and provides actionable guidance including severity analysis, precautions, and nearby hospital recommendations.

Python Scikit-learn Streamlit Machine Learning Geolocation

Problem And Solution

Problem

Most symptom-based systems provide limited or vague predictions without actionable guidance, leaving users uncertain about next steps or severity of their condition.

Solution

Built a system that predicts diseases and enhances the output with severity levels, precautions, and real-world recommendations to help users make informed decisions.

Key Features

Disease Prediction

  • Multi-class classification (41 diseases)
  • Top 3 predictions with confidence levels

Health Guidance

  • Severity scoring (1–5 scale)
  • Precautions and do’s/don’ts
  • Home remedies and suggestions

Smart Recommendations

  • Actionable insights based on severity
  • Suggestions like rest, medication, or emergency care

Location Intelligence

  • Detects user location using JavaScript
  • Recommends top nearby hospitals
  • Interactive map integration

Output System

  • Email-based report delivery
  • Complete health summary report

Model And Data

Model

Built a multi-class classification model trained on symptom-disease relationships to predict the most likely conditions.

Dataset

Synthetic healthcare dataset (IBM), enabling fast prototyping and controlled experimentation.

Results And Observations

Achieved high accuracy (~99%) due to synthetic dataset. Real-world performance may vary, and future work includes training on real clinical datasets for improved generalization.

Future Improvements

  • Integration with real medical datasets
  • Improved explainability of predictions
  • Enhanced recommendation engine