An advanced healthcare monitoring solution that uses Machine Learning and Deep Learning to predict patient risk levels in real-time. Unlike traditional threshold-based systems, this project analyzes multivariate sensor data to provide early warnings for cardiac events.
Our system operates on a multi-layered validation logic:
- XGBoost Classifier: Real-time risk status (Normal vs. Critical) with 100% Accuracy.
- LSTM (Deep Learning): Analyzes temporal patterns to predict future cardiac risk probability.
- Isolation Forest: Monitors sensor integrity to detect malfunctions and prevent false alarms.
- Balanced Data: Achieved perfect Mean-Median alignment using 2.5% strict percentile capping.
- Precision: 0.99 for critical patient detection.
- Features: Age, HR, SpO2, Systolic/Diastolic BP, Temp, HRV, MAP, and Pulse Pressure.
- Languages: Python
- AI/ML: XGBoost, TensorFlow (Keras), Scikit-learn
- Data: Pandas, Numpy
- Backend/Web: Flask/FastAPI (Work in Progress by Varun)
- IoT: Arduino/ESP32 (Work in Progress by Vishal & Suhani)
- Shivam Gupta: Data Cleaning Pipeline & XGBoost Risk Model.
- Sanya Gupta: LSTM Predictive Analytics & Anomaly Detection.
- Varun Gupta: Web Dashboard & API Integration.
- Vishal & Suhani Gupta: IoT Sensor Hardware & Cloud Data Flow.