Proactive Health Insight System (PHIS)

PHIS is a fully local, privacy-preserving pipeline for wearable data analysis. It simulates health metrics (heart rate, steps, sleep, etc.), detects anomalies with ML (Z-score + Isolation Forest), generates insights, and visualizes via Streamlit dashboard—all without cloud [file:75].

🎯 Objectives

• Simulate wearable data from Kaggle datasets (Wearable Health Device Performance + Sports Health Monitoring).
• Flask + PostgreSQL backend for storage.
• Per-device baselines, anomaly detection, rule-based insights.
• Interactive dashboard with Plotly charts [file:75].

🛠️ Tech Stack

Layer	Technologies
Backend	Flask, PostgreSQL, psycopg2
ML	scikit-learn (Isolation Forest, Z-score)
Dashboard	Streamlit, Plotly
Data	Pandas, NumPy
Other	python-dotenv, requests [file:75]

🚀 Quick Start

1. Clone: git clone https://github.com/[yourusername]/phis_project.git
2. Env: conda create -n phis python=3.9 && conda activate phis
3. Deps: pip install -r requirements.txt
4. Setup DB (PostgreSQL): Update .env, create healthdata/insights tables.
5. Run:

streamlit run dashboard.py # Dashboard python app.py # Backend jupyter notebook # Simulator/ML notebook

Demo: Insights on anomalies like "High heart rate above baseline" [file:75].

📊 Key Results

• Processed 410k+ simulated records across 15 devices (e.g., Apple Watch, WHOOP).
• Generated 1,285+ insights (e.g., low sleep + high HR stress patterns).
• Local end-to-end latency: 2-3s [file:75].

Metric	Value
Records	410,380
Insights	1,285
Anomaly Rate	~1% [file:75]

🗂️ Structure

phis_project/ ├── README.md ├── .gitignore ├── requirements.txt ├── app.py # Flask backend ├── .env.example ├── notebooks/ # Simulator + ML ├── dashboard.py # Streamlit └── data/ # Processed Kaggle CSVs

📖 Background

Capstone for MSc Software Engineering, University of Europe for Applied Sciences. Addresses cloud/privacy issues in wearables with local ML insights. Author: Bhanu Teja Malineni (btm portfolio) [file:75].

🔮 Future Work

• Real wearable APIs (Fitbit/Garmin).
• LSTM/Transformers for time-series.
• Docker deployment [file:75].

📄 License

MIT. See your report appendices for full schema/code snippets [file:75].