Advanced Predictive Analytics for Higher Education Retention

A cloud-native platform engineered to transform raw academic data into actionable retention strategies. Powered by Google BigQuery and Streamlit, this system provides a real-time dashboard for university administrators to monitor, predict, and prevent student dropout.

Access Student Intelligence Hub

• Real-Time KPIs: Total enrollment, risk cases, satisfaction pulse
• Priority Intervention Queue: Students requiring immediate attention
• Strategic Insights: Satisfaction drivers and cluster trends

• Risk Tiers: Critical (>75%), Monitor (35-75%), Safe (<35%)
• Actionable Lists: Export at-risk cohorts for outreach
• Model Disclaimer: Built-in warning about prediction polarization

• Behavioral Clustering: K-Means segmentation into 4 archetypes
• Silent Burnout Detection: High grades + low satisfaction alerts
• Feature Importance: Model explainability

• Multilingual Support: Real-time interaction in English, Italian, Spanish, and French
• Priority Context Mapping: Logic refined to prioritize analytical advice over greetings
• Page-Aware Responses: Answers adapt to current view (Dashboard, Console, 360)

studenti-analytics/
├── streamlit_app.py    # Main application (Optimized callbacks)
├── ml_utils.py         # Polyglot AI (EN, IT, ES, FR) & ML logic
├── data_utils.py       # Optimized BigQuery data loading
├── constants.py        # Configuration and table metadata
├── styles_config.py    # LinkedIn-style CSS theme
├── comprehensive_test.py # Automated test suite
├── requirements.txt    # Python dependencies
├── SQL_QUERIES.md      # BigQuery ML queries
└── README.md           # Documentation

• Python 3.9+
• Google Cloud Service Account (JSON Key)

# Clone
git clone https://github.com/Giacomod2001/studenti-analytics.git
cd studenti-analytics

# Install dependencies
pip install -r requirements.txt

# Run Tests
python comprehensive_test.py

# Run App
streamlit run streamlit_app.py

Note: Model predictions may show polarized distributions. See SQL_QUERIES.md for implementation details.

• Alessandro Geli
• Giacomo Dellacqua
• Paolo Junior Del Giudice
• Ruben Scoletta
• Luca Tallarico

Apache License 2.0 - See LICENSE