EduPulse Platform

Flask-based student early-warning prototype for training academic-risk classifiers, reviewing predictions, and comparing predictions with later outcomes.

Live demo: https://edupulse-platform.onrender.com/

What It Implements

• CSV validation and upload flows for training, prediction, and actual-result datasets
• Reproducible comparison of Logistic Regression, Decision Tree, Random Forest, and SVM classifiers
• Persisted best-model artifact and evaluation metrics
• Student-level risk bands and deterministic support recommendations
• Prediction-versus-actual comparison
• Render deployment blueprint, tests, and GitHub Actions CI

Evidence Boundary

The included data is synthetic demonstration data. Model metrics show the behavior of this controlled prototype and should not be interpreted as evidence of effectiveness in a real educational setting.

Recommendations are deterministic rules based on input indicators and model risk. The application does not autonomously contact students, change academic records, or replace qualified human review.

Architecture

flowchart LR
    A[Synthetic training CSV] --> B[Validate and preprocess]
    B --> C[Compare classifiers]
    C --> D[Persist selected model and metrics]
    E[Prediction CSV] --> F[Risk scores and support recommendations]
    D --> F
    G[Actual outcomes CSV] --> H[Prediction comparison]
    F --> H

Loading

Security Controls

• Flask secret is supplied through FLASK_SECRET_KEY in deployment
• Uploads are restricted to CSV filenames and a 5 MB request limit
• The repository excludes local virtual environments and secrets

This is a public demonstration application without authentication. Do not upload confidential or personal student data.

Run Locally

python -m venv .venv
python -m pip install -r requirements.txt
python app.py

Verify

python -m pip install pytest
pytest -q
python -m compileall app.py utils tests