AutoML Assumption Checker 🔍

An interactive tool for checking and visualizing the assumptions of statistical learning models — built for learning, teaching, and robust modeling.

Feature	Status
Linearity Check	✅ Done
Homoscedasticity	✅ Done
Normality Check	✅ Done
Multicollinearity	🛠️ In Progress
Independence Check	⬜ Planned
Outlier Detection	⬜ Planned

🎯 Purpose

Most AutoML tools skip statistical assumptions. This project flips the script:

• Step-by-step assumption checking
• Visual diagnostics and interpretation helpers
• Built for students, bloggers, and practitioners

🧠 What It Does

✅ Upload your dataset or simulate one ✅ Walk through common linear model assumptions:

• Linearity
• Homoskedasticity
• Normality
• Multicollinearity
• Outliers & Influential Points

✅ Visual + statistical tests side-by-side ✅ Model summaries with plain-English diagnostics

🧰 Tech Stack

• Python 3.11+
• Streamlit (or FastAPI/Gradio TBD)
• Pandas / Statsmodels / Scikit-learn
• Matplotlib / Seaborn / Plotly

🚀 Getting Started

# Clone the repo
git clone https://github.com/W-Thurston/automl_assumption_checker.git
cd automl_assumption_checker

# (Optional) Create a virtual environment
python -m venv .venv
source .venv/bin/activate  # or .venv\Scripts\activate on Windows

# Install dependencies
pip install -r requirements.txt

# Run the app (placeholder — full UI coming soon)
python app/main.py

# For now: Generate report on simulated data
python app/report.py

# Output (console):
# ✅ Assumption: Linearity
#    R² = 0.86 → Pass
#    Recommendation: —

🧪 Tests

pytest tests/

🔴 Live Demo

Launch the interactive version (coming soon)

📦 Docker (coming soon)

Want full reproducibility? A containerized version will be available shortly.

📜 License

MIT License © 2025 Will Thurston