Exoplanet Habitability Explorer

◉ An end-to-end data-science & ML project that explores, ranks, and visualizes thousands of confirmed exoplanets using data from the NASA Exoplanet Archive.

◉ The project demonstrates data sourcing, feature engineering, interactive visualization, and ML classification in a single deployable web app.

---

◉ Features

• Live NASA data pull – fetches the latest PSCompPars catalog from the NASA Exoplanet Archive via its TAP API (src/fetch_data.py)

• Explainable habitability score – composite of physical parameters such as insolation, radius, distance, stellar temperature, etc.

• Interactive web app (Streamlit) —
  • Filter by radius, insolation, discovery year, distance, etc.
  • Weight presets (Conservative HZ / Optimistic HZ / Observation-friendly) plus sliders for custom scoring
  • Click-to-inspect planet details with score-component breakdown
  • Compare tray for side-by-side comparison of up to 3 planets
  • Downloadable filtered table

• ML classifier – trains a Random-Forest model to predict “optimistic habitable-zone candidate” label from non-leaking astrophysical & engineered features
  (e.g. luminosity proxy, semi-major axis, estimated insolation)
  • ROC-AUC / PR-AUC / F1 displayed in the app
  • Optional toggle to show predicted probability & label in the UI

• Clean architecture & reproducibility – separate training script, model artifacts in /models, Streamlit app in app.py

---

◉ Tech Stack

• Python: pandas, numpy, requests, pyarrow
• Data science / ML: scikit-learn, joblib
• Web app / viz: Streamlit, Plotly, streamlit-plotly-events
• Data source: NASA Exoplanet Archive TAP API

---

◉ Workflow

# 1. Create environment & install deps
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# 2. Pull latest data
python src/fetch_data.py

# 3. (optional) Train / update ML model
python src/train_classifier.py

# 4. Run interactive app locally
streamlit run app.py