Back to research
Machine Learning · Group project · 2023

Exoplanet Habitability Classification & Analysis

A full machine-learning pipeline over the PHL Exoplanet Catalog (5,600+ planets, 100+ features) to predict habitability class, with explanations for individual predictions.

View on GitHub
[ ARC ]

Architecture

  1. 01

    Catalog cleaning

    • Drop error / ID columns
    • Missing-value audit

  2. 02

    Feature engineering

    • KNN & iterative imputation
    • Scaling, encoding
    • SMOTE for imbalance

  3. 03

    Model zoo

    • LogReg, RF, SVM, KNN, MLP
    • XGBoost, LightGBM, CatBoost
    • RandomizedSearchCV

  4. 04

    Explainability

    • LIME
    • Permutation importance
    • Partial dependence
[ KEY ]

Technical highlights

  1. 01

    Compared eight classifiers under stratified cross-validation, with imbalance handled by SMOTE inside the training folds.

  2. 02

    Explained individual predictions with LIME alongside global permutation importance.

[ STK ]

Stack

  • Python
  • scikit-learn
  • XGBoost
  • LightGBM
  • CatBoost
  • LIME
  • imbalanced-learn
  • Dask