Skip to content

Latest commit

 

History

History
49 lines (31 loc) · 1.53 KB

File metadata and controls

49 lines (31 loc) · 1.53 KB

Predicting Stellar Class | Kaggle Competition

Multi-class classification predicting stellar object type (Galaxy, Star, QSO) from astronomical measurements.

Competition Result

  • Competition: Kaggle Playground Series Season 6 Episode 6
  • Public Score: 0.96517 (Balanced Accuracy)
  • Leaderboard Rank: 1203 / 2,666 teams (Top 45%)
  • Submissions: 1

Approach

  • Feature engineering: color indices (u-g, g-r, r-i, i-z, u-z) from photometric filters
  • Model: LightGBM with class_weight='balanced'
  • Metric: Balanced Accuracy (handles class imbalance across 3 classes)

Dataset

  • 577,347 training rows, 247,435 test rows
  • 3 classes: GALAXY (65.4%), QSO (20.3%), STAR (14.3%)
  • Key feature: redshift — almost perfectly separates QSOs from stars and galaxies

Key Features

Feature Description
redshift Most powerful feature — separates QSO from GALAXY and STAR
u_g, g_r, r_i, i_z, u_z Color indices — brightness differences between wavelength filters
spectral_type Star spectrum classification
galaxy_population Galaxy population type

Project Structure

stellarclass_model/

├── config/

├── processing/

├── trained_models/

├── pipeline.py

train_pipeline.py

predict.py

research.ipynb