In June 2026 kaggle playground series, we have a great dataset for astronomy enthusiasts and I am one of them.
Access the competition from here: Competition Page.
In this episode we have the following 10 features:
| S.No. | Feature | Description |
|---|---|---|
| 1 | alpha | Right Ascension, RA. Equivalent to longitude on Earth. Measured in degrees (0°-360°). Specifies east-west poisiton on celestial sphere. |
| 2 | delta | Declination, Dec. Equivalent to latitude on Earth. Measured in degrees (-90° to +90°). Specifies north-south position. |
| 3 | u | Ultravoilet |
| 4 | g | Green |
| 5 | r | Red |
| 6 | i | Near Infrared |
| 7 | z | Infrared |
| 8 | redshift | Tells us how far a celestial object is |
| 9 | spectral_type | It is the classification based on temperature from hottest to coolest |
| 10 | galaxy_population | It tells what kind of object do we have - a blue cloud or red sequence |
alpha and delta are sky coordinates. They tell where the object is located in the sky.
u, g, r, i, z are SDSS photometric bands. Each measures brightness through a different filter. Every celestial object emit light differently.
Due to expansion of the universe, nearby objects have small redshift while distant galaxies have larger.
Redshift, z =
spectral_type is the stellar spectral class, that classifies objects based on temperature. It is written as O, B, A, F, G, K, M from hottest to coolest, repectively.
In our dataset, galaxy_population has only two unique values - Blue_Cloud and Red_Sequence. It generally tells us if the object is newly born or old.
The dataset have 577k rows, so we have a quite a big dataset. Using these feature, we need to predict what kind of celestial object it is from the given target features.
Target: Galaxy | QSO | Star
| Notebook | Description |
|---|---|
| Catboost | Benchmark and newer versions of CatBoost |
| LightGBM | Benchmark and newer versions of LightGBM |
| XG Boost | Benchmark and newer versions of xg-boost |
| MLP | Benchmark and newer versions of MLP |