S-Phot Stellar Classifier

Description

An XGBoost based approach to classification of highly sparse, sampling biased photometric stellar data with extreme class imbalance using Data generated by PySSED.

File Descriptions

1. data/
   1. random_10perc PySSED generated labels and data for ten percent SIMBAD sample
   2. ms_augmented PySSED generateed labels and data for ten percent SIMBAD sample augmented with additional main sequence star data
2. tuning_results_5foldcv_2000_iter_ms_augmented/ Contains outputs generated by running tuning_script.py with data from data/ms_augmented
3. tuning_results_5foldcv_2000_iter_no_ms/ Contains outputs generated by running tuning_script.py with data from data/random_10perc
4. tuning_script.py Script used to tune XGBoost model hyperparameters
5. main.py Script used to load and train XGBoost models using parameters saved from running tuning_script.py 6 create_imbalance_plots.py Used in sparsification experiments
6. explore_commons.py Contains helper functions
7. XGBoost_Weighted.py XGBoost model with additional parameters for tuning of class weights.
8. requirements.txt libraries and versions used in order to generate experiment results.