Exploring Synergistic Approaches to Enhance Predictive Modeling in Astronomical Classification

This research article focused on testing the performance of three machine learning classification algorithms- RandomForest, XGBoost, and CatBoost for classifying stellar multiple classes allows for an evaluation of their performance metrics-accuracy, precision, recall, and F1-score. For all the cross-validation cases, RandomForest demonstrated constant dominance over the other algorithms, which achieved the highest accuracy of 98.29% and the best F1 scores in all classes. The ensemble approach it used to aggregate multiple decision trees considerably reduced overfitting and was capable of addressing complex feature interactions inherent in datasets of astronomy with minimal hyperparameter tuning. The recall of RandomForest was strongly concentrated on Class 1, at 99.97% which again highlighted its strength in the correct classification of positive instances. Although XGBoost and CatBoost's reported competitive performance, the strength of the system, good tradeoff between precision and recall, and interpretability provided by RandomForest are more proficient for stellar classification. Future work can be oriented towards hybrid models and feature engineering to improve classification accuracy.