“School of Astronomy”
Back to Papers HomeBack to Papers of School of Astronomy
| Paper IPM / Astronomy / 18366 |
|
||||||||||||||
| Abstract: | |||||||||||||||
|
Dusty stellar point sources contribute to stellar dust emissions and enrichment of galaxies such as the Magellanic Clouds (MCs). These objects can be resolved and classified using photometric and spectroscopic observation. We applied supervised machine learning classifiers for dusty stellar point source classification in the MCs, including young stellar objects (YSOs) and evolved stars. We implemented typical algorithms to classify dusty stellar categories comprising YSOs, oxygen- and carbon-rich asymptotic giant branch stars (AGBs), red supergiants (RSGs), and post-AGB stars. In the following, we created synthetic samples based on the minority class using the Synthetic Minority Oversampling Technique (SMOTE) method to resolve the Surveying the Agents of Galaxy Evolution (SAGE) catalogs' imbalanced and limited spectral counts. The classification accuracy of these sources, which include 619 spectral labeled counts, was assessed using infrared features and multiwavelength filters. According to the results, the Probabilistic Random Forest (PRF) classifier, a tuned Random Forest (RF), performed best among all the models applied before and after data augmentation with the SMOTE for categorizing dusty point sources. We reached an accuracy of ~89 percent based on the recall metric for the spectral classification of dusty stellar sources. In this study, using the SMOTE technique does not improve the accuracy of the best model for the CAGB, PAGB, and RSG classes; it stays at 100% and 88%, respectively. However, there are variations for the CAGB, OAGB, and YSO classes. We collected photometric labeled data with properties similar to the training dataset and classified them using the top four models with more than 87% accuracy. The labeled data is available for three dusty stellar types, YSOs, oxygen- and carbon-rich AGBs, due to the reliable accuracy we achieved for them in models.
Download TeX format |
|||||||||||||||
| back to top | |||||||||||||||
