Ensemble Learning for Probabilistic Classification of Astronomical Objects in Deep Fields Using Artificial Intelligence and Machine Learning Method

Understanding cosmic structure and evolution requires the classification of stars, galaxies, and quasars; however, supervised learning techniques are still limited by the lack of labeled spectroscopic data, and the large amount of data from contemporary surveys makes manual classification unfeasible. Massive amounts of image data encompassing dim and densely populated celestial objects are produced by deep astronomy surveys. It is difficult to classify these artifacts accurately, especially in deep-field areas without authorized catalog labeling. This paper presents an Artificial Intelligence (AI) and Machine Learning (ML)–driven framework for probabilistic classification of astronomical objects using coordinate-based catalog labeling combined with ensemble learning techniques. Objects detected in cataloged sky regions are cross-matched using celestial coordinates to obtain reliable ground-truth labels from established astronomical databases 6. These labeled samples are used to train a One-vs-Rest ensemble of Machine Learning classifiers, including Random Forest, Gradient Boosting, Support Vector Machine, in addition to the Logistic Regression models. Physics-based detection using a signal-to-noise ratio criterion is integrated with feature extraction to ensure scientific consistency prior to ML inference 5. The trained ensemble is applied to unlabeled deep fields to generate probabilistic predictions while explicitly accounting for uncertainty. Experimental validation through injection–recovery tests demonstrates increasing detection completeness with higher signal-to-noise ratios. The proposed framework PCAO method provides a scalable, scientifically consistent, besides of the uncertainty-aware approach for astronomical object classification in modern deep-field surveys. The primary contribution of this paper is the integration of physics-based detection, coordinate-derived ground truth, and ensemble Machine Learning within a probabilistic classification framework. By explicitly modeling uncertainty and incorporating scientific constraints prior to inference, the PCAO proposed methodology ensures both computational robustness and astrophysical consistency.