Machine Learning-Based Analysis of Celestial Objects Using the SDSS17 Dataset

This study explores the application of machine learning (ML) techniques on the Sloan Digital Sky Survey (SDSS) Data Release 17 dataset, for the classification, regression and clustering of celestial objects, including stars, galaxies and quasars, using spectral features, redshift and position coordinates. Using Python and libraries such as scikit-learn, LightGBM and XGBoost, the research applies the classification algorithm (Logistic Regression, k-Nearest Neighbors, Random Forest), regression (Multiple Linear Regression, Polynomial Regression, LightGBM) and clustering (K-means, DBSCAN) on this data. The results indicate a maximum accuracy of 0.9796 in the classification with Random Forest, obtained by integrating all features, and a coefficient of determination of 0.4328 in the regression with k-NN, highlighting the combined contribution of the data. In clustering, K-means achieved a silhouette score of 0.6692 per redshift, reflecting the ability of distance-based separation. The study highlights the potential of SDSS17 as an "open science" resource for the practical application of artificial intelligence in astronomy. The conclusions confirm the effectiveness of nonlinear methods in cosmic data analysis and the essential role of open access to data in advancing research.