Analysis of the Schmidt Camera Archive: Variable Object Search and Classification Using Machine Learning

Digitized archival astronomical data serve as a valuable resource for studying variable stars and their evolution. This study analyzes variable stars based on archival photometric data obtained with the Schmidt camera at the Fesenkov Astrophysical Institute from 1960 to 1989.Cross-identification with the General Catalog of Variable Stars catalog revealed 20,961 variable stars in the archival frames. Additionally, matching with Gaia DR3 and TESS expanded the characterization of these stars, including galactic height, variability amplitude, spatial distribution, and identification of objects with archival observations. These parameters had not been previously compiled in a single catalog.To further enrich the data, machine learning methods were applied, allowing the classification of objects with incomplete characteristic data and the prediction of their spectral types and variability classes. The analysis demonstrated that XGBoost achieved the highest accuracy among the tested models (78.9% for variability types and 86.46% for spectral types) and identified the most significant parameters influencing stellar variability.The final catalog has been published on the Kazakhstani National Virtual Observatory platform and is available for further research.