Browsing by Author "Uvarova, Elizaveta"

Asteroids within our Solar System attract considerable attention for their potential impact on Earth and their role in elucidating the Solar System's formation and evolution. Understanding asteroids' composition is crucial for determining their origin and history, making spectral classification a cornerstone of asteroid categorization. Spectral classes, determined by asteroids' reflectance spectrum, offer insights into their surface composition. Early attempts at classification, predating 1973, utilized photometric observations in ultraviolet and visible wavelengths. The Chapman-McCord-Johnson classification system of 1973 marked the beginning of formal asteroid taxonomy, employing reflectance spectrum slopes for classification. Subsequent developments included machine learning techniques, such as principal component analysis and artificial neural networks, for improved classification accuracy. Gaia mission's Data Release 3 has significantly expanded asteroid datasets, allowing more extensive analyses. In this study, I examine the relationship between asteroid photometric slopes, spectra, and taxonomy using a feed-forward neural network trained on known spectral types to classify asteroids of unknown types. Our classification gained the mean accuracy of 80.4 ± 2.0 % over 100 iterations and separated successfully three asteroid taxonomic groups (C, S, and X) and the asteroid class D.