IN SEARCH OF TABBY-AI’S SISTER (STAR KIC 8462852)

Abstract

Tabby’s Star (KIC 8462852) -a star that undergoes chaotic and irregular brightness changes- exhibits dimming events of approximately 20%. This means that the measured light from the star drops by up to 20% at certain times relative to its average observed brightness. The reduction in Tabby’s Star’s brightness is non-periodic, which is extremely unusual. These fluctuations -identified through the Planet Hunters project led by Tabetha Boyajian using Kepler telescope data- do not correspond to spherical bodies and have given rise to various hypotheses. This observation leads to the following research question: How can potential planets, structures, or other celestial bodies be efficiently located and categorized using data discarded by missions such as Kepler and TESS? To address this, TABBY-AI was developed -an interactive web application built using Python 3.11 and Streamlit that integrates: light curve visualization; comparison of similar curves; anomaly detection using LSTM (Long Short-Term Memory) neural networks; and a reference for key astronomical concepts. The platform is bilingual (Spanish/English), features a NASA-inspired design, and is updated daily. Its primary objective is to facilitate the analysis of unusual stellar brightness patterns, promoting the identification of cases comparable to Tabby’s Star through artificial intelligence and data science techniques.

In this study, the dependent variable is the variation in stellar brightness (light curves), while the independent variables correspond to astrophysical and observational factors (observation time, space mission, orbital parameters, potential transiting bodies, dust clouds, or other structures). The results demonstrate that TABBY-AI holds great potential for democratizing access to astronomical analysis. Its implementation facilitates the detection of unusual stellar phenomena, aiding in the discovery of as-yet unexplained structures and contributing to the advancement of astrophysical research through artificial intelligence.