As a prominent binary star system in the constellation Virgo, Spica (α-Virginis) offers valuable insights into stellar interiors and dynamics. In binary systems, gravitational interactions between the stars cause subtle deformations that affect their orbital paths, and the steady rate of this change is referred to as the apsidal constant. This constant provides critical information about a star’s internal structure and evolutionary stage. Traditionally, stellar environments are studied through simulations like MESA (Modules for Experiments in Stellar Astrophysics), but solving for the apsidal constant through such simulations is computationally intensive and time-consuming. This research seeks to integrate machine learning techniques to efficiently solve for the apsidal constant, with two key objectives: (1) implement a variety of machine learning and deep learning models to constrain the apsidal constant, and (2) apply feature engineering techniques to identify the most significant variables influencing its determination.
Both machine learning and deep learning approaches demonstrate significant potential in estimating Spica’s apsidal constant. Through evaluations ranging from simple machine learning models to more advanced architectures like transformers, the relationships within the underlying dataset were uncovered, revealing the key features influencing the determination of the apsidal constant. The most impactful features identified include luminosity, age, and effective temperature, as determined through both model-based and model-agnostic methods. In summary, integrating machine learning not only improves the efficiency of estimating Spica’s apsidal constant but also offers a novel approach that could be applied to other areas of stellar astrophysics.