Potential Exoplanet Detection Using Feature Selection, Multilayer Perceptron, and Supervised Machine Learning

Since the discovery of the first exoplanet in 1992, advancements in technology have enabled the identification of numerous additional exoplanets. The recently launched James Webb telescope, succeeding the Hubble telescope, is set to enhance our understanding by scrutinizing exoplanet surroundings. Exoplanet detection, traditionally labor-intensive and reliant on experts, is now undergoing a transformation. Leveraging the wealth of data from the NASA Exoplanet Archive at Caltech, we employ techniques such as forward feature selection, Information Gain, and machine learning models like Logistic Regression and ensemble learning. Dimensionality reduction aids in selecting the most crucial features among the 49 available. The performance of these models is then rigorously evaluated through various metrics and visualizations, aiming to streamline the certification process for prospective exoplanets.