Exoplanet Detection Using Kepler Mission Data with Machine Learning

Abstract

The search for habitable planets beyond our solar system has long captivated the scientific community and remains one of the foremost pursuits in modern astronomy. With the advent of space-based missions, such as NASA’s Kepler telescope, our observational capabilities have expanded significantly, resulting in vast volumes of high-quality astronomical data. This data deluge necessitates the development of scalable, automated methods to support astronomers in efficiently analyzing and interpreting these observations. In recent years, machine learning has emerged as a powerful paradigm for automating complex, human-intensive tasks. This study investigates the application of supervised machine learning techniques to the detection of exoplanets using data from NASA’s Kepler mission. The data set comprises Kepler Objects of Interest (KOIs), including both physical and orbital parameters, along with their confirmed classification. We evaluate a range of supervised classifiers, spanning probabilistic, decision tree-based, and neural network models. Our best-performing model, Histogram Gradient Boosting, achieves a precision of 94.6% and a recall of 94.1% on a held-out test set. These results underscore the promise of machine learning in advancing exoplanet detection and offer a pathway toward automating the discovery of planetary systems beyond our own.