DATA-DRIVEN CHARACTERIZATION AND CLASSIFICATION OF EXOPLANETS FROM KEPLER SPACE TELESCOPE OBSERVATIONS

Abstract

Background: Space missions such as the Kepler Space Telescope which are available and offer large scale observational data have played a significant role in the discovery and analysis of exoplanets. Purpose: The proposed paper is an information-based approach to describing and categorizing exoplanets in terms of a properly structured data set of planetary and stellar characteristics. The parameters that are important in the dataset are orbital period, planetary radius, equilibrium temperature, stellar properties and observational signal characteristics, which offer a thorough basis to examine. 

Methodology: A systematic approach, which included data cleaning, exploratory data analysis, feature selection, and machine learning model, was used to identify exoplanet candidates as confirmed planets, candidates, and false positives. Several classification algorithms were tested, and ensemble-based models showed the best result as they can reflect the nonlinear relationships in the data.

Results: The findings show that the classification accuracy and the ability to distinguish between classes are high and they are justified by the detailed evaluation measures and the confusion matrix analysis.

Conclusion: In addition, feature importance analysis revealed planetary radius, orbital period and signal strength as the most significant variables in the classification and this aligns with well-known astrophysical knowledge. The results indicate that using machine learning-based methods and astrophysical data to enhance the characterization and classification of exoplanets is successful.