Kepler Exoplanet Search Result

An end-to-end machine learning pipeline for exoplanet classification is presented in this project. It performs feature selection, preprocesses and balances the features of the Kepler dataset, and trains several models, such as KNN, Random Forest, SVM, and XGBoost. Metrics like accuracy, precision, recall, F1/F2 scores, and ROC-AUC are used to assess model performance. SHAP and LIME are used to further interpret the optimal model in terms of explainability and feature importance. Potential exoplanets can be accurately detected and interpreted from astronomical data thanks to this pipeline.