Exoplanet Habitability Prediction Model Development: A Comparative Study

Exoplanets refer to planets existing beyond our solar system. Whether or not life exists on exoplanets is debatable. Studying and predicting habitability on exoplanets helps us learn how these particular systems are formed and how they have evolved. It provides clues to understand whether life exists elsewhere in the universe. This model aims to predict habitability using multiclass classification. After the data pre-processing, feature engineering was performed using three classifiers: Random Forest Classifier, Ada Boost Classifier, and Extra Trees Classifier. Features were selected, and the model was trained on three cases. The intersection of common features from three different classifier sets was obtained, the training of the model was done on features that were obtained from the classifier that gave maximum accuracy, and lastly, the features were taken from all the combined sets. This variation in the selection of features resulted in different confusion matrices, F1 scores, precision, and accuracy scores, which were taken as parameters for comparison. Lastly, the accuracy achieved with the proposed model is compared with K-Nearest Neighbors, Multiclass Decision Tree classification, and Gradient Boosting Classification. It achieved 72.9% accuracy with Decision Tree Classification, 98.2% accuracy with K-Nearest Neighbors Classification, and 96.4% accuracy with Gradient Boosting Classification. These findings highlight the influence of feature selection and classifier choice on prediction performance, offering insights for improving habitability prediction models.