Building Accurate Machine Learning Models for Predicting the Habitability of Exo-Planets

For the last 20 years, due to the rapid advancement in space technology, mainly advanced telescopes, astronomers, and astrobiologists have been able to find the number of planets outside the solar system, and there is a greatzeal in the scientific community regarding the knowledge of the worlds and possibility of their habitability. NASA has done commendable work in this field; over time, they have found much information regarding extra-terrestrial objects.According to some studies, there are around 700 quintillion celestial objects in the Universe. Fortunately, with technological advancement, humans can find 5000 plus exo-planets. To find the habitability of the exo-planets depending on various factors such as the planet’s size, the temperature on the planet, its position in the habitable zone, the distance of it from the start and other parameters. Some work has been done previously in this field, and we tried to get more accurate and suitable results for the available data. Building a model that could better predict the planet’s habitability and help astrobiologists get better results in further experiments on the exoplanets. The data set is taken from the PHL exoplanet Habitability website. For this we used Logistic Regression (LR), which provided S9% accuracy, SVM which provided 98% accuracy, K-fold cross-validation, which provided 94% accuracy, XGBoost which provided 99.6% accuracy; and Stratified K-fold cross-validation, which delivered 99.95% accuracy. Hence, after checking the above models, itis found that Stratified K-fold is the best model for habitability prediction.