Analysis of Exoplanet Habitability Using RNN and Causal Learning

The quest to discover exoplanets with the potential for habitability is a captivating area of research that offers valuable insights into the origins of life in the universe. The proposed model investigates the application of causal learning and deep learning techniques to analyze the TESS, JWST, and KEPLER datasets. The primary objective is to classify exoplanets based on their habitability by uncovering significant features and relationships within the data. Using the transit method and other observations from these telescopes, the most essential elements such as planet gravity, eccentricity, mass, radius, and stellar temperature are extracted. The model is then trained in a deep learning architecture, specifically an RNN (Recurrent Neural Network), to recognize patterns in the dataset and predict which exoplanets are most likely to be habitable. The JWST dataset consists of approximately 8000 planetary observations, TESS comprises of around 6000 observations and the KEPLER dataset consists of around 9000 entries. Upon careful observations it is noticed that the LSTM model outperforms the RNN model in terms of accuracy and related scores.