Machine Learning in Astrophysics: A Multi-Algorithm Comparison for Pulsar Prediction

Pulsar stars were discovered over 50 years ago. Known for their radiation pulses observed at regular intervals from planet Earth, they are a type of neutron star that possesses extremely powerful magnetic fields which are a result of the collapsed cores of the stars. Prediction of pulsar stars can be beneficial to scientists as they continue to uncover the mysteries relating to these exemplary celestial bodies. The purpose of this study is to compare the performance of various machine learning models to determine which one of them is the most capable of predicting pulsar stars. The models were built on the algorithms of Support Vector Machine (SVM), Random Forest and Naïve Bayes. The dataset used in this project is the widely known HTRU_2 dataset. The assessment of the models was done using evaluation metrics like the F1-score, recall, classification accuracy and precision. The metrics provided a comprehensive understanding of each model’s performance. The training and prediction times were also investigated to evaluate the scalability of the models. Of the models developed, Random Forest soared as the highest performer in terms of accuracy with 98.32%. Following closely behind was SVM with 97.56%, a result achieved when combined with the RBF kernel. The Naïve Bayes model, though it was faster than the other two models, attained an accuracy of 95.42%, as the classification performance was slightly on the lower side. The concluding study aims to encapsulate the performance of each model in terms of its strengths and weaknesses, thereby determining its potential suitability for pulsar prediction.