Stellar Classification based on Various Star Characteristics using Machine Learning Algorithms

Abstract

The task of stellar classification can be tedious and lengthy when done manually. One can expedite stellar classification by creating an artificial intelligence model to automate the process. As we as a species continue to explore the frontier of the observable universe, we should seek to automate time-intensive problems like stellar classification. The current stellar classification model serves to effectively categorize stars for research purposes regarding their distribution around the universe, so automating the development of this resource would allow professionals to allocate more time to explore the bounds of our current understanding of space and the universe. After finding and analyzing a dataset containing numerical and categorical features, a supervised learning approach was then used to train and test different models on their ability to classify the stars. A Decision Tree Classifier, Random Forest Classifier, Ridge Classifier, and Support Vector Classifier were trained and tested. The most successful models were the Decision Tree Classifier and Random Forest Classifier, each with about a 94 percent prediction accuracy across different accuracy metrics on the test data. Despite some drawbacks in regard to the availability of usable data, four models were trained and two were proven to be consistently and successfully accurate. Any future attempts at developing models for stellar classification should concentrate more on gathering data as to have a more thoroughly trained set of models.