You can access paper here
Author: Ahmed Bera Pay
Course: CMPE 49F
Professor: Fatih Alagoz
Date: January 2024
This research paper explores the application of machine learning techniques to classify star types, focusing on key stellar attributes such as temperature, luminosity, and spectral class. Conducted as part of the CMPE 49F course, the study investigates the use of various algorithms—Random Forest, k-Nearest Neighbors, Support Vector Classifier (SVC), XGBoost, and Logistic Regression— to predict star types. Random Forest demonstrated the highest accuracy in this classification task.
As fundamental celestial bodies, stars have intrigued astronomers for centuries towards new dimensions in understanding stellar evolution and cosmic dynamics, which also serve as cornerstones in understanding the universe’s evolution. In this context, various techniques and methods have been used to classify stars based on features such as temperature, luminosity, and magnitude. In the contemporary era, machine learning emerges as a powerful and comprehensive concept, used in astronomy to process astronomical datasets and train models for predictive analysis. This paper investigates the relationships, correlations, and significance of different parameters, including luminosity, temperature, and spectral characteristics, in classifying different star types. This research employs some of the machine learning algorithms, namely Random Forest, kNN, SVC, XGB, and Logistic Regression, to construct prediction models to classify stars such as Red Dwarf, Main Sequence, Super Giants, etc. The Random Forest algorithm exhibited the highest accuracy in correctly predicting the star types.
- Algorithms Used: Random Forest, k-Nearest Neighbors, Support Vector Classifier, XGBoost, Logistic Regression
- Core Attributes Analyzed: Temperature, Luminosity, Radius, Absolute Magnitude, Spectral Class
- Results: Highest accuracy achieved by Random Forest, confirming strong correlations across stellar features