Using Machine Learning to Classify Stars, Quasars, and Galaxies

In this project, we looked at data from stars, quasars, and galaxies from the sixteenth data release of the Sloan Digital Sky Survey Telescope. The aim of the project was to accurately and quickly classify these three types of objects using machine learning. We used three machine learning algorithms, namely logistic regression, multi-layer perceptron, and decision tree classifier.

In this project, we looked at data from stars, quasars, and galaxies from the sixteenth data release of the Sloan Digital Sky Survey Telescope. The aim of the project was to accurately and quickly classify these three types of objects using machine learning. We used three machine learning algorithms, namely logistic regression, multi-layer perceptron, and decision tree classifier. The classification was done based on measurements of the object’s redshift and its u,g,r,i, and z light emissions. Logistic regression offered the least accurate results, with an overall accuracy of 94.83%, and a runtime of 28.2 seconds. Better in both aspects was the decision tree classifier, with an overall accuracy of 98.90%, and a runtime of 6.36 seconds. Several structures with different kinds of neuron-layer arrangements were tried for the MLP classifier. While all yielded around a 98% total accuracy, the more complex a structure was, or the more nodes and layers it had, the more time it took to run, and none even came close to the low amount of time taken by the decision tree classifier.

Explore More!

Research Paper