Application of Machine Learning to Interpret Predictability of Different Models: Approach to Classification for SDSS Sources

Abstract

In astronomy, the classification of galaxies has a variety of uses. Many studies of the universe have benefited from extensive collections of categorized galaxy pictures. However, Sloan Digital Sky Survey (SDSS), which is the most critical dataset source in astronomy, does not categorize 50 million galaxy images it contains. Machine learning algorithms are essential for automating this procedure since there are too many things to classify manually. Artificial Intelligence (AI) and Machine learning algorithms can match human performance in transforming data from the SDSS survey into physical objects, artifacts, etc. Humans still do this phase in any temporary scientific workflow, but future surveys, the Large Synoptic Survey Telescope, will ask for machine-enabled alternatives. This project aims to correctly classify SDSS data and use principle component analysis to find the best feature from the dataset. Though the decision tree shows a maximum of 97.1692 % accuracy, all other algorithms also show good accuracy. The second dataset from feature selection shows 83.0729 % accuracy(Decision Tree), which is the maximum from the other two formats of PCA data.