Morphological Classification of Galaxies using Enhanced Deep Learning Model

Galaxies in astrophysics are one of its fundamental, their morphological classification remains an essential key to unlock countless doors about the structure, formation and history of our Universe. For nearly a century, astronomers have assigned galaxies to perceptual categories such as ellipticals, spirals and irregulars. However, the extraordinary expansion of the galaxy database catalysed and instigated by sky surveys of ever-increasing scales rendered this classification more manual than practical. Because of the learning capability and its scaling nature, to learn hierarchical abstract features from the visual data Convolutional Neural Networks (CNN) are powerful tools to recognize complex patterns and structures in images, such as spiral arms or central bulges that contribute significantly to accurate morphological classification. That is why CNN are particularly well suited to the problem of galaxy classification. While traditional classification work used to solve the problem, it is not feasible for today's huge amounts of generated data. This attempt to build and train a CNN model using galaxy images as our domain. The Enhanced CNN model can achieve the high accuracy to morphological type of galaxies. The Enhanced CNN increase both efficiency and classification accuracy, particularly when processing large dataset. The Galaxy Zoo dataset indicated that our study can provide a tool for other astronomical research in future. The Enhanced CNN gain the overall accuracy rate are 95.29% with the performance indicator were calculated including average precision, recall, F1-score and Area under curve (AUC) is 0.9513, 0.9531, 0.9525 and 0.9853 respectively. The Enhanced CNN outperforms the various existing state-of-the models.