Abstract:
The light curve analysis of the heavenly bodies is an indispensable tool for understanding the physical phenomena that govern them. Doing so not only leads to new discoveries but also enhances our understanding of the universe. Large telescopes like the Large Synoptic Survey Telescope (LSST) will produce an excess of data that will necessitate the need for automated methods to sift through it quickly and efficiently, as doing so manually can be truly laborious. Furthermore, such a method should be able classify the observed astronomical objects accurately. Keeping this in view, this research presents an automated classification method using the simulated, photometric light curves provided in the Kaggle Challenge PLAsTiCC hosted by the LSST Team, in to 14 different classes. The classification model has been built around extracting several features and employing three different classifiers: Random Forest, eXtreme Gradient Boosting and Light GBM into an ensemble rounded off by a 5-layer Multilayer Perceptron (MLP). The training dataset containing 7848 samples has been used to train all three classifiers with different subsets of features sorted on the bases of their importance to the classifier. The MLP has then been trained on the concatenated probabilities of the three classifiers to predict the probabilities for 14 classes. The proposed methodology performs reasonably well for most of the classes achieving around an accuracy of 85% on the 3.5 million testing samples present in the test dataset. As the proposed methodology relies on features extracted from photometric light curves, therefore it can be adapted and extended for use in other fields that rely on similar light curves.
