Abstract |
In this project, our goal is to improve the existing methods for identifying the compact objects of X-ray
Binary systems and classifying the accretion state of their compact object.
X-ray Binary systems contain a stellar remnant, which we call a compact object and a companion star.
The characterization of these systems can be done either based on the compact object or based on the
companion star; in our research, we deal with the first type of characterization. The traditional
classification method is based on the visual examination the X-ray spectrum of each individual system.
This process is very time consuming and mistakes can easily be made.
In the past decade, there has been a lot of research on the development of machine learning algorithms
to automate this process and achieve results that are more accurate in much less time.
In our effort to develop accurate classification models, we used the Random Forest Classifier and
Artificial Neural Networks on the NuSTAR simulated datasets based on RXTE observations of X-ray Binary
systems with Neutron Stars, Black Holes and Pulsars. We characterized these systems based on the
compact object and the accretion state in which it resides.
Our results show that both the Random Forest and the Artificial Neural Network perform extremely well
with an average accuracy, across all datasets, of 97.31% for the Random Forest and 96.63% for the
Artificial Neural Network. Comparing our models with another machine learning method called SVM
that has been used on the same data and achieved an average accuracy of 92.9%, we can see that both
the Random Forest and the Artificial Neural Network classifiers perform better and we propose them as
better alternatives to the SVMs. Although the difference between the two models are very low, we can
see that the Random Forest classifier seems to perform slightly better on characterizing a system
according to the accretion state, while the Artificial Neural Network seems to perform slightly better on
characterizing according to the compact object.
|