Detection of Quasi-Periodic Oscillations in Time Series of a Cataclysmic Variable Using Support Vector Machine

Abstract

Quasiperiodic oscillations (QPO's) in cataclysmic variables (CV) can be very subtle as well as their confidence of detection, indicating less significance than what the reality is. In our observed object, MV Lyrae, we focus on such QPO's. We simulated the QPO according to Timmer and Koenig (Astron. Astrophys. 300:707–710, 1995) and estimated its confidence intervals. Some known (not obvious) QPO's fell under 1- σ and therefore are not significant regarding this method. We propose and evaluate Support Vector Machine (SVM) models trained to identify those QPO's. Our main goal is to testify whether the accuracy of QPO detection using machine learning methods is higher than the significance of confidence levels obtained by the use of Timmer and Koenig’s (Astron. Astrophys. 300:707–710, 1995) simulations.