Organised by:
in collaboration with:



social media:


Accesso Utenti

Warsaw University Astronomical Observatory, Warsaw, Poland
Miscellaneous Information:

Abstract Reference: 30800
Identifier: O9.2
Presentation: Oral communication
Key Theme: 1 Reduction and Analysis Algorithms for Large Databases and Vice-versa

Machine Learning Variability Classification in the OGLE Project

Pawlak Michal

The OGLE project is one of the largest photometric variability surveys regularly monitoring about one billion sources in the densest sky regions. The huge amount of data collected gives a unique possibility to detect and study different types of variables but on the other hand it makes the automatisation of the classification process a must. Machine learning approach has been successfully applied to solve this problem, both in case of transient events like gravitational microlensing, as well as periodic variables, especially eclipsing binaries In each case a multi-step classification process based on Random Forest algorithm is proposed. The performance of the methods is evaluated resulting in high accuracy, in each case over 80%.