Comparison of outliers and novelty detection to identify ionospheric TEC irregularities during geomagnetic storm and substorm
publication date Aug 2016 publication description Journal of Physics
Houw Liong Thee, Agus Pattikawa , Acep Purqon
Conference Series 739(1):012015 · August 2016 DOI: 10.1088/1742-6596/739/1/012015
In this study, we compare two learning mechanisms: outliers and novelty detection in order to detect ionospheric TEC disturbance by November 2004 geomagnetic storm and January 2005 substorm. The mechanisms are applied by using v-SVR learning algorithm which is a regression version of SVM. Our results show that both mechanisms are quiet accurate in learning TEC data. However, novelty detection is more accurate than outliers detection in extracting anomalies related to geomagnetic events. The detected anomalies by outliers detection are mostly related to trend of data, while novelty detection are associated to geomagnetic events. Novelty detection also shows evidence of LSTID during geomagnetic events.