Efficiently Detecting Light Events in Astronomical Temporal Data.

Detecting light events is a crucial problem in astronomical temporal data, the task of which is to find gravitational microlensing and flare star among the Milky Way. We propose an efficient method called FLMM (Fast Locating based on Median and Mean) to detect light events. The idea is to determine the location of suspicious light events in astronomical data by means of median-mean value and peak value in data after preprocessing. We combine the respective advantages of FLMM and Dynamic Time Warping (DTW) to improve the recognition accuracy. In order to balance between efficiency and accuracy, we use DTW to match the raw data of the suspicious light event. The experimental results demonstrate that for the processing of 0.93 million pieces of historical data, the combined method is nearly 3 times faster than DTW. The processing accuracy of abnormal data is improved by 2 orders of magnitude.