Event Discovery in Time Series - Astrophysics > Instrumentation and Methods for AstrophysicsReport as inadecuate




Event Discovery in Time Series - Astrophysics > Instrumentation and Methods for Astrophysics - Download this document for free, or read online. Document in PDF available to download.

Abstract: The discovery of events in time series can have important implications, suchas identifying microlensing events in astronomical surveys, or changes in apatient-s electrocardiogram. Current methods for identifying events require asliding window of a fixed size, which is not ideal for all applications andcould overlook important events. In this work, we develop probability modelsfor calculating the significance of an arbitrary-sized sliding window and usethese probabilities to find areas of significance. Because a brute force searchof all sliding windows and all window sizes would be computationallyintractable, we introduce a method for quickly approximating the results. Weapply our method to over 100,000 astronomical time series from the MACHOsurvey, in which 56 different sections of the sky are considered, each with oneor more known events. Our method was able to recover 100% of these events inthe top 1% of the results, essentially pruning 99% of the data. Interestingly,our method was able to identify events that do not pass traditional eventdiscovery procedures.



Author: Dan Preston, Pavlos Protopapas, Carla Brodley

Source: https://arxiv.org/







Related documents