Change point analysis as a tool to detect abrupt cosmic ray muons variations

Abstract

Recently, there have been an increasing number of studies using Big Data. They rely on large data sets of time series to detect artificial or natural patterns in processes of natural sciences and economy. The most possible outcome due to lack of rigid data processing is data contamination with abrupt drifts and regime shifts. They yield either inclusion of undetected errors or missed detection of important observations and events. Possible automatic tools for detection of regime shifts could be delivered from change point statistical methods. However, a major drawback for the most of the currently available change point (CP) methods is the challenge of complex temporal variations in non-stationary natural processes like cosmic rays observed at Earth. This kind of data analysis is applied to experimentally acquired time series from cosmic ray measurements. The observed parameters are muons produced in cosmic ray cascades in atmosphere and acquired in parallel with atmospheric and other meta-data. In this study, we test different approaches for change point detection in compound particle counting process.