A Fuzzy Shape-Based Anomaly Detection and Its Application to Electromagnetic Data

The problem of data analytics in real-world electromagnetic (EM) applications poses a lot of algorithmic constraints. The process of big datasets, the requirement of prior knowledge, unknown location of anomalies, and variable length patterns are all issues that need to be addressed. In this application, we address those issues by proposing a fuzzy shape-based method with anomaly detection. This method is evaluated against 12 benchmark datasets of different kinds of anomalies and provides promising results based on the use of a new performance metric that takes into account the distance between the predicted and actual anomalies. Real-world EM data from the Earth's magnetic field are provided by the SWARM satellite constellation relating to regions in China, Greece and Peru. The seismic events that occurred in those regions are compared against the SWARM data. Moreover, three other methods: GrammarViz, HOT-SAX, and CUSUM-EWMA are also applied to further investigate the possible linkages of EM anomalies with seismic events. The findings further our understanding of real-world data analytics in EM data and seismicity. Some proposals regarding the limitations of available data for the real-world datasets are also presented.