Standardization framework of ionospheric Very Low Frequency (VLF) signal amplitude classes for machine learning-based anomaly detection: from calm ionospheric conditions to solar activity-induced dynamics

Machine learning (ML) techniques are extensively employed in the domain of near-Earth physics.An application of ML techniques is the anomaly detection of Very Low Frequency (VLF) ionospheric amplitude data.Prior research focused on the binary classification task, yielding promising results, and the subsequent exploration involves the multi-label classification of a broader spectrum of VLF amplitude signal features.This research paper introduces a standardization framework for labeling multi-class VLF amplitude features, including normal (daytime) signals, solar flare effects, nighttime signals, instrumental errors, and outlier data points.The primary aim of this standardization framework is to define all main VLF amplitude features, specify the conditions under which each VLF amplitude feature can be classified, and outline future initiatives for the development of additional tools to facilitate the labeling process.Future research will focus on developing supplementary tools and software packages for this purpose, with the ultimate objective of establishing a streamlined process from the Worldwide Archive of Low-Frequency Data and Observations (WALDO) database to labeled data and subsequently to ML models.