An Algorithm for Early Detection of Monitoring Data Deviations Arising Out of Natural or Man-Made Emergencies

An algorithm for early detection of extraneous changes in multidimensional environmental monitoring data resulting from natural or man-made emergencies has been developed based on the combined use of self-organizing Kohonen maps (SOM) and the k-nearest neighbors method. The main purpose of the algorithm is to increase the interpretability of changes in environmental monitoring data that signal the occurrence of an emergency situation. The algorithm determines the spontaneous data changes in two stages in order to minimize the number of false positives: first, potential deviations from typical values are determined at the level of individual nodes on the SOM map, then at the level of the entire sample. A freelance observation is defined as deviating at both stages. Verification of the proposed algorithm was carried out on a set of environmental monitoring data. The algorithm revealed observations corresponding to documented emergency discharges and emergencies, which confirms its practical significance. The algorithm is reproducible, scalable, allows working with data of various nature and can be integrated into environmental monitoring systems for data analysis. The presented implementation demonstrates the effectiveness of the approach using the example of an environmental monitoring task, which confirms its applicability to solving practical problems.