Experience of using autoencoders in solving problems of detecting anomalies in time series of measurement information

E.N. Tsyba1, O.A. Volkova2, N.A. Vostrukhov1

1 FSUE “VNIIFTRI”, Mendeleevo, Moscow region, Russia;
2 LLC Mosinzhiniring Grupp, Moscow, Russia;
tsyba@vniiftri.ru

Al’manac of Modern Metrology № 2 (38) 2024, pages 150–160

Abstract. Detecting and handling outlier values in the dataset is a critical issue in machine learning. The work presents analyzed and systematized information on methods for searching for anomalies in time series of measurement information. The issue of the potential use of autoencoders in the context of detecting anomalies in measurement information is considered using the example of time series of GNSS station coordinates (MDVJ and IRKJ). It is concluded that autoencoders allow detecting anomalies in time series to a fairly complete extent.

Keywords: autoencoders, anomalies, measurement information, neural network modeling.

Full texts of articles are available only in Russian in printed issues of the magazine.

Previous article ……. Contents ……. Next article