TY - GEN

T1 - Analysis of the LSTM-AE OCSVM Method for Finding Anomalies in Multivariate Time Series

AU - Klekchyan, Pavel

AU - Mayatskaya, Ekaterina

AU - Chernyshov, Yury

PY - 2023/5/15

Y1 - 2023/5/15

N2 - The problem of anomaly detection is one of the most pressing tasks of data mining and is actively studied. Anomaly detection is applied in areas such as medical diagnostics, network traffic analysis, text and image processing and recognition, as security and information protection, industrial monitoring, risk management. The obvious approach to solving the anomaly detection problem is as follows: we need to define a domain corresponding to normal behavior. Then any observation lying in this area will be considered normal, and outside the area - abnormal. In this paper, we will consider an anomaly search method based on deep learning without a teacher, where no marked-up data is used. A machine learning algorithm will be paired with deep learning to classify the data. Threshold will also be used in the comparison of classification results. The method will be examined on a dataset of a simulated gasoil heating loop process, which is a multivariate time series.

AB - The problem of anomaly detection is one of the most pressing tasks of data mining and is actively studied. Anomaly detection is applied in areas such as medical diagnostics, network traffic analysis, text and image processing and recognition, as security and information protection, industrial monitoring, risk management. The obvious approach to solving the anomaly detection problem is as follows: we need to define a domain corresponding to normal behavior. Then any observation lying in this area will be considered normal, and outside the area - abnormal. In this paper, we will consider an anomaly search method based on deep learning without a teacher, where no marked-up data is used. A machine learning algorithm will be paired with deep learning to classify the data. Threshold will also be used in the comparison of classification results. The method will be examined on a dataset of a simulated gasoil heating loop process, which is a multivariate time series.

UR - http://www.scopus.com/inward/record.url?partnerID=8YFLogxK&scp=85164926888

U2 - 10.1109/USBEREIT58508.2023.10158820

DO - 10.1109/USBEREIT58508.2023.10158820

M3 - Conference contribution

SP - 332

EP - 335

BT - Proceedings - 2023 IEEE Ural-Siberian Conference on Biomedical Engineering, Radioelectronics and Information Technology, USBEREIT 2023

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2023 IEEE Ural-Siberian Conference on Biomedical Engineering, Radioelectronics and Information Technology (USBEREIT)

Y2 - 15 May 2023 through 17 May 2023

ER -