Comparative study of CV methods for outlier detection in time series using deep learning models†
Suntae Park1 · Changryong Baek2
12Department of Statistics, Sungkyunkwan University, Seoul, Korea.
Correspondence to: † This work was supported by the Basic Science Research Program from the National Research Foundation of Korea (NRF-2022R1F1A1066209).
1 Graduate student, Department of Statistics, Sungkyunkwan University, Seoul 03063, Korea.
2 Corresponding author: Professor, Department of Statistics, Sungkyunkwan University, Seoul 03063, Korea. E-mail: crbaek@skku.edu
1 Graduate student, Department of Statistics, Sungkyunkwan University, Seoul 03063, Korea.
2 Corresponding author: Professor, Department of Statistics, Sungkyunkwan University, Seoul 03063, Korea. E-mail: crbaek@skku.edu
Received September 15, 2024; Revised October 7, 2024; Accepted October 30, 2024.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
- Abstract
- In this paper, we compared and analyzed the performance of various cross-validation methods to identify optimal hyperparameters in deep learning models for time series outlier detection. We evaluated the performance and training time of LSTM-AE, LSTM-VAE, and USAD models. These models were applied to datasets with different outlier patterns, including trend, seasonality, and shapelet. The cross-validation methods used were the so-called KF, TS, and RW. We analyzed both synthetic and real-world datasets. The results show that the KF method generally outperformed the others based on F1 scores. However, the performance differences among the three methods were minimal for datasets with short periods and repetitive patterns. The RW validation method consistently demonstrated the fastest training time. These findings suggest that, while RW may not provide the best performance, it can serve as an efficient alternative in scenarios requiring rapid model updates. This is particularly useful in real-time outlier detection, where speed is critical without significantly compromising performance.
- Keywords : Cross-validation, deep learning, LSTM-AE, LSTM-VAE, outlier detection, times-series, USAD
Full Text(PDF) Free
