| Title |
Anomaly Detection of SE-ImDiffusion Network using Time Series Data |
| Authors |
김동현(Dong Hyun Kim) ; 황호성(Ho Seong Hwang) ; 김호철(Ho Chul Kim) |
| DOI |
https://doi.org/10.5573/ieie.2024.61.7.69 |
| Keywords |
Time series anomaly detection; Factory facility; Imputation; Diffusion model |
| Abstract |
In this paper, we propose an anomaly detection network that is robust to the frequently encountered issue of anomaly concentration in time-series data analysis and demonstrates high accuracy. To achieve this, we employed an imputation-based technique resistant to anomaly concentration and incorporated the squeeze-and-excitation block, an attention mechanism, into the network structure, resulting in the development of a novel SE-ImDiffusion network. We conducted experiments using the Server Machine Dataset and Mars Science Laboratory Dataset, which are widely used for the performance comparison of anomaly detection in time-series data. We evaluated the models using Precision, Recall, and F1 Score, and observed superior performance across all metrics. In conclusion, we have confirmed that utilizing imputation-based techniques and attention mechanisms yields superior performance compared to other anomaly detection networks that utilize time-series data. |