Full metadata record
DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | 조인휘 | - |
dc.contributor.author | 장지원 | - |
dc.date.accessioned | 2023-09-27T02:12:13Z | - |
dc.date.available | 2023-09-27T02:12:13Z | - |
dc.date.issued | 2023. 8 | - |
dc.identifier.uri | http://hanyang.dcollection.net/common/orgView/200000684167 | en_US |
dc.identifier.uri | https://repository.hanyang.ac.kr/handle/20.500.11754/187334 | - |
dc.description.abstract | Time series data is ubiquitous in our daily lives and plays a pivotal role in a wide range of fields, including industrial inspection, finance, network security, healthcare, and environmental monitoring. The ability to detect anomalies in time series data is of paramount importance in these domains as it enables us to identify critical events, outliers, or deviations from expected patterns. With the rapid advancements in deep learning, there has been an increasing emphasis on leveraging these powerful techniques to tackle the challenge of anomaly detection in time series data. In this paper, we study into the application of a recently emerged generative model, the diffusion model, for detecting anomalies in time series. The diffusion model offers a promising approach by capturing the important dynamics and dependencies within the data through a sequence of transformations. It allows us to model the underlying distribution of the time series and generate high-quality samples that adhere to the observed patterns. To operationalize this approach, we specifically utilize CSDI as the foundational model for anomaly detection. Our proposed methodology involves comparing the generated predicted distribution obtained from CSDI with the actual distribution of the observed time series data. By evaluating the gaps between these distributions, we can effectively identify and flag potential outliers or anomalous patterns. This approach leverages the strengths of deep learning and the capability of the diffusion model to capture the complex nature of time series data. To validate the performance and efficacy of our proposed approach, we conduct extensive experimental evaluations on multiple real-world datasets. These datasets encompass various scenarios, encompassing different types of anomalies, and provide a robust evaluation framework for our method. In summary, this paper contributes to the field of anomaly detection in time series data by exploring the application of the diffusion model, specifically CSDI, as a powerful tool for this task. Through rigorous experimentation and evaluation, we presented the efficacy and superiority of our proposed approach, providing valuable insights for researchers and practitioners in anomaly detection and deep learning. | - |
dc.publisher | 한양대학교 | - |
dc.title | Anomaly Detection of Time Series Data Based on Diffusion Model | - |
dc.type | Theses | - |
dc.contributor.googleauthor | 장지원 | - |
dc.sector.campus | S | - |
dc.sector.daehak | 대학원 | - |
dc.sector.department | 컴퓨터·소프트웨어학과 | - |
dc.description.degree | Master | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.