Thông tin chung
  English
  Đề tài NC khoa học
  Bài báo, báo cáo khoa học
  Hướng dẫn Sau đại học
  Sách và giáo trình
  Các học phần và môn giảng dạy
  Giải thưởng khoa học, Phát minh, sáng chế
  Khen thưởng
  Thông tin khác
  Tài liệu tham khảo
  Hiệu chỉnh
 
Số người truy cập: 107,402,381

 A General Model for Long-short Term Anomaly Generation in Sensory Data
Tác giả hoặc Nhóm tác giả: Thien-Binh Dang, Duc-Tai Le, Moonseong Kim, Hyunseung Choo
Nơi đăng: IEEE International Conference on Ubiquitous Information Management and Communication (IMCOM); Số: 10.1109/IMCOM53663.2022.9721783;Từ->đến trang: 1-5;Năm: 2022
Lĩnh vực: Công nghệ thông tin; Loại: Báo cáo; Thể loại: Quốc tế
TÓM TẮT
Anomaly detection algorithms play an important role in Internet of Things (IoT) where a significant amount of data is processed every second. The abnormal data can seriously affect the decision-making of data analysts that may lead to system failure. Hence, anomaly detection algorithms are useful tool to identify anomaly. However, detection accuracy of these algorithms is affected by the amount and quality of training data. In fact, the well-known-published datasets are limited. Moreover, they are not labeled and are hard to use for training. In this paper, we propose a general model for artificial anomaly generation. The proposed model can generate six typical forms of anomalies in IoT time-series data including stuck-at, offset, drift, noise, outlier, and spike. The model allows users not only to straightforwardly generate anomalies under various parameters but also generate combined anomalies which are the combination of those six typical forms of anomalies.
ABSTRACT
Anomaly detection algorithms play an important role in Internet of Things (IoT) where a significant amount of data is processed every second. The abnormal data can seriously affect the decision-making of data analysts that may lead to system failure. Hence, anomaly detection algorithms are useful tool to identify anomaly. However, detection accuracy of these algorithms is affected by the amount and quality of training data. In fact, the well-known-published datasets are limited. Moreover, they are not labeled and are hard to use for training. In this paper, we propose a general model for artificial anomaly generation. The proposed model can generate six typical forms of anomalies in IoT time-series data including stuck-at, offset, drift, noise, outlier, and spike. The model allows users not only to straightforwardly generate anomalies under various parameters but also generate combined anomalies which are the combination of those six typical forms of anomalies.
© Đại học Đà Nẵng
 
 
Địa chỉ: 41 Lê Duẩn Thành phố Đà Nẵng
Điện thoại: (84) 0236 3822 041 ; Email: dhdn@ac.udn.vn