Citation

Mohd Zebaral Hoque, Jesmeen and Hossen, Md. Jakir and Abd. Aziz, Azlan (2021) Unsupervised Novelty Detection for Time Series using Deep Learning. In: 2nd FET PG Engineering Colloquium Proceedings 2021, 1-15 Dec. 2021, Online Conference. (Unpublished)

Text 07 Jesmeen.pdf Restricted to Repository staff only Download (19kB)

Abstract

The function uses unlabeled extracted clean features to train the learning model using the normal value of the time series. The forecasting time series model was trained using deep convolutional neural network (CNN) and long short-term memory (LSTM). The model was stored for the real-time anomaly detection process. In the real application, the trained model was loaded. This process takes in a range of windows of series data and predicts the next value of the window. Following with a few steps of algorithm which finds out the predicted value is how close to the actual value. According to the anomaly fraction percentage and threshold (a distance value) the data is distinguished between normal and abnormal data.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Anomaly detection, deep learning, feature extraction, time series, novelty
Subjects:	Q Science > Q Science (General) > Q300-390 Cybernetics
Divisions:	Faculty of Engineering and Technology (FET)
Depositing User:	Ms Nurul Iqtiani Ahmad
Date Deposited:	25 Jan 2022 14:00
Last Modified:	06 Mar 2023 06:47
URII:	http://shdl.mmu.edu.my/id/eprint/9878

Downloads

Edit (login required)