Cover Image

Impact of Resampling and Deep Learning to Detect Anomaly in Imbalance Time-Series Data

Deep neural networks; Time series; Anomaly detection; Capture time; Data imbalance; Electricity theft detection; Imbalance time series data; Over sampling; Resampling; Resampling technique; Time-series data; Under-sampling; Anomaly detection

Saved in:
Bibliographic Details
Main Authors: Saripuddin M., Suliman A., Sameon S.S.
Other Authors: 57220806580
Format: Conference Paper
Published: Institute of Electrical and Electronics Engineers Inc. 2023
Tags: Add Tag
No Tags, Be the first to tag this record!
id my.uniten.dspace-27223
record_format dspace
spelling my.uniten.dspace-272232023-05-29T17:41:13Z Impact of Resampling and Deep Learning to Detect Anomaly in Imbalance Time-Series Data Saripuddin M. Suliman A. Sameon S.S. 57220806580 25825739000 36683226000 Deep neural networks; Time series; Anomaly detection; Capture time; Data imbalance; Electricity theft detection; Imbalance time series data; Over sampling; Resampling; Resampling technique; Time-series data; Under-sampling; Anomaly detection In the domain of anomaly detection, it is common that the data presented has lower amount of anomaly cases that cause data imbalance. Resampling technique has been one of the fastest and reliable way to overcome data imbalance but its effectiveness on time series data is yet to be proven. Deep learning is a good approach to work with time series data since it can capture time shift that exists in the data but what if the data is highly imbalance? Thus, this study aims at investigating whether resampling technique and deep learning can work best on highly imbalance time series data. The experiments will be made by applying three famous resampling techniques: SMOTE, ROS and RUS on an ANN algorithm. The ANN is also modified into a deep learning named as DANN by increasing the number of hidden layers. Different training-testing ratio is used since resampling is challenged with underfit and overfit issues. Five evaluation metrics are used to record the result which are the AUC, Accuracy, Recall, Precision and Fl-Score. Consequently, Random Un dersampling with lowest training sample performs the best with the deep neural network model to detect anomaly in imbalance time series data. � 2022 IEEE. Final 2023-05-29T09:41:13Z 2023-05-29T09:41:13Z 2022 Conference Paper 10.1109/ICCRD54409.2022.9730424 2-s2.0-85127771308 https://www.scopus.com/inward/record.uri?eid=2-s2.0-85127771308&doi=10.1109%2fICCRD54409.2022.9730424&partnerID=40&md5=dc015024ac1575260dc602f2e33302fa https://irepository.uniten.edu.my/handle/123456789/27223 37 41 Institute of Electrical and Electronics Engineers Inc. Scopus
institution Universiti Tenaga Nasional
building UNITEN Library
collection Institutional Repository
continent Asia
country Malaysia
content_provider Universiti Tenaga Nasional
content_source UNITEN Institutional Repository
url_provider http://dspace.uniten.edu.my/
description Deep neural networks; Time series; Anomaly detection; Capture time; Data imbalance; Electricity theft detection; Imbalance time series data; Over sampling; Resampling; Resampling technique; Time-series data; Under-sampling; Anomaly detection
author2 57220806580
author_facet 57220806580
Saripuddin M.
Suliman A.
Sameon S.S.
format Conference Paper
author Saripuddin M.
Suliman A.
Sameon S.S.
spellingShingle Saripuddin M.
Suliman A.
Sameon S.S.
Impact of Resampling and Deep Learning to Detect Anomaly in Imbalance Time-Series Data
author_sort Saripuddin M.
title Impact of Resampling and Deep Learning to Detect Anomaly in Imbalance Time-Series Data
title_short Impact of Resampling and Deep Learning to Detect Anomaly in Imbalance Time-Series Data
title_full Impact of Resampling and Deep Learning to Detect Anomaly in Imbalance Time-Series Data
title_fullStr Impact of Resampling and Deep Learning to Detect Anomaly in Imbalance Time-Series Data
title_full_unstemmed Impact of Resampling and Deep Learning to Detect Anomaly in Imbalance Time-Series Data
title_sort impact of resampling and deep learning to detect anomaly in imbalance time-series data
publisher Institute of Electrical and Electronics Engineers Inc.
publishDate 2023
_version_ 1806426501398134784
score 13.19449

Similar Items