Anomaly Novelty Detection For Univariant Time- Series Isung Hybrid Deep Learning

Citation

Mohd Zebaral Hoque, Jesmeen and Hossen, Jakir and Abd. Aziz, Azlan (2023) Anomaly Novelty Detection For Univariant Time- Series Isung Hybrid Deep Learning. In: 2nd FET PG Engineering Colloquium Proceedings 2023, 1-31 December 2023, Multimedia University, Malaysia. (Submitted)

[img] Other (Poster)
2. PPM Poster Jesmeen Mohd Zebaral Hoque - Noor Ziela.pdf - Submitted Version
Restricted to Repository staff only

Download (744kB)

Abstract

Time series and streaming data generated by sensors in Smart Home Systems(SHS) crucial to integrate with anomaly detection for identifying irregular patterns, potential fraud, and ensuring the accuracy of collected information. The proposed approach enables learning and the extraction of features from time-series data automatically. Evaluation methods include common distance-based (LOF, Mahalanobis), common probabilistic (Isolation Forest, OneClass SVM, Elliptic Envelope), and proposed forecastingbased (CNN, LSTM, CNN-LSTM). Notably, the algorithm distinguishes seasonal abnormalities from true abnormalities, emphasizing an unsupervised approach without the need for explicit feature extraction.

Item Type: Conference or Workshop Item (Poster)
Uncontrolled Keywords: Smart Home Systems(SHS), deep learning
Subjects: Q Science > Q Science (General) > Q300-390 Cybernetics
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7800-8360 Electronics
Divisions: Faculty of Engineering and Technology (FET)
Depositing User: Ms Rosnani Abd Wahab
Date Deposited: 03 Apr 2024 02:36
Last Modified: 03 Apr 2024 02:36
URII: http://shdl.mmu.edu.my/id/eprint/12354

Downloads

Downloads per month over past year

View ItemEdit (login required)