Attention-Augmented LSTM Autoencoders for Improved Anomaly Detection in Time-Series Data

Naik, Karthik Shankar (2025) Attention-Augmented LSTM Autoencoders for Improved Anomaly Detection in Time-Series Data. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (622kB) | Preview

Abstract

Detecting anomalies in time-series data is critical in many domains, including transportation, healthcare, and cybersecurity. Traditional autoencoders often rely solely on the final hidden state to reconstruct input sequences, limiting their ability to focus on subtle, localized deviations. This study leverages an attention-enhanced LSTM autoencoder that incorporates a temporal attention mechanism to dynamically weigh the importance of each timestep during reconstruction, thereby improving the model’s sensitivity to anomalies, using a time series dataset and compared the leveraged model against baseline deep learning architectures including RNN, GRU, and vanilla LSTM autoencoders. Experimental results show that our model significantly outperforms these baselines, achieving a macro-averaged F1-score of 0.82 and an anomaly-specific F1-score of 0.66. These improvements highlight the benefit of integrating attention mechanisms into sequence modelling for more accurate anomaly detection. The findings suggest strong potential for the model's application in real-time monitoring systems. It also discusses limitations such as computational overhead and dataset specificity, and proposes future directions, including expansion to multivariate datasets and lightweight deployment on edge devices.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Niculescu, Hamilton UNSPECIFIED
Uncontrolled Keywords:	Time-Series Anomaly Detection; LSTM Autoencoder; Attention Mechanism; Deep Learning; Temporal Modelling
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	25 Aug 2026 15:18
Last Modified:	25 Aug 2026 15:18
URI:	https://norma.ncirl.ie/id/eprint/9637

Actions (login required)

View Item