Abstract

Multivariate time series anomaly detection is a crucial task in various domains, including industrial monitoring, cybersecurity, and healthcare. Traditional methods often struggle to capture complex temporal dependencies and inter-variable correlations within high-dimensional data. This paper proposes a novel hybrid deep learning architecture that combines Long Short-Term Memory (LSTM) networks, Convolutional Neural Networks (CNNs), and attention mechanisms for enhanced anomaly detection in multivariate time series data. The LSTM network captures temporal dependencies, while the CNN extracts spatial features from the time series. Attention mechanisms are incorporated to focus on the most relevant features and time steps for anomaly detection. Furthermore, we introduce a feature engineering approach to derive meaningful features from the raw time series data, which are then fed into the deep learning model. The proposed approach is evaluated on several benchmark datasets and compared to state-of-the-art anomaly detection methods. The results demonstrate that our hybrid architecture with attention mechanisms and feature engineering significantly improves anomaly detection performance in terms of precision, recall, and F1-score.