Leveraging Empirical Mode Decomposition for Time Series Anomaly Detection in IoT and Smart City Applications

Abstract

This paper proposes a novel forecasting-based anomaly detection method that leverages Empirical Mode Decomposition (EMD) to break down complex univariate time series into Intrinsic Mode Functions (IMFs). Gated Recurrent Units are tasked with processing IMFs to predict future values. By reconstructing the time series from predicted IMFs, anomalies are detected when deviations between predicted and observed values of the time-series occur. The proposed method is validated on real-world datasets from smart city applications, demonstrating its efficiency in handling noisy data and multiscale seasonal trends while maintaining low computational overhead, making it suitable for deployment in resource-constrained IoT environments.