Model-centered Ensemble for Anomaly Detection in Time Series

Time-series anomalies detection is a fast-growing area of study, due to the exponential growth of new data produced by sensors in many different contexts as the Internet of Things (IOT). Many predictive models have been proposed, and they provide promising results in differentiating normal and anomalous points in a time-series. In this paper, we aim to find and combine the best models on detecting anomalous time series, so that their different strategies or parameters can contribute to the time series analysis. We propose TSPME-AD (stands for Time Series Prediction Model Ensemble for Anomaly Detection). TSPME-AD is a model-centered based ensemble that trains some of the state-of-the-art predictive models with different hyper-parameters and combines their anomaly scores with a weighted function. The efficacy of our proposal was demonstrated in two real-world time-series datasets, power demand, and electrocardiogram.