Forecasting across time series databases using recurrent neural networks on groups of similar series: A clustering approach

• Published in: Expert systems with applications Vol. 140; p. 112896
• Main Authors: Bandara, Kasun, Bergmeir, Christoph, Smyl, Slawek
• Format: Journal Article
• Language: English
• Published: New York Elsevier Ltd 01.02.2020; Elsevier BV
• Subjects: Algorithms; Big data forecasting; Clustering; Competition; Datasets; Domains; Forecasting; Forecasting techniques; LSTM; Mathematical models; Neural networks; Recurrent neural networks; RNN; Subgroups; Time series; Time series clustering; Big data forecasting; RNN; LSTM; Time series clustering; Neural networks
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2019.112896

•Building a model on a set of related time series can improve the forecast accuracy.•Performance of the global models can degenerate if built on disparate time series.•A subgrouping strategy then augments the accuracies of the baseline global models. With the advent of Big Data, nowadays in many applications databases containing large quantities of similar time series are available. Forecasting time series in these domains with traditional univariate forecasting procedures leaves great potentials for producing accurate forecasts untapped. Recurrent neural networks (RNNs), and in particular Long Short Term Memory (LSTM) networks, have proven recently that they are able to outperform state-of-the-art univariate time series forecasting methods in this context, when trained across all available time series. However, if the time series database is heterogeneous, accuracy may degenerate, so that on the way towards fully automatic forecasting methods in this space, a notion of similarity between the time series needs to be built into the methods. To this end, we present a prediction model that can be used with different types of RNN models on subgroups of similar time series, which are identified by time series clustering techniques. We assess our proposed methodology using LSTM networks, a widely popular RNN variant, together with various clustering algorithms, such as kMeans, DBScan, Partition Around Medoids (PAM), and Snob. Our method achieves competitive results on benchmarking datasets under competition evaluation procedures. In particular, in terms of mean sMAPE accuracy it consistently outperforms the baseline LSTM model, and outperforms all other methods on the CIF2016 forecasting competition dataset.