Recurrent Neural Networks for Time Series Forecasting: Current Status and Future Directions

TL;DR

This paper provides an extensive empirical evaluation of RNNs for time series forecasting, comparing them with traditional models like ETS and ARIMA, and offers practical guidelines for their application.

Contribution

It presents a comprehensive empirical study and open-source framework for RNN forecasting models, establishing best practices and comparison benchmarks.

Findings

RNNs can model seasonality with homogeneous patterns

Deseasonalization improves RNN performance on complex series

RNNs are competitive alternatives to classical models in many cases

Abstract

Recurrent Neural Networks (RNN) have become competitive forecasting methods, as most notably shown in the winning method of the recent M4 competition. However, established statistical models such as ETS and ARIMA gain their popularity not only from their high accuracy, but they are also suitable for non-expert users as they are robust, efficient, and automatic. In these areas, RNNs have still a long way to go. We present an extensive empirical study and an open-source software framework of existing RNN architectures for forecasting, that allow us to develop guidelines and best practices for their use. For example, we conclude that RNNs are capable of modelling seasonality directly if the series in the dataset possess homogeneous seasonal patterns, otherwise we recommend a deseasonalization step. Comparisons against ETS and ARIMA demonstrate that the implemented (semi-)automatic RNN models are no silver bullets, but they are competitive alternatives in many situations.