We present a method for conditional time series forecasting based on the recent deep convolutional WaveNet architecture. The proposed network contains stacks of dilated convolutions that allow it to access a broad range of history when forecasting; multiple convolutional filters are applied in parallel to separate time series and allow for the fast processing of data and the exploitation of the correlation structure between the multivariate time series. The performance of the deep convolutional neural network is analyzed on various multivariate time series and compared to that of the well-known autoregressive model and a long-short term memory network. We show that our network is able to effectively learn dependencies between the series without the need for long historical time series and can outperform the baseline neural forecasting models.

International Conference on Artificial Neural Networks, ICANN 2017
Centrum Wiskunde & Informatica, Amsterdam (CWI), The Netherlands

Borovykh, A.I, Bohte, S.M, & Oosterlee, C.W. (2017). Conditional time series forecasting with convolutional neural networks. doi:10.48550/arXiv.1703.04691