Proceedings chapter
Open access

Forecasting and Granger Modelling with Non-linear Dynamical Dependencies

Presented at Skopje, Macedonia, 18-22 September 2017
PublisherCham : Springer
Publication date2017

Traditional linear methods for forecasting multivariate time series are not able to satisfactorily model the non-linear dependencies that may exist in non-Gaussian series. We build on the theory of learning vector-valued functions in the reproducing kernel Hilbert space and develop a method for learning prediction functions that accommodate such non-linearities. The method not only learns the predictive function but also the matrix-valued kernel underlying the function search space directly from the data. Our approach is based on learning multiple matrix-valued kernels, each of those composed of a set of input kernels and a set of output kernels learned in the cone of positive semi-definite matrices. In addition to superior predictive performance in the presence of strong non-linearities, our method also recovers the hidden dynamic relationships between the series and thus is a new alternative to existing graphical Granger techniques.

Citation (ISO format)
GREGOROVA, Magda, KALOUSIS, Alexandros, MARCHAND-MAILLET, Stéphane. Forecasting and Granger Modelling with Non-linear Dynamical Dependencies. In: Machine Learning and Knowledge Discovery in Databases, European Conference, ECML PKDD 2017, Proceedings, Part II. Skopje, Macedonia. Cham : Springer, 2017. p. 544–558. doi: 10.1007/978-3-319-71246-8_33
Main files (1)
Proceedings chapter (Accepted version)

Technical informations

Creation06/13/2019 3:23:00 PM
First validation06/13/2019 3:23:00 PM
Update time03/15/2023 5:25:46 PM
Status update03/15/2023 5:25:45 PM
Last indexation01/17/2024 5:43:59 AM
All rights reserved by Archive ouverte UNIGE and the University of GenevaunigeBlack