Improving multivariate time series forecasting with random walks with restarts on causality graphs

Piotr Przymus, Youssef Hmamouche, Alain Casali, Lotfi Lakhal

The Laboratoire d’Informatique Fondamentale de Marseille (LIF), Aix-Marseille University.

BDSR 2017: Workshop on Big Data & Data Science in Retail in conjunction with ICDM (The IEEE International Conference on Data Mining series)

Experiment details

Selection files

Here you can find csv files generated after first (selection) phase of forecasting.

Forecasting files

Here you can find csv files generated after second (model fitting) phase of forecasting. Note that not all selection files allowed proper model fitting, for details see next section.

Details for models that could not be fitted

Here you will find details for which files and why a model could not be fitted.