The repository demonstrates some of the models fitted as part of seminar research (link to paper) at HU Berlin on exploring econometric as well as LSTM neural network-based forecasting approaches to predict step-ahead prices of Brent crude oil 1-month futures.
The out-of-sample forecast period is from September 2016 to April 2023.
Custom functions for grid search and stateful training used in the LSTM notebooks were later refined, extended and wrapped up in a StatefulLSTM class in the following repository.
Note that the repository was created from some of the notebooks and files that had to be migrated from several different devices therefore paths and filenames specified within the notebooks will not be the same as that in the repository itself.