Fuzzy cognitive maps can be used as a simple model for multivariate data forecasting. Reacreation of this approach in Python 3.8 yields interesting results, albeit weaker than expected. This method is tested against the UWave time series dataset.
dataset_instructions.txt- explains how to set up the UWave dataset.requirements.txt- use for Python pip install.
fcm_train.py- trains the model and stores it in a json file.fcm_test.py- tests the models given in a json file.fcm.py- shortcut for doing both of the above.
All the additional branches in the repository are different configurations of the model used for experiments. Most of these have been used in the project report, which you may read below or on Overleaf.
