We used the following github project as base code structure https://github.com/hfawaz/dl-4-tsc. We added the augmentation methods, changed the main.py and added plotting/result overview functions to utils.py.

@article{IsmailFawaz2018deep, Title = {Deep learning for time series classification: a review}, Author = {Ismail Fawaz, Hassan and Forestier, Germain and Weber, Jonathan and Idoumghar, Lhassane and Muller, Pierre-Alain}, journal = {Data Mining and Knowledge Discovery}, Year = {2019}, volume = {33}, number = {4}, pages = {917--963}, }

SeSePj2020-2

1). before start, please setup the enviroment according to utils/pip-requirements.txt
2). parameters for augmentation and classification could be modified in utils/constants.py
3). download and extract the UCR datasets https://www.cs.ucr.edu/~eamonn/time_series_data_2018/ into the folder archives/UCRArchive_2018
4). you can choose which UCR datasets to use by changing the list UNIVARIATE_DATASET_NAMES_2018 in utils/constants.py

Two experimental approaches:

approach 1 (not used in our experiments):

normal approach to run serval iterations on the original splits of the dataset

approach 2 (we use this approach in our experiments):

stratified cross validation, the average accuracy of cross validation is taken as evaluation metric

Time Series Classification without augmentation:

approach 1

python main.py --approach 1 --aug noAug --cls allCls --generate_results_overview

approach 2

python main.py --approach 2 --aug noAug --cls allCls --generate_results_overview

Time Series Classification with augmentation:

approach 1

python main.py --approach 1 --aug allAug --cls allCls --generate_results_overview

approach 2

python main.py --approach 2 --aug allAug --cls allCls --generate_results_overview

Generate results overview anytime:

python main.py --generate_results_overview

Plot epochs_loss(& val_loss & accuracy & val_accuracy) overview anytime:

python main.py --plot_epochs_overview