Dataset: Evaluation of post-hoc interpretability methods in time-series classification

This repository contains the dataset, trained models as well as results for the article Evaluation of post-hoc interpretability methods in time-series classification.

The code to reproduce the results presented in the article is available on GitHub. More details on the data and results can be found in the article.

Files:

datasets.zip: Include the three datasets used in the article:

ECG: Processed version of the CPSC dataset from Classification of 12-lead ECGs: the PhysioNet - Computing in Cardiology Challenge 2020.

fordA: Dataset from the UCR Time Series Classification Archive

synthetic: Synthetic dataset developed specifically for the purpose of the article

trained_models.zip: Include CNN, transformer and bi-lstm trained on the three datasets

results_paper.zip: Computed relevance and evaluation metrics for the trained models

model_interpretability: Include the relevance computed using the different interpretability methods as well as the computed metrics for each method 

summary_results: Summary of the evaluation metrics across all interpretability methods for each dataset as well as an excel file summarising the metrics across all datasets.