OS-PGSM and OEP-ROC

Code accompanying the papers Explainable Online Ensemble of Deep Neural Networks Pruning for Time Series Forecasting and Explainable Online Deep Neural Network Selection Using Adaptive Saliency Maps for Time Series Forecasting.

Datasets

Some datasets need to be downloaded and placed in code/datasets/monash_ts for the experiments to work. Specifically, you need to download the Electricity (Hourly), KDD Cup 2018, Pedestrian Counts, Solar (10 minutes) and Weather datasets from https://forecastingdata.org/. If available, choose the datasets without missing values. Also, you need to rename them to electricity_hourly.tsf, kdd_cup_2018.tsf, pedestrian_counts.tsf, solar_10_minutes.tsf and weather.tsf.

Installing dependencies

We use python version 3.7.10. Install all dependencies via

pip install -r requirements.txt

Overview of the repository

• code/compositors.py contains the code for OEP-ROC and for OS-PGSM.
• code/experiments.py contains all experiment parameters and configurations.