/explainable_ts

Explainable Neural Networks for Time Series Forecasting

Primary LanguageJupyter Notebook

Explainable Neural Networks for Time Series Forecasting

Citation

SSRN - DNNLITS

@article{ozyegendnnlits,
  title={Dnnlits: Deep Neural Networks for Locally Interpretable Time Series Forecasting},
  author={Ozyegen, Ozan and Cevik, Mucahit and Basar, Ayse},
  journal={Available at SSRN 4179881}
}

Experiments

  • Train single model
    • python exps/run_single.py
  • Tune a dataset-model pair via optuna framework
    • python exps/run_tuning.py
  • Analyze the model performances
    • python exps/review_results.ipynb
  • Analyze model forecasts and explanations
    • python notebooks/polar_rnn_rossmann.ipynb
    • A trained Rossmann $DNNLITS^{RNN}$ model checkpoint is available under models/3r5uvsl1

Data Sources

Rossmann

Walmart

Requirements

  • python 3.8
  • pytorch 1.11
  • pytorch-forecasting 0.10.1
  • wandb - Weights and Biases is used for tracking the experiments
  • exp_ts.yml contains all the package dependencies