Model Monitoring and Dynamic Model Selection in Travel Time-series Forecasting

This repository contains the code and the useful links to reproduce the experiments in the paper Model Monitoring and Dynamic Model Selection in Travel Time-series Forecasting.

The M4 competition dataset, together with the monitored models used in the framework, are available in the original M4 repository.

The implementation of the two meta-learning approaches used for the comparison with our method can be found in the following links:

• FFORMS: https://github.com/thiyangt/WorkingPaper1
• ADE: https://github.com/vcerqueira/tsensembler

Getting started

Requirements

• Tensorflow
• Keras
• GPy

Usage

To use one of the available monitoring models, you can run the corresponding python script, specifying the path of the needed files. Example for LSTM model:

python lstm.py --observations=<OBSERVATIONS_PATH> --true_values=<TRUE_VALUES_PATH> --forecasts=<FORECASTS_PATH>

Parameters:

• OBSERVATIONS_PATH: path of csv file containing the observations
• TRUE_VALUES_PATH: path of csv file containing the true_values
• FORECASTS_PATH: path of csv file containing the forecasts

To perform dynamic model selection, you can run the script dynamic_model_selection.py, specifying the path of the needed files:

python dynamic_model_selection.py --observations=<OBSERVATIONS_PATH> --true_values=<TRUE_VALUES_PATH> --forecasts_folder=<FORECASTS_FOLDER>

Parameters:

• OBSERVATIONS_PATH: path of csv file containing the observations
• TRUE_VALUES_PATH: path of csv file containing the true_values
• FORECASTS_FOLDER: path of the folder containing the csv files of forecasts

Citation

Please cite it as follows:

@inproceedings{c2020model,
    year =        {2020},
    title =       {{M}odel monitoring and dynamic model selection in travel time-series forecasting},
    author =      {{C}andela, {R}osa and  {M}ichiardi, {P}ietro and  {F}ilippone, {M}aurizio and  {Z}uluaga, {M}aria {A}},
    booktitle =   {{ECML}-{PKDD} 2020, {T}he {E}uropean {C}onference on {M}achine {L}earning and {P}rinciples and {P}ractice of {K}nowledge {D}iscovery in {D}atabases, 14-18 {S}eptember 2020, {G}hent, {B}elgium},
    address =     {{G}hent, {BELGIUM}},
    month =       {09},
    url =         {https://arxiv.org/abs/2003.07268}
}