Experiments with Synthetic Data for Human-in-the-Loop Large-Scale Predictive Maintenance of Workstations

This repository contains toy experiments with synthetic data for the publication

• Alexander Nikitin and Samuel Kaski (2022). Human-in-the-Loop Large-Scale Predictive Maintenance of Workstations.
  [ACM] | [Arxiv]

For production implementation, check Sections "5. Industrial Implementation" and "6.2 Online Experiments" of the paper.

Use

The repo uses git-lfs to store datasets. To fetch the data use:

git lfs fetch

The code was tested with python>=3.6.

To install the required packages, run:

pip install -r requirements.txt
pip install -e .

Structure

The repository contains and implementation of the approach for predictive maintenance of the worksations. The structure is as follows:

• dre_pdm contains utilities for decision rule elicitation modeling and training of the models,
• experiments/simulator.ipynb contains an implementation of the synthetic data simulator,
• experiments/analysis.ipynb contains the experiments with synthetic data from the article, and visualizations,
• data contains a generated dataset.

Experiments.

Simulator Experiments:

Open with jupyter-notebook:

./experiments/analysis.ipynb

Citation

If you found the publication useful for your research, please cite the paper as follows:

@inproceedings{nikitin2022human,
  title={Human-in-the-loop large-scale predictive maintenance of workstations},
  author={Nikitin, Alexander and Kaski, Samuel},
  booktitle={Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
  pages={3682--3690},
  year={2022}
}

License

This software is provided under the Apache License 2.0.