Prolothar Queue Mining

Algorithms for queue mining (discovering discrete event simulations based on waiting queue models) from event logs

Based on the publication

Boris Wiegand, Dietrich Klakow, and Jilles Vreeken. Why Are We Waiting? Discovering Interpretable Models for Predicting Sojourn and Waiting Times. In: Proceedings of the SIAM International Conference on Data Mining (SDM), Minneapolis, MN. 2023, pp. 352–360.

Prerequisites

Python 3.11+

Usage

If you want to run the algorithms on your own data, follow the steps below.

Installing

pip install prolothar-queue-mining

Creating a dataset

from prolothar_queue_mining.model.job import Job
from prolothar_queue_mining.inference.queue import CueMin

# our input data are jobs with an ID and their corresponding arrival resp. departure time
observed_arrivals = [
    (Job('A'), 3),
    (Job('B'), 4),
    (Job('C'), 5),
    (Job('D'), 6),
    (Job('E'), 7),
    (Job('F'), 8),
]
observed_departues = [
    (Job('A'), 4),
    (Job('B'), 7),
    (Job('C'), 11),
    (Job('D'), 12),
    (Job('E'), 13),
    (Job('F'), 14),
]

#you can add additional features to a job, example:
Job('4711', {'color': 'blue', 'size': 12})

cuemin = CueMin(verbose=True)

#if your jobs have features, which can have an influence on the service order or service time:
cuemin = CueMin(verbose=True, categorical_attribute_names = ['color'], numerical_attribute_names = ['size'])

#find and a print a waiting queue model
queue = cuemin.infer_queue(observed_arrivals, observed_departues)
print(queue)

#if you want to use domain knowledge to restrict the number of servers, e.g. min 2 and max 4:
cuemin = CueMin(search_strategy_name='linear-2-4')
queue = cuemin.infer_queue(observed_arrivals, observed_departues)

Development

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

Additional Prerequisites

• make (optional)

Compile Cython code

make cython

Running the tests

make test

Deployment

Optional requirement: Create a .pypirc file in the project root directory with your pypi authentication token:

[pypi]
username = __token__
password = pypi-AgEIcH...

1. Change the version in version.txt
2. Build the package

make clean_package
make package

3. Deploy the version to Pypi:

 make publish

or

twine upload --skip-existing --verbose --config .pypirc dist/*

4. Create and push a tag for this version by

git tag -a $(cat version.txt) -m "describe this version"
git push --all && git push --tags

Devcontainer

There is a decontainer definition in this project, which helps you to set up your environment. At Stahl-Holding-Saar, we are behind a corporate proxy and cannot install dependencies from PyPi directly. I yet have not found a stable solution to set the PIP_INDEX_URL and PIP_TRUSTED_HOST variables dynamically. In the current Dockerfile, I hardcoded the values, so you have to adapt them. If you know a solution to this problem, please contact me.

Versioning

We use SemVer for versioning.

Authors

If you have any questions, feel free to ask one of our authors:

• Boris Wiegand - boris.wiegand@stahl-holding-saar.de