dynamik is an algorithm for analyzing the dynamics of a process execution, detecting changes in the cycle time and,
if present, providing insights about the causes of the change.
dynamik is designed to be executed using an event log as it's input, processing it sequentially, event by event,
mimicking an online environment where events are consumed from an event stream.
To run the algorithm in your command line, you need an event log in CSV format.
The file is expected to have a headers row which is followed by a row per event:
case, start, end, activity, resource
2182, 2023-02-24T05:28:00.843, 2023-02-24T05:28:00.843, START, resource-000001
2182, 2023-02-24T05:28:00.843, 2023-02-24T05:34:31.219, Activity 1, resource-000044
2182, 2023-02-24T05:34:31.219, 2023-02-24T05:47:25.817, Activity 2, resource-000024
2182, 2023-02-24T05:47:25.817, 2023-02-24T05:59:46.195, Activity 3, resource-000010
2182, 2023-02-24T05:59:46.193, 2023-02-24T05:59:46.193, END, resource-000001
7897, 2023-03-01T08:39:42.861, 2023-03-01T08:39:42.861, START, resource-000001
7897, 2023-03-01T08:39:42.861, 2023-03-01T08:53:41.167, Activity 1, resource-000029
7897, 2023-03-01T08:53:41.167, 2023-03-01T08:56:46.299, Activity 2, resource-000007
7897, 2023-03-01T08:56:46.299, 2023-03-01T09:12:49.468, Activity 3, resource-000018
7897, 2023-03-01T09:12:49.468, 2023-03-01T09:12:49.468, END, resource-000001
... ... ... ... ...
If you need a different mapping for processing your log files, you can specify it using a JSON file:
{
"case": "<your case attribute name>",
"activity": "<your activity attribute name>",
"resource": "<your resource attribute name>",
"start": "<your start time attribute name>",
"end": "<your end time attribute name>"
}Then, you can then run dynamik with the option --mapping MAPPING_FILE to use your custom mapping:
$> dynamik ./log.csv --mapping ./my_custom_mapping.jsonRun dynamik --help to check the additional options:
$> dynamik --help
usage: dynamik [-h] [-f FORMAT] [-m MAPPING_FILE] [-t TIMEFRAME] [-a ALPHA] [-v] LOG_FILE
Explainable Performance Drift is an algorithm for finding actionable causes for drifts in the
performance of a process execution. For this, the cycle time of the process is monitored, and,
if a change is detected in the process performance, the algorithm finds the actionable causes
for the change.
positional arguments:
LOG_FILE The event log, in CSV or JSON format.
options:
-h, --help show this help message and exit
-f FORMAT, --format FORMAT specify the event log format
-m MAPPING_FILE, --mapping MAPPING_FILE provide a custom mapping file
-t TIMEFRAME, --timeframe TIMEFRAME provide a timeframe size, in days, used
to define the reference and running models.
-a ALPHA, --alpha ALPHA specify the confidence for the statistical tests
-v, --verbose enable verbose output. High verbosity level can
drastically decrease dynamik performance
dynamik is licensed under the Apache License, Version 2.0
Aside of providing an executable command, dynamik can be fully customized by using it as a Python package.
If you use poetry you can install dynamik directly from git:
$> poetry add "https://gitlab.citius.usc.es/ProcessMining/explainable-performance-drift.git"
When using it as a package, the drift detection algorithm can be located at dynamik.drift.detect_drift.
dynamik provides you with all the utilities needed to read CSV files in a performant way.
However, we are aware that your logs may be in a different format. You may even want to read your logs from a database
or message queue! Don't worry, we've got you covered. The representation we use for logs in expert is a
simple Generator object, so you can implement any data source you need, and as long as it returns an Generator
of expert.model.Event objects you shouldn't have any problems. As a reference, you can check the implementations
from expert.input.csv.read_csv_log.