This repository contains a prototype implementation of a framework for the construction of an Event Knowledge Graph from a stream of events.

The prototype implementation accompanies the paper "Enabling Multi-dimensional Analytics of Digital Traces with Ambiguity" by Marco Franceschetti, Dominik Manuel Buchegger, Ronny Seiger, and Barbara Weber, submitted to CAiSE2025.

The implementation requires Neo4j as the graph database and Python to run the scripts. In our tests we employed Neo4j 5.24 and Python 3.11.9.

The libraries required to run the program can be installed from the requirements.txt file. They include matplotlib, neo4j, numpy, paho-mqtt, pandas, and validators.

In main.py change lines 20-34.

• 20-22: Contains the Neo4j connection settings. These need to be changed according to your Neo4j setup.
• 25-31: Contains the MQTT setup and connection settings. Leave USE_MQTT=False to stream directly from an XES file. Alternatively, you can publish and receive the contents of the file via MQTT (USE_MQTT="SenderReceiver") or only receive MQTT messages if the data is published by another Script (USE_MQTT="Receiver"). Change the connection settings according to your MQTT setup.
• 34: Sets the XES file to stream data from (only set xes_file_path=None if USE_MQTT="Receiver").

Ensure Neo4j is running. Then run main.py.

After executing, the results folder contains a csv file and a graph with the times (in seconds) that Neo4j needed to process each new event in the event stream.

Logs are provided in the logs folder. Log with name ending in "16032" is the log with 400800 events and must be unzipped to be used; log with name ending in "12032" has 300800 events; log with name ending in "8032" has 200800 events; log with name ending in "6432" has 160800 events; log with name ending in "4832" has 120800 events; log with name ending in "3232" has 80800 events; log with name ending in "1632" has 40800 events.

We provide a small sample file under logs/extended_event_log_mock.xes. We show some screenshots of the graph in Neo4j that is constructed from this sample file.

Here we can see an overview of what the Neo4j graph tracks:

Here we can see ambiguous events and the resulting DF (left side: DF labeled with corresponding entity, right side: DF labeled with corresponding probability):

Here we can see ambiguous correlations and the resulting DF (left side: DF labeled with corresponding entity, right side: DF labeled with corresponding probability):

Here we can see ambiguous events & correlations and the resulting DF labeled with probabilities: