-Object-centric process mining provides a more holistic view of processes where we analyze processes with multiple case notions. However, most object-centric process mining techniques consider the whole event log rather than the comparison of existing behaviors in the log. In this paper, we introduce a stand-alone object-centric process cube tool built on the PM4PYMDL process mining framework. Our infrastructure uses both object and event attributes to build the process cube which leads to different types of materialization. Furthermore, our tool is equipped with the state of the art object-centric process mining techniques. Through our tool the user can visualize the extracted object-centric event log from process cube operations, export the object-centric event log, discover the state-of-the-art objectcentric process model for the extracted log, and compare the process models side-by-side. Index Terms-Object-Centric Process Mining, Object-Centric Event Logs, Process Comparison, Process Cubes
Process mining is a field of data science that aims to
bridge the gap between business process model-based
analysis and data-oriented analysis. Process mining techniques
include process discovery, conformance checking, and process
enhancement methods [
Event logs are the starting point to apply process mining
techniques. Event logs consist of events where each event
refers to one case notion, activity, timestamp, and some
additional attributes such as resource. Common process mining
techniques are based on event logs with one case notion,
however, in reality, several case notions are involved in one
process, e.g., a simple Order-to-Cash (O2C) process where
orders, offers, and invoices are involved. Object-centric
process mining is a novel branch of process mining that aims
to develop process mining techniques on top of
ObjectCentric Event Logs (OCELs) [
In this demo paper, we present an interactive tool that
permits the user to compare object-centric processes with each
other through process cube operations. Furthermore, it permits
to discover object-centric process models (object-centric
Directly Follows Graphs [
The remainder of this paper is organized as follows. In Section II, we describe the main functionalities of the tool that are provided. In Section III, we evaluate the scalability of the tool. Finally, Section IV concludes the paper and presents some future work for the extension of our work.
In our open-source tool1, called OCPC (Object-Centric Process Cube), we have used Tk GUI toolkit as the user interface which is one of the popular standard Python interfaces. A snapshot of the tool is shown in Figure 1 where three sub windows are highlighted:
Input: the input is JSON-OCEL/XML-OCEL. In OCEL, an event is related to event attributes and some possible objects related to that event. Furthermore, each object can have its own properties in another table. Therefore, as shown in Figure 1, in the first highlighted sub window, we have shown events with event attributes and their objects in one table, whereas, objects within their attributes are shown in separate tables. The user can select event attributes and object attributes as the dimensions of the process cube and build the cube.
Wizard: After creating the cube, we see overviews of the cube in the second sub window. It is possible to select dimensions of the cube as rows and columns and see the number of events in the selected slices/dices.
The combo box ”Materialization” is related to design choices that we have in this tool for materializing that did not exist in the previous process cubes. In the ”Existence” option for materialization, there should exist an object in that event that satisfies the property. For example, in the left-view in Figure 1, the option is ”Existence” and the highlighted green dice in this view shows in 557 events Echo show 8 was involved. In the ”All” option 1https://github.com/AnahitaFarhang/object-centric-process-cube for materialization, all the objects, in the event, should satisfy that property. An example is shown in right-view where this option is ”All”. The highlighted green cell in this view shows that in 8 events all the items were Echo show 8.
Output: In the third sub window, we have compared the process models of the selected cells shown in the second sub window. As we see, there are differences in the duration of activities for the selected cells. Furthermore, depending on the operation selection in the second sub window, it is possible to compare the extracted event logs, and Object-Centric Petri nets of the selected cells.
The scalability of the tool in terms of number of events, event attributes, and object attributes are shown in Tables I, II, and III, respectively2. The result of the analysis with different settings shows the time required for creating the cube increases linearly, linearly, and non-linearly when increasing the number of events, event attributes, and object attributes, respectively. These relationships are justifiable by the nature of OCELs where an event can contain multiple objects and there is a oneto-many relationship between an event and its objects which results in a non-linear relationship shown in Table III.
Here, we present an interactive object-centric process cube tool that enables the exploration, discovery of object-centric 2All the above experiments were performed on a laptop with the specifications: PC Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz. process models along with side-by-side model comparison. Using the tool that is implemented in PM4PY-MDL, users explore object-centric event logs using process cube operations, export the partitioned OCEL, and discover the object-centric process models, annotated with frequency and performance, which helps in understanding processes better. A video displaying the functionalities of our tool is available at the address https://youtu.be/zenHt3wdZP4. We aim in the future to extend this tool with an automatic method to choose interesting slices and dices to explore the cube.