Models depicting the behavior and performance of processes can be discovered through Process Mining. These models often facilitate visualizing alternate paths and rework through control-flow logic, but identifying all process variants in such models can become a complex endeavor. To address this, a tool that allows visualizing all process variants within a single model is proposed, in the form of Directed Rooted Trees whose leaves are the end states of variants. The generated trees are also decorated with cost information for quickly comparing the cost performance of variants. The tool is expected to be useful for performing exploratory analyses.
the two without applying filters or analyzing the data within the event log.
To address the above, this work proposes a tool that allows the visualization of all process
variants within a single process model. Specifically, the proposed process model is in the form
of a transition system constructed as a Directed Rooted Tree (DRT). In transition systems, the
events of the process are the transitions that imply the flow from one state of the system to
another [
Transition systems can be extended to include additional information [
The DRT generated by the proposed tool possess similarities with context trees [
The remainder of this work is structured as follows. Section 2 describes the main features of the tool, as well as its interface. In Section 3, the maturity of the tool is discussed. Section 4 provides links for the demonstration of the tool in the form of a tutorial document and video. Finally, the conclusions of this work are presented in Section 5.
The proposed tool is available at https://bit.ly/drt-variant-costs. It has been developed as a
web application using the Angular framework [
The buttons in area (1) enable the distinct features of the tool. The "Load Event Log" button visualizes a modal for loading an event log to the tool. Once loaded, the tool automatically generates the resulting DRT. The "Filter Event Log" button visualizes a modal containing three types of filters: “Activities to Include”, which filters the diagram to only visualize events of selected activities; “Start Activities”, which filters the diagram to only include cases that start with selected activities; and “End Activities”, which filters the diagram to only include cases that end with selected activities. The "Configuration" button visualizes a modal that allows changing the displayed currency, and the visualization of additional metrics in every state and transition of the DRT. Finally, the "Download Diagram (PNG)" and "Download Diagram (SVG)" buttons allow the download of a *.png or a *.svg file, respectively, containing the generated DRT.
By default, the states (nodes) of the transition system of the generated DRT contain the average case cost of executions flowing through them, whereas transitions (arcs) contain the activities executed between every state, and their frequency. States are colored using a trafic light color range (from green, to yellow, to red). The closer to green, the cheaper the average cost of executions flowing through the state. Conversely, the closer to red, the more expensive the average cost of flowing executions. The line width of transitions relates to the case frequency, i.e., a wider transition line corresponds to one with a higher occurrence frequency.
The tool is currently in a stable version. That is, no major bugs are known and all the features described in this work are implemented. Nevertheless, the maturity of the tool can be described in relation to its current limitations.
First, as any transition system, the generated DRT has problems expressing parallelism.
Parallelism occurs when two activities of a case overlap [
Second, only visualization of cost performance is currently supported. As future work, the possibility of extending the tool to also support waiting times between activities is considered. The visualization of any metric that can be aggregated is also foreseen.
Third, the tool has been designed for processing any event log that complies with the required
format. However, loading event logs containing large numbers of cases (e.g., over 10,000) will
result in long processing times. Moreover, the utility of the tool is reduced when loading event
logs with excessive numbers of variants (e.g, over 100), as their comparison would be dificult
without applying filters. To facilitate the analysis of processes with several variants, the trace
clustering techniques used for context trees [
Fourth, design limitations, such as the currently required event log format, could be addressed through future work. Namely, through input boxes for users to indicate column names of the event log or the format of timestamps. The possibility of downloading the DRT in editable formats, such as JSON, or the notation of the used diagramming library, is also considered.
The tool has not been used for real case studies yet. However, it was tested with the road
trafic fine management event log, which contains real-life records of an information system
from an Italian city [
A tutorial document containing step-by-step instructions on how to utilize the tool has been constructed. This document can be found at https://bit.ly/drt-tutorial-document. A video summarizing this tutorial can be found at https://bit.ly/drt-video-tutorial.
In the tutorial, an example is used for illustrative purposes. The example is based on the
blasting process of a Chilean copper mine. A model with the activities of this process is publicly
available in Appendix 4 of [
This work proposes a tool for the visualization of process variants within a single process model, in the form of a Directed Rooted Tree. Every leaf of the DRT is the end state of a variant, which facilitates their side-to-side comparison. The above is enhanced by the inclusion of cost metrics, which are supported by a color range for the identification of cheaper and more expensive process variants.
The proposed tool is expected to serve as a complementary visualization of discovered processes when performing exploratory analyses of event logs.
This work was funded by the Agencia Nacional de Investigación y Desarrollo de Chile. Grant numbers ANID FONDECYT 1230697 and ANID-Subdirección de Capital Humano/Doctorado Nacional/2021-21210022.