Authors: Raffaele Conforti, Marcello La Rosa, Adriano Augusto, Marlon Dumas
This plugin allows users to discover a process map or a BPMN model from an event log. A process map (a.k.a. directly-follows graph) is a visual representation of the log as a graph where nodes capture process activities and directed arcs between these nodes capture sequential order relations between activities. For example, an arc going from activity “Accept order” to activity “Check order” indicates that in the log it has been observed that process cases flow from “Accept order” to “Check order”. Process maps are a simple yet effective means to understand the basic order relationships between process activities. As such, they are the most common type of model discovered by commercial process mining tools.
A window will open up, showing the process map discovered from the log.
Using the two sliders at the top, one can adjust the complexity of the discovered map by increasing and/or reducing the frequency of nodes and arcs that are visualized in the map. This will lead to more, or less, nodes and arcs. For example, this abstraction mechanism can be used to focus on the most frequent process paths. The default values for these sliders are Nodes = 100%, Arcs = 10%.
The plugin also provides the possibility to focus on the most infrequent paths. This is achieved by selecting the options “Invert Priority Nodes” and/or “Invert Priority Arcs”. When these options are selected the sliders will allow you to abstract focusing this time prioritizing the less frequent over the most frequent nodes/arcs.
The plugin provides simple statistics on the event log such us the total, minimum and maximum number of times an activity is executed. This information is provided both as a label on the activities/arcs as well as encoded in the colour of activities (the darker the blue colour, the higher the number of times that activity has been observed in the log) and in the thickness of arcs (the thicker the black line, the higher the frequency of that arc). To change what statistics to show, use the Frequency drop-down list on the top bar.
Statistics are also provided on the time performance of the activities and arcs in the process map. These are total, mean, median, minimum and maximum duration of each arc (indicating the waiting time before starting a given activity, once the previous one has been completed), and total, mean, median, minimum and maximum duration of each arc duration of an activity (a.k.a. the activity’s processing time). If the log only has completion timestamps for each activity and not their start timestamp, these performance statistics will combine both processing time and waiting time into a single time statistics visualized on the arc, while activities will be shown as having an instantaneous duration.
Similar to frequency statistics, time performance statistics are also visualized via labels on activities and arcs, as well as via colours and line thickness (on a red scale) for activities and arcs.
Sometimes, we may be interested in abstracting the process map by focusing only on slow/fast paths. This can be achieved by selecting the “Unlocked” option while being in the performance perspective. Locking or unlocking a perspective allows us to apply the sliders on a specific perspective while visualizing another one.
Visualizing the handover between activities is not the only way a process can be analysed. There will be times where we may be interested in assessing if a certain resource or group of resources are overloaded with work. When clicking on “Selector” a user can decide which attribute of the log will be the focus of the process map.
For example, to visualize the handover of work among a group of resources (i.e. a role withing out organization) we can select the option “org:group”. This option will map the organizational role of each actor involved in our process to a node and will connect two nodes if a handover of work occurs between the two nodes.
When analysing an event log we may be interested in isolating a certain type of behaviour or removing a specific activity and so on. An event log can be filtered clicking on “filter” and creating a new filter that fits our needs. Currently, the plugin allows us to remove all or to retain only events belonging to a specific activity, being performed by a specific resource, or having a specific value for a certain attribute. Additionally, it is also possible to remove or to retain traces containing a certain activity, directly follow-dependency (i.e. an arc in the process map), taking longer than a certain amount of time, or occurring within a given time-frame.
Filters can also be defined by right-clicking on elements of the process map. The following shortcuts exist:
- Right-click on an arc: remove all traces containing the selected direct-follow relation
- CTRL + Right-click on an arc: retain only traces containing the selected direct-follow relation
- Right-click on a node: remove all traces with an event containing the selected attribute
- CTRL + Right-click on a node: retain only traces with an event containing the selected attribute
- Alt (Option) + Right-click on a node: remove all events containing the selected attribute
- Alt (Option) + CTRL + Right-click on a node: retain only events containing the selected attribute
Whenever the insights deriving from the analysis of a process map are not sufficient, the same functionalities are offered on top of a BPMN model. Clicking on the option “BPMN” will automatically discover a BPMN model from an event log. When visualizing a BPMN model, the slider “Parallelism” offers the possibility to adjust the amount of parallelism (e.g. AND and OR gateways) discovered by the plugin.
To assess the quality of a discovered process map/BPMN with respect to a log, the fitness of the model can be measured by clicking on the “Fitness” button.
A filtered log or discovered process map/BPMN can be exported by clicking on the Export button (models can be exported as a “.bpmn”, a PDF, or a PNG file). The Animate button allows one to replay the log on top of the process map, using the Animate a log on top of process map plugin.
Example logs for this plugin can be found in folder “4TU Center Event Logs” of the Apromore repository.
For an advanced analysis of the behavioral relations between process activities we suggest to use the Discover BPMN model (advanced) plugin. Among others, this latter plugin allows the discovery of hierarchical BPMN models with nested sub-processes, activity markers, boundary events and event sub-processes.