Genetic Process Mining: Alignment-Based Process Model Mutation

Abstract

The Evolutionary Tree Miner (ETM) is a genetic process discovery algorithm that enables the user to guide the discovery process based on preferences with respect to four process model quality dimensions: replay fitness, precision, generalization and simplicity. Traditionally, the ETM algorithm uses random creation of process models for the initial population, as well as random mutation and crossover techniques for the evolution of generations. In this paper, we present an approach that improves the performance of the ETM algorithm by enabling it to make guided changes to process models, in order to obtain higher quality models in fewer generations. The two parts of this approach are: (1) creating an initial population of process models with a reasonable quality; (2) using information from the alignment between an event log and a process model to identify quality issues in a given part of a model, and resolving those issues using guided mutation operations. (a) The four quality dimensions used to calculate process model fitness (from [1]). Stop? Select Change Evaluate Create [ Wednesday 30 th April, 2014 at 17:56-?? pages-version 0.1 ] (b) The different phases of a genetic process discovery algorithm, as given in [3].

Monitoring Interactions Across Multi Business Processes with Token Carried Data

“…Traditionally, process discovery is to mine a process model from the event data on the viewpoint of control flow [19], [20]. There are many process discovery algorithms, such as α [21], α + [22], α ++ [23], α # [24], β [25], λ [26], τ, region-based [27] and genetic ones [28]. Recently, more and more works related to data perspective have been done and the mined achievements are applied to process data management successfully, e.g., processing queries over workflow executions [29].…”

Human Performance-Aware Scheduling and Routing of a Multi-Skilled Workforce

“…Process models usually combine both perspectives and describe the allowed ordering of activities [6], [7], which means that they contain information on the jobs to be scheduled and their precedence relations. In this section we provide an approach to incorporate process models in the form of process trees [35], [36], a formalism to specify block-structured workflow nets [4], into our scheduling problem.…”

“…Process trees, however, are guaranteed to represent sound process models and their structure enables a more straightforward mapping between process knowledge and scheduling input. Figure 2 shows the possible operators that process trees can be composed of, and their translations to BPMN constructs [35], [36]. Operator nodes specify the relation between their children.…”

“…Hence a non-exclusive choice becomes an exclusive choice in a scheduling context and the loop becomes a sequence of a single 'do' and 'exit'. Figure 3 shows an example process model for a simple hardware maintenance process in the form of a process tree [35], [36] and a BPMN model [4]. Each instance of this process starts with the execution of a single Problem Intake (PI) activity.…”

Reduced alignment based on Petri nets

“…In order to achieve the optimal alignments, it needs a lot of time to consume and memory space to occupy. Since the alignment approach was proposed, many studies have been carried out based on the alignment, for example, alignment‐based model repair,(–) precision checking,() genetic process mining, and so on. Hence, the alignment is more and more important in the field of process mining.…”