Diversity Guided Evolutionary Mining of Hierarchical Process Models

Abstract

Easy-to-understand and up-to-date models of business processes are important for enterprises, as they aim to describe how work is executed in reality and provide a starting point for process analysis and optimization. With an increasing amount of event data logged by information systems today, the automatic discovery of process models from process logs has become possible. Whereas most existing techniques focus on the discovery of well-formalized models (e.g. Petri nets) which are popular among researchers, business analysts prefer business domain-specific models (such as Business Process Model Notation, BPMN) which are not well formally specified. We present and evaluate an approach for discovering the latter type of process models by formally specifying a hierarchical view on business process models and applying an evolution strategy on it. The evolution strategy efficiently finds process models which best represent a given event log by using fast methods for process model conformance checking, and is partly guided by the diversity of the process model population. The approach contributes to the field of evolutionary algorithms by showing that they can be successfully applied in the real-world use case of process discovery, and contributes to the process discovery domain by providing a promising alternative to existing methods.

Automated Discovery of Process Models from Event Logs: Review and Benchmark

“…Another method to discover BPMN models has been presented in [72]. In this approach, a hierarchical view on process models is formally specified and an evolution strategy is applied on it.…”

“…[19], [25], [26], [33], [67], [68], [69], [75], [77], [87], [93]) were further tested against synthetic logs, while 13 approaches (cf. [12], [16], [19], [55], [56], [63], [68], [71], [72], [79], [83], [84], [96]) against artificial logs. Finally, one method was tested both on synthetic and artificial logs only (cf.…”

Evaluation of the Dynamic Construct Competition Miner for an eHealth System

“…The extraction of a business process model from a given event log without the usage of any a-priori information is addressed by Process Discovery algorithms [19], e.g. [14,16,18,20].…”

“…In Figure 2 the agents and models of the Process Monitor, a module to discover the process behaviour and changes in the process at run-time, are displayed in more detail. The static process extractor employed in the DP Acquisition module could not be utilized for this task due to its non-deterministic behaviour and the non-compliance with run-time requirements (the extractor is based on a genetic algorithm [14]). Instead the Process Monitor was implemented with the Dynamic Constructs Competition Miner as introduced earlier.…”

“…Two types of Process Extractors have been developed in the project: (1) a statically operating genetic miner that works on the historic business process logs [14] and is suitable for long-life business processes that do not change over time, such as the one from Civil Infrastructure domain. However, due to the characteristics of the static genetic miner (non-deterministic as well as long and unpredictable execution time) it is not applicable for monitoring fast and dynamically changing processes such as the ones from the eHealth domain [3].…”