A heuristic method for detecting problems in business process models

The purpose of this paper is to present a new heuristic approach for finding errors and possible improvements in business process models.

First, the paper translates the information that is included in a model into a set of Prolog facts. It then searches for patterns which are related to a violation of the soundness property or bad modeling style or otherwise gives rise to the assumption that the model should be improved. To validate this approach, the paper analyzes a repository of almost 1,000 business process models. For this purpose, three different model‐checkers that explore the state space of all possible executions of a model are used. The result of these tools are compared with the results given by this heuristic approach.

The paper finds that the heuristic approach identifies violations of the soundness property almost as accurate as model‐checkers. However, other than these tools, the approach never ran into state space explosion problems. Furthermore, this heuristic approach can also detect patterns for bad modeling style which can help to improve the quality of models.

Heuristic checks can run in the background while the modeler works on the model. In this way, feedback about possible modeling errors can be provided instantly. This feedback can be used to correct possible problems immediately.

Current Prolog‐based validation tools check mainly for syntactical correctness and consistency requirements. This approach adds one more perspective by also detecting control‐flow errors (like deadlocks) and even pragmatic issues.