An Empirical Analysis of Diagnosis of Industrial Business Processes at Sub-process Levels

Business process models expressed in languages such as BPMN (Business Process Model and Notation) play a critical role in implementing the workflows in modern organizations. However, control flow errors such as deadlock and lack of synchronization as well as syntactic errors arising out of poor modeling practices often occur in industrial process models. In this paper, we provide an empirical diagnostic analysis of such errors for real-life industrial process models. The investigation involved models from different application domains. It turns out that error frequency has non-linear relation with error depth (the maximum depth at which an error occurred) across models from all domains. Error occurrence has statistically significant correlations (p <; 0.0001) with the size of sub-processes as well as with the swim-lane interactions, however only the former correlation is strong (Spearman's ρ = 0.579). We also develop a logistic regression model to estimate error probability in terms of the following metrics: sub-process size, coefficient of connectivity, sequentiality and structuredness; the predictive model fits well with the data (χ2(4, N = 1261) = 720.68, p <; 0.001).