An approach for analyzing business process execution complexity based on textual data and event log
With the advent of digital transformation, organizations increasingly rely on various information systems to support their business processes (BPs). Recorded data, including textual data and event log, expand exponentially, complicating decision-making and posing new challenges for BP complexity analysis in Business Process Management (BPM). Herein, Process Mining (PM) serves to derive insights based on historic BP execution data, called event log. However, in PM, textual data is often neglected or limited to BP descriptions. Therefore, in this study, we propose a novel approach for analyzing BP execution complexity by combining textual data serving as an input at the BP start and event log. The approach is aimed at studying the connection between complexities obtained from these two data types. For textual data-based complexity, the approach employs a set of linguistic features. In our previous work, we have explored the design of linguistic features favorable for BP execution complexity prediction. Accordingly, we adapt and incorporate them into the proposed approach. Using these features, various machine learning techniques are applied to predict textual data-based complexity. Moreover, in this prediction, we show the adequacy of our linguistic features, which outperformed the linguistic features of a widely-used text analysis technique. To calculate event log-based complexity, the event log and relevant complexity metrics are used. Afterward, a correlation analysis of two complexities and an analysis of the significant differences in correlations are performed. The results serve to derive recommendations and insights for BP improvement. We apply the approach in the IT ticket handling process of the IT department of an academic institution. Our findings show that the suggested approach enables a comprehensive identification of BP redesign and improvement opportunities.
Is Part Of
Published in: Information Systems, 10.1016/j.is.2023.102184, Elsevier