SOFTWARE DEVELOPMENT AND RESEARCH FOR MACHINE LEARNING-BASED STRUCTURAL ERRORS DETECTION IN BPMN MODELS

The most important tool for process management is business process modeling. Business process models allow to graphically represent the sequences of events, activities, and decision points that make up business processes. However, models that contain errors in depicting the business process structure can lead to misunderstanding of a business process, errors in its execution, and associated expenses. Thus, the aim of this study is to ensure the comprehensibility of business process models by detecting structural errors in business process models and their subsequent correction. During the analysis of the Business Process Management (BPM) lifecycle, it was found that the created business process models do not have a stage of control for the presence of errors in them. Therefore, the paper analyzes and improves the BPM lifecycle using the proposed approach. In the improved BPM lifecycle, it is proposed to take into account the correctness validation stage of business process models using the developed software. The paper proposes to process created BPMN (Business Process Model and Notation) models as connected directed graphs. To detect errors in business process models, one of the Machine Learning methods, K-Nearest Neighbors, is chosen, which is a fairly simple and effective classification method. The study also includes the software design and development, its performance validation, and usage to solve the given problem. To analyze the obtained results, the confusion matrix was used and the corresponding quality metrics were calculated. The obtained results confirm the suitability of the developed software for detecting structural errors in business process models. This web application, which is based on the created classification model, allows all interested users to upload business process models in BPMN 2.0 format, view the uploaded models, and analyze them for structural errors.