Evaluating the Effectiveness of Multi-Filtering Techniques on Comprehensibility Improvement of Spaghetti Models

• Published in: 2023 14th International Conference on Computing Communication and Networking Technologies (ICCCNT) pp. 1 - 6
• Main Authors: Singh, Uphar, Hoskere, Yash, Tapas, Nachiket, Vyas, Ranjana, Vyas, O.P.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 06.07.2023
• Subjects: Analytical models; Computational modeling; Data models; Data preprocessing; Filtering; Filtering algorithms; Heuristic algorithms; Lasagna Model; Process discovery; Process Mining; Quality criteria; Spaghetti Model
• ISSN: 2473-7674
• DOI: 10.1109/ICCCNT56998.2023.10308399

While approaching the process enhancement, discovery algorithms become the initial step with event-log-driven techniques. Process discovery techniques work with the aim of finding a model from the logs that can accurately describe the observed behaviour. The process model's fitness increases with the amount of data it can capture. However, due to the very complex event log, the resulting output is a complex web-like unstructured model, also known as the spaghetti model. Resolving this into a simple model, also known as the lasagna model, is critical to the process model's comprehensibility. Range filtering is one of the methods suggested to address the issue with unstructured models. In this paper, we evaluate the efficacy of a multi-range filtering technique for improving the comprehensibility of the spaghetti model. We utilize a Python-based library to evaluate multirange filtering on BPI challenge 2013 and clinical dataset. Process conformance is measured using various quality measures such as fitness, precision, generalization, and simplicity by applying alpha, inductive, and heuristic mining algorithms. The analysis suggests that multi-range filtering techniques produce models that can significantly vary in quality and deviate from generally perceived notions about these algorithms. We conclude that data preprocessing, like cohort analysis, would be helpful before applying any filtering technique. Furthermore, region-based mining algorithms may provide a lossless way to improve comprehensibility.