EUVIMEDEUVIMED entdecken
Crossref

Method for refining an explanation of an intelligent system decision based on event log data

D.O. Yerokhin

Radiotekhnika · 30.09.2026 · (226) · 17-24

DOI: 10.30837/rt.2026.3.226.02 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Radiotekhnika, online erschienen am 30.09.2026. Verfasst von D.O. Yerokhin.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Intelligent systems for process management support use machine learning models to determine decisions that affect the subsequent execution of business processes. To substantiate such decisions, a user needs an explanation that relates the result of an intelligent system to the data and conditions of its formation. In process management implementation, the course of decision formation is recorded in a process journal as a sequence of events containing data on process execution traces, activities, timestamps, and attributes. Accordingly, such data can be used to refine an explanation. The object of the study is the process of refining an explanation of an intelligent system decision based on event log data. The purpose of the study is to develop a method for refining an explanation of an intelligent system decision based on event log data, which provides the formation, verification, and registration of admissible refinements of an explanation of such a decision. To achieve this purpose, the following tasks are solved: development of a method for refining an explanation of an intelligent system decision based on event log data; experimental verification of the proposed method. A method is developed that includes the selection of identified event log records, the formation of a refinement candidate, evaluation of data sufficiency in the event description, selection of an admissible refinement operation, verification of constraints, and registration of the result. To evaluate data sufficiency, a rule dictionary is used that contains activity classes, candidate templates, operation types, mandatory attributes, and domain interpretation rules. The method distinguishes admissible explanation refinement candidates, event-based candidates whose use for explanation requires additional substantiation, and inadmissible event-based candidates. The method is experimentally verified on an event log fragment. In the test sample, candidates for inclusion in an explanation are identified with the statuses accept, revise, and reject. The results of the experimental verification of the developed method confirm the possibility of analysing candidates for explanation refinement, revising candidates with insufficient attributes, insufficient context, or insufficient interpretation rules, and rejecting operations that have no connection to event log data, create cyclic support dependencies, or change the recorded decision result. Further research is related to expert verification of the rule dictionary, evaluation of the comprehensibility of refined explanations, and verification of the causal validity of statements used in explanation refinement.

Zitieren

APA

Yerokhin, D. (2026). Method for refining an explanation of an intelligent system decision based on event log data. Radiotekhnika, 17-24. https://doi.org/10.30837/rt.2026.3.226.02

Vancouver

Yerokhin D. Method for refining an explanation of an intelligent system decision based on event log data. Radiotekhnika. 2026:17-24. doi:10.30837/rt.2026.3.226.02

BibTeX
@article{Yerokhin2026,
  author = {Yerokhin, D.O.},
  title = {Method for refining an explanation of an intelligent system decision based on event log data},
  journal = {Radiotekhnika},
  year = {2026},
  number = {226},
  pages = {17--24},
  doi = {10.30837/rt.2026.3.226.02}
}
Kostenlos starten →