INEXA: Interactive and Explainable Process Model Abstraction Through Object-Centric Process Mining

TL;DR

The paper tackles the challenge of discovering and exploring process models from multi-granularity event logs in real-world settings. It introduces INEXA, an interactive, explainable method that leverages object-centric process mining and a log–model link to apply and trace abstractions while preserving a direct connection to the original event log. The approach starts with discovering a model on the original log, then reduces it to a displayable starting point and enables iterative, explainable granularity exploration through an abstraction history and augmented logs. Evaluation on a real manufacturing dataset demonstrates substantial model-size reduction (from 1,489 to about 58 elements) and shows how analysts can navigate granularity levels, while noting limitations such as the requirement for perfect log fitness and a finite set of initial abstractions.

Abstract

Process events are recorded by multiple information systems at different granularity levels. Based on the resulting event logs, process models are discovered at different granularity levels, as well. Events stored at a fine-grained granularity level, for example, may hinder the discovered process model to be displayed due the high number of resulting model elements. The discovered process model of a real-world manufacturing process, for example, consists of 1,489 model elements and over 2,000 arcs. Existing process model abstraction techniques could help reducing the size of the model, but would disconnect it from the underlying event log. Existing event abstraction techniques do neither support the analysis of mixed granularity levels, nor interactive exploration of a suitable granularity level. To enable the exploration of discovered process models at different granularity levels, we propose INEXA, an interactive, explainable process model abstraction method that keeps the link to the event log. As a starting point, INEXA aggregates large process models to a "displayable" size, e.g., for the manufacturing use case to a process model with 58 model elements. Then, the process analyst can explore granularity levels interactively, while applied abstractions are automatically traced in the event log for explainability.

Figures (6)

• Figure 1: Two process models of a bank's account opening business process tsagkani_process_2022 abstracted with INEXA. The upper process model shows four transitions with fine-grained activity labels denoted in orange, while the lower process model exhibits only a single transition $t5$ as an aggregate of the four transitions with fine-grained activity labels.
• Figure 2: Workflow object type classes conceptualized with the aim of representing these artifacts based on russell2004workflowbuhner_working_2006reijers_ha_usefulness_2010smirnov_business_2012dumas_fundamentals_2013mangler_cpee_2014kumar_optimal_2013pika_mining_2017van_der_aalst_object-centric_2019van_der_aalst_discovering_2020turetken_influence_2020ghahfarokhi_ocel_2021ehrendorfer_assessing_2021park2022detectingadams_defining_2022fdhila_verifying_2022tsagkani_process_2022.
• Figure 3: Model abstraction object type classes classified by "abstraction target" based on smirnov_business_2012tsagkani_process_2022. All types in black in column "abstraction target" can be applied in INEXA. INEXA's initial abstraction repository (cf. \ref{['ssec:prelim']}) consists of the seven process model abstraction instances in column "abstraction repository" from coarse-grained to fine-grained granularity level.
• Figure 4: General idea for interactive, explainable model abstraction INEXA with separated presentation and domain layer for operations based on Domain-Driven Design evans2004domain. $pd$ is an object-centric process discovery technique, $abs$ a process model abstraction, $st_{abs}$ the corresponding augmented event log transition, and $overlay$ the projection of an augmented event log $L$ onto the original process model $AN_{original}$.
• Figure 5: INEXA's initialized process model of the manufacturing process in mangler_xes_2023. The initialized process model depicts the orchestration process for producing batches of chess turm pieces. Redoing the "Spawn Production" subprocess aggregation shows how the orchestration process spawns the actual production process that calls the respective machine programs to produce the parts.

Theorems & Definitions (3)

• definition thmcounterdefinition: Workflow, Model Abstraction and Abstraction History Object Type(s)
• definition thmcounterdefinition: Event Log
• definition thmcounterdefinition: Model Abstraction Repository and Admissibility