Object-Centric Data Model Taxonomies for Process Mining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object-centric data models are unstructured, making type definitions difficult to reuse and leading to inconsistent event and object types across different source systems, limiting the ability of process mining software to exploit knowledge within type definitions.
Innovation Solution
A computer-implemented method generates an object-centric data model using taxonomies to classify object and event types, allowing for hierarchical relationships and decoupling type definitions from taxonomies, enabling efficient data cleaning and analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If traditional unstructured type definitions using mere strings are used, then implementation is simple, but reuse of type definitions is difficult and data cleaning is complicated
Solution Approach 1:
The patent transforms type definitions from unstructured strings to structured data models with defined attributes and relationships. This parameter change enables type definitions to be reused across different organizations and source systems while maintaining consistency, resolving the contradiction between implementation simplicity and reusability.
Solution Approach 2:
The patent creates a universal type definition framework that can accommodate multiple organizations and source systems through standardized structures. This universal approach allows the same type definition schema to serve multiple purposes across different contexts, improving reusability without significantly complicating implementation.
2Device complexity
If traditional unstructured type definitions are used, then data model creation is straightforward, but data cleaning and establishing relations between different source systems is complicated
Solution Approach 1:
The patent performs preliminary structuring of type definitions during the data model creation phase, establishing standardized attributes and relationships before data extraction and cleaning. This preliminary action reduces the time and effort required for subsequent data cleaning and integration tasks across different source systems.
3Device complexity
If instance-level relations are used in traditional data models, then implementation is simple, but type-level knowledge cannot be exploited
Solution Approach 1:
The patent adds a type-level dimension to the data model by introducing explicit type definitions with attributes and relationships that exist independently of instances. This dimensional addition allows the model to capture and exploit type definition knowledge while maintaining instance-level relations, resolving the contradiction between model simplicity and knowledge exploitation.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided is a computer-implemented method to generate an object-centric data model of recorded process data, wherein each object is assigned to exactly one object type and each event is assigned to exactly one event type. The method comprises providing at least one first taxonomy, wherein each object type comprises a first pointer to at least one class per each first taxonomy. Preferably, each event type similarly comprises a second pointer to at least one class per each second taxonomy. Each class enforces the presence of a number of attributes for each corresponding type. Due to their hierarchical structure, taxonomies allow for an aggregation at type level. Further, taxonomies enable a reuse of types, since the respective type definitions are abstracted to at least one class per taxonomy.