Adaptive Multi-Entity Process Graph Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current process discovery algorithms are inadequate for multi-entity data models, as they are optimized for single-case event logs and fail to capture optimizations for interrelated entities, leading to suboptimal process graphs that may not suit the specific use case or data model.
Innovation Solution
A computer-implemented method that determines an optimized process graph by selecting a subset of entity instances that maximizes a utility function, balancing quality requirements such as readability, coverage, precision, and generalizability, using a parametrized utility function and a predefined order to adapt the search space to specific use cases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If process discovery algorithms are applied directly to multi-entity data models, then the algorithms can process the data, but they fail to capture optimizations for interrelated entities and produce suboptimal process graphs
Solution Approach 1:
The patent segments the multi-entity process discovery problem into multiple single-entity process discovery subproblems. Each entity type is processed independently to generate its own process graph, which is then merged into a comprehensive multi-entity process model. This segmentation allows standard algorithms to be applied effectively while still capturing inter-entity relationships through the merging step.
Solution Approach 2:
The patent introduces an intermediary merging step that combines multiple single-entity process graphs into a unified multi-entity process model. This intermediary component handles the complexity of inter-related entities by integrating information from multiple entity-specific process graphs, thereby improving the overall accuracy and reliability of the discovered process model.
2Adaptability or versatility
If a single multi-entity process graph is determined for a given data model, then the process graph provides one view onto the data model, but it may not be the best suited view to understand the captured behavior
Solution Approach 1:
The patent makes the process graph selection dynamic by introducing a utility function that can be parameterized according to different use cases. Instead of providing a single static process graph, the system allows users to adjust parameters of the utility function to generate different optimized process graphs suited for different purposes (e.g., readability vs. coverage vs. precision). This dynamic approach enables adaptability without requiring complex manual selection processes.
Solution Approach 2:
The patent employs parameter changes in the utility function to transform a single process graph into multiple optimized views. By varying the parameters of the utility function (such as weighting factors for different quality metrics), the system can generate different process graphs optimized for specific use cases. This approach provides flexibility in selecting appropriate views while keeping the underlying mechanism relatively simple.
3Quantity of substance
If the search space is expanded to include every entity in multi-entity data models, then more comprehensive process coverage is achieved, but the complexity of finding the optimal process graph increases significantly
Solution Approach 1:
The patent reduces optimization complexity by segmenting the comprehensive search space into smaller entity-specific search spaces. Instead of searching for the optimal process graph across all entities simultaneously (which would be computationally intractable), the system performs separate optimizations for each entity type and then merges the results. This segmentation makes the optimization problem manageable while still achieving comprehensive coverage.
Solution Approach 2:
The patent combines multiple simplified optimization problems (one for each entity type) into a comprehensive multi-entity process model through a merging step. This merging approach allows the system to achieve comprehensive entity coverage without directly solving the computationally complex problem of optimizing all entities simultaneously. The combination of segmented optimizations produces a globally optimized process graph.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a computer-implemented method to determine an adaptive optimized single-entity process graphs from an entity-centric data model, and an adaptive optimized multi-entity process graph based on the optimized single-entity process graphs.