Linear Programming Reformulation for Optimal Conformance Checking
A new paper on arXiv introduces a totally unimodular linear program (LP) for alignment-based conformance checking, reformulating the problem on the reachability graph of the synchronous product. The approach exploits network-flow structure to guarantee integral optimal solutions via LP relaxation, avoiding integer variables and branch-and-bound search. Empirical evaluation on over 2.1 million instances from real-world and synthetic benchmarks shows that the LP method complements the standard A*-based heuristic search, particularly for long traces or substantial deviations where A* exhibits exponential runtime.
Key facts
- The paper introduces a totally unimodular LP for alignment-based conformance checking.
- The reformulation is defined on the reachability graph of the synchronous product.
- It guarantees an integral optimal extreme-point solution through LP relaxation.
- The method avoids combinatorial overhead of integer variables and branch-and-bound.
- Empirical evaluation covers more than 2.1 million conformance checking instances.
- Datasets include real-world and synthetic benchmarks.
- The LP approach complements A*-based heuristic search.
- A* can exhibit exponential runtime for long traces or substantial deviations.
Entities
Institutions
- arXiv