LP2Graph Mines Optimization Models for Railway Rescheduling

The field of railway rescheduling, like many optimization-driven domains, heavily relies on Mixed-Integer Linear Programming (MILP). However, the vast body of modeling knowledge is fragmented across numerous academic papers, often presented in incompatible notations. Existing narrative surveys tend to classify models subjectively, focusing on vocabulary rather than underlying structure, and fail to reproduce the models themselves. To address this, the researchers present LP Mining with LP2Graph. This method systematically extracts the structural information from published LP and MILP formulations, organizing it into a reproducible dataset and an induced taxonomy. At its core, LP2Graph represents each formulation, admitted by its canonical grammar, as a typed variable-equation graph derived from a single canonical model. Once a source is processed into this model, all subsequent steps are deterministic. Each source is parsed, homologized, and then clustered both bottom-up (by variables, constraints, objective, and whole-model structure) and separately by application domain and solution approach. The resulting groups are automatically labeled using a rule-seeded, self-updating classifier. The representation is validated by regenerating per-cluster representatives in LaTeX and re-solving them against reported optima using various solvers. This process yields an objective, repeatable taxonomy of variables, constraints, and model types, forming a principled foundation for automated railway-rescheduling model development.