Knowledge assisted integer programming model formulation

Abstract

UNIK-OPT is a knowledge-assisted optimization model formulator. The first version of UNIK-OPT, UNIK-LP covers the linear programming models. As a next step, the scope of application is extended to include integer programming models, thus named UNIK-IP. However, the interpretation as well as formulation of the IP model is not an easy task at all, because the model implicitly has logical meaning among the variables and constraints. Since the transformation of logical relationship to IP formulation is very difficult to even IP experts, supporting the transformation process is very important. UNIK-IP allows model builders to represent IP models semantically at a high level and automatically transforms such a high level model to a solvable base level. Since the base level model can be solved by a Branch & Bound algorithm, all the transformable high level representations can be used to aid the formulation of the IP models with more intuitive perspective of problems themselves. Next concerns are a compression toward efficient IP model and a transformation of composite expression which is composed of multiple logical relationships. For that purpose, rules in the form of theorems and formulation reasoning strategies are developed.