부산대학교 도서관

The present paper presents a methodology for solving a given linear programming problem. We extend the idea for the location of the binding constraints from two to n dimensions and we implement it in order to solve such LP problems. The proposed methodology classifies the constraints into two classes. The first of them consists of the majority of the binding constraints and is utilized for solving the corresponding linear programming problem. The second class is addressed primarily for verifying the validity of the remaining constraints and secondarily for the fine-tuning of the solution using the rest of the binding constraints. We test the proposed methodology in a set of random linear programming problems with interesting results.