Mid Sweden University

Several methods have been developed over a number of years for solving decision problems when vague and numerically imprecise information prevails. However, the DELTA method and similar methods give rise to particular bilinear programming problems that are time consuming to solve in a real-time environment. This paper presents a set of benchmark tests for non-linear programming solvers for solving this special type of problems. With two existing linear programming based algorithms, it also investigates the performance of linear programming solvers for special decision situations in decision support systems.

Computational decision analysis methods (such as DELTA) have been developed and implemented over a number of years for solving decision problems when vague and numerically imprecise information prevails. However, the evaluation phases in those methods often give rise to bilinear programming problems, which are time-consuming to solve in an interactive environment with general nonlinear programming solvers such as SNOPT. This paper proposes a linear programming based algorithm for solving this type of problems, thus enabling interactive use of such methods.

Several computational decision analysis approaches have been developed over a number of years for solving decision problems when vague and numerically imprecise information prevails. However, the evaluation phases in the DELTA method and similar methods often give rise to special bilinear programming problems, which are time-consuming to solve in an interactive environment with general nonlinear programming solvers. This paper proposes a linear programming based global optimization algorithm that combines the cutting plane method together with the lower bound information for solving this type of problems. The central theme is to identify the global optimum as early as possible in order to save additional computational efforts.

Computational decision analysis methods, such as the DELTA method, have been developed and implemented over a number of years for solving decision problems where vague and numerically imprecise information prevails. However, the evaluation phases in those methods often give rise to bilinear programming problems, which are time-consuming to solve in an interactive environment with general nonlinear programming solvers. This paper proposes a linear programming based algorithm that combines a cutting plane method with the lower bounding technique for solving this type of problem. The central theme is to identify the global optimum as early as possible in order to avoid generating unnecessary cuts in the convergent cutting plane procedure.