MICROCOMPUTER-BASED MULTIPLE CRITERIA DECISION MAKING : IMPLEMENTATION, COMPARATIVE EVALUATION, AND A HYBRID APPROACH

Abstract

An approach to comparative study of multicriteria decision making algorithms is presented in this thesis. The Analytic Hierarchy Process, a well known technique, is the proposed methodology for the comparative study. Fourteen multicriteria decision making algorithms are compared using the proposed methodology. Implementability, Computational Efficiency, Usability and Solution process are the major factors considered. These major factors are further divided into thirteen sub-criteria, organized in a hierarchical structure. Inputs to the Analytic Hierarchy Process have been collected from the computational experiments conducted in a laboratory environment. Relative rankings of all the algorithms with respect to all the criteria are then produced. For the purpose of the laboratory experiment, the algorithms are implemented on microcomputer. A package called MOLP-PC is developed using "Turbo-Pascal" which makes use of a customized and proprietary spread sheet on the front-end to enhance its user friendliness and man-machine interface. Based on the experience gathered in the comparative study, a new, practical, man-machine interactive algorithm for solving multiple objective linear programs is proposed. The new method is a hybrid of three well known methods from the literature. It thus possesses the good attributes of these methods and avoids the weaknesses of other methods. The method makes use of the two-person zero sum game with mixed strategies approach to arrive at an initial solution. The decision maker then gives his or her level of relaxation for one or more objective functions and the method proceeds to modify the feasible region in order to arrive at the next solution. The cycle is repeated until a satisfactory solution is found. A numerical example is presented to illustrate the algorithm. The proposed method has been implemented on microcomputer in Pascal.