Intelligent Multiple Attribute Decision System

Chua Hong Keong

Abstract

In this project, the feasibility of developing an intelligent Multiple Attributes Decision Making (MADM) system had been studied. In the process, a prototype system had been developed.

Because of its wide scope, many mathematical methodologies have been developed over the years to solve MADM problems. Naturally, some methods are more suitable and efficient than others in the solution of each particular decision problem. However, such knowledge is not readily available and this has deterred many potential users from utilizing these methods. From research, it was found that several methods were used for solving a decision problem throughout the entire process. Hence, the system should preferably contain sufficient number of methods for the decision makers' use during the decision making process.

To harness the potential of these methods effectively, the system must possess two important characteristics. Firstly, it should possess the flexibility to be modified so that new or revised methods could be introduced when the need arises. Secondly, it should be capable of guiding users possessing various levels of expertise on the usage of these MADM methods.

For this project, the 17 well-established methods compiled by Hwang & Yoon (1981) had been studied and 11 of them had been implemented in the prototype system. These were developed as independent executables to facilitate the flexibility required of the system. As these methods share similar data acquisition routines, these routines were developed as independent modules so that data acquired could be accessed by all the methods. This also contributes to the flexibility of the system.

For the second characteristic, Knowledge-Based Systems (KBS) have been utilized to provide guidance on the selection of suitable methods for MADM problems. To provide appropriate guidance for users possessing different levels of expertise, three modes of guidance had been incorporated, namely the advanced, intermediate and novice modes.

The advanced mode was designed for user who are familiar with the various methods. These users should be capable of selecting an appropriate method for their problems. The job of the KBS is to detect the existence of the necessary inputs and prompt the users for them if they do not exist.

The intermediate mode operates in the opposite manner. It will attempt to find methods corresponding to a set of inputs. This was designed for users who are not so familiar with the methods but do have access to the various inputs. They would like to know which are the methods that correspond to a set of inputs.

For users who are totally unfamiliar with MADM, the novice mode will prompt the user with questions regarding the decision problem, expected solutions and the users' preferences. From the answers obtained, suitable methods will be recommended.

In conclusion, this project had displayed the feasibility of building a decision support system that permits ease of integration; i.e., MADM methods can be added or removed without much difficulties. In addition, this project had demonstrated how KBS could be used to assist users in areas of scare expertise and facilitate operation of a flexible modular decision making system.