A Simulation Based Comparative Study of Normalization Procedures in Multiattribute Decision Making SUBRATA CHAKRABORTY Student Member, IEEE Clayton School of Information Technology, Monash University, Clayton, Victoria 3800. AUSTRALIA. CHUNG-HSING YEH Senior Member, IEEE Clayton School of Information Technology, Monash University, Clayton, Victoria 3800. AUSTRALIA. Abstract: – Normalization procedures are required in multiattribute decision making (MADM) to transform performance ratings with different data measurement units in a decision matrix into a compatible unit. MADM methods generally use one particular normalization procedure without considering the suitability of other available procedures. This study compares four commonly known normalization procedures in terms of their ranking consistency and overall preference value consistency when used with the most widely used simple additive weight method. To achieve this, new performance measure indices are introduced and new simulation settings are devised for dealing with various measurement settings. A wide range of MADM problems with various measurement scales are generated by simulation for the comparison study. The experiment result shows that vector normalization and linear scale transformation (the max method) outperforms other normalization procedures when used with SAW. Key-Words: - MADM, SAW, Normalization, Decision making, Decision support systems, Simulation study, Method comparison, Decision consistency. 1 Introduction Multiattribute decision making (MADM) problems involve ranking or evaluating a finite number of alternatives with multiple, often conflicting, attributes. Various MADM methods have been developed to solve different problem settings. MADM methods have shown their suitability to particular decision problems. The sheer complexity in MADM problems and the need for a timely solution prompts to choose any MADM method suitable to a problem on the basis of experience, knowledge and intuition. With quite a few available methods, it is extremely difficult to choose the best method that meets all the requirements. Several comparative studies on MADM methods have shown that certain methods are more suitable for specific decision settings as compared to other methods [2][7][12][13][15]. In MADM problems, each alternative has a performance rating for each attribute, which represents the characteristics of the alternative. It is common that performance ratings for different attribute are measured by different units. To transform performance ratings into a compatible measurement unit, normalization procedures are used. MADM methods often use one normalization procedure to achieve compatibility between different measurement units. For example, SAW uses linear scale transformation (max method) [1][3][5][6][13][14], TOPSIS uses vector normalization procedure [13][14][16], ELECTRE uses vector normalization [4][14] and AHP uses linear scale transformation (sum method) [8][9][10][14]. Enormous efforts have been made to comparative studies of MADM methods, but no significant study is conducted on the suitability of normalization procedures used in those MADM methods. This leaves the effectiveness of various MADM methods in doubt and certainly raises the necessity to examine the effects of various normalization procedures on decision outcome when used with given MADM methods. Proceedings of the 6th WSEAS Int. Conf. on Artificial Intelligence, Knowledge Engineering and Data Bases, Corfu Island, Greece, February 16-19, 2007 102