Comparison of rank-based weighting methods for multi-criteria decision making

Abstract

In engineering and management science, decision makings are always influenced by several criteria. In general, degrees of contribution of each criterion to the decision, or the weights of criteria, are not equal. Among various weighting methods suggested in the decision literature, rank-based methods which convert a criteria ranking order into numerical weights have been claimed as a good compromise choice between ease of implementation and quality of the decision result. While previous studies compared quality of several rank-based methods through computer simulations, this study conducts an empirical experiment using actual decision data in order to validate theoretical conclusions. Four weighting methods are compared based upon the hit percentage and the rank-order conformity. The results support previous studies that rank order centroid (ROC) method tends to perform the best due to its steepness and non-linear function of the weights which are likely to be mostly consistent with decision makers’ behaviour.