GENERALIZED ENTROPY-BASED WEIGHTS FOR MULTIATTRIBUTE VALUE MODELS

The 'nearly-equal' weights procedure for additive multiattribute value models has been developed by Barron and Schmidt (1988) using the maximum entropy principle. The maximum entropy principle uses no other information except some constraints on the weights. When it is required that a set of weights closely approximate a set of given weights and satisfy some constraints, then the Minimum Discrimination Information (MDI) principle is the appropriate entropy-based allocation procedure. The MDI, also known as the minimum cross-entropy, reduces to the maximum entropy when the given weights are all equal.