Techniques for ranking simple fuzzy numbers are abundant in the litera
ture. However, we lack efficient methods for comparing complex fuzzy n
umbers that are induced by arithmetic operations. In this paper a prob
abilistic approach is taken. Membership functions of fuzzy numbers are
first converted into probability density functions. The Mellin transf
orm is then introduced to compute the mean and the variance of a compl
ex fuzzy number. The fuzzy number with the higher mean is ranked highe
r. If the means are equal, the one with the smaller variance is judged
higher rank. Two numerical examples and a fuzzy multiple attribute de
cision analysis are illustrated.