Abstract: In reality,the weights of attributes(e.g.indices of a product) often depend on the requirement of the customer.Hency the customer preference information is important for weight identifyiing.Such information may be not direct expressed as importance attached to attributes,but implied in their preference on a set of given alternatives.Along with the attribute evaluation,there also maybe exists uncertainty of ignorance.For this kind of circumstance,an attribute weight estimating method based on mining the latent customer requirement is proposed.The theory of evidence combination and goal programming is employed in the procedure.An illustration example is given.

Key words: weight estimation, evidence combination, goal programming