Aggregating multiple expert data for linear case valuation models using the MDE principle

摘要

The Maximum Decisional Efficiency Principle is a recently developed technique for parameter estimation based on actual expert decisions. This paper illustrates how the method may be used to aggregate multiple expert data such as for developing expert system logic for binary classification problems. Previously the method has only been applied to derive a scoring function of Leontief type. In this paper the method is applied to the more important problem of deriving a linear scoring function. Hence the results here are more directly comparable to Linear Discriminant Analysis. Also an example is developed to illustrate the performance of the method when experts disagree on one or more cases to be classified.