Effects of Design Parameters in Cost Estimation and Cost Uncertainty Modeling

Case-based reasoning (CBR) is an approach which uses old cases/experiences to understand and solve new problems. In CBR, a previous case similar to the current case is used to generate a solution for the current case and usually involves adaptation of the generated solution to suit the current case. The CBR approach consists of creating a knowledge-base (or database) containing past cases (products), defining a new case, retrieving cases similar to the new case, and adjusting the solution (cost) of the retrieved cases to the new case. This paper compares CBR approach with regression analysis approach in studying the effects of varying design attribute specifications on cost estimation accuracy and cost distribution reliability. These approaches are compared and effects of defining a concept with varying design attribute specifications are studied by applying leave-one-out cross-validation to a knowledge-base of automobiles.