This chapter discusses artificial computational intelligence methods as applied to cost prediction. We present the development of a suite of hybrid fuzzy-neural systems for predicting the cost of performing engine tests at NASA’s Stennis Space Center testing facilities. The system is composed of several adaptive network-based fuzzy inference systems (ANFIS), with or without neural subsystems. The output produced by each system in the suite is a rough order of magnitude (ROM) cost estimate for performing the engine test. Basic systems predict cost based solely on raw test data, whereas others use preprocessing of these data, such as principal components and locally linear embedding (LLE), before entering the fuzzy engines. Backpropagation neural networks and radial basis functions networks (RBFNs) are also used to aid in the cost prediction by merging the costs estimated by several ANFIS into a final cost estimate.
Kaminsky, E., H. Danker-McDermott, and F. Douglas, "Fuzzy-Neural Cost Estimation for Engine Tests", Chapter 9 in Computational Economics: A Perspective from Computational Intelligence, Idea Group Publishing (Hershey, PA), ISBN: 1-59140-649-8, 2006, pp. 178-204.