An Application of Data Mining Algorithms for Shipbuilding Cost Estimation
Bohdan L. Kaluzny, Sorin Barbici, G. M. van den Berg, Renzo Chiomento, Dimitrios Derpanis, Ulf Jönsson, R. H. A. David Shaw, Marcel C. Smit, Franck Ramaroson · Journal of Cost Analysis and Parametrics · 2011
This article presents a novel application of known data mining algorithms to the problem of estimating the cost of ship development and construction. The work is a product of North Atlantic Treaty Organization Research and Technology Organization Systems Analysis and Studies 076 Task Group “NATO Independent Cost Estimating and its Role in Capability Portfolio Analysis.” In a blind, ex post exercise, the Task Group set out to estimate the cost of a class of Netherlands' amphibious assault ships, and then compare the estimates to the actual costs (the Netherlands Royal Navy withheld the actual ship costs until the exercise was completed). Two cost estimating approaches were taken: parametric analysis and costing by analogy. For the parametric approach, the M5 system (a combination of decision trees and linear regression models) of Quinlan (1992) Quinlan, J. 1992. “Learning with continuous classes”. In Proceedings AI'92 Singapore: World Scientific Edited by: Adams, A. and Sterling, L. 343–348. [Google Scholar] for learning models that predict numeric values was employed. Agglomerative hierarchical cluster analysis and non-linear optimization was used for a cost estimation by analogy approach void of subjectivity.