Application of machine learning methods to cost estimation of product life cycle

Industry competitiveness depends on the cost, performance, and timely delivery of the product. Thus, an accurate, rapid, and robust product cost estimation model for the entire product life cycle is essential. This research applies two machine learning methods - back-propagation neural networks (BPNs) and least squares support vector machines (LS-SVMs) - to solve product life cycle cost estimation problems. The performance of a number of cost estimation models, statistical regression analyses, BPNs and LS-SVMs, are compared in terms of their performance. The estimation results and performance reveal that both LS-SVMs and BPNs can both provide more accurate performance than conventional models. We also make the first attempt to combine LS-SVMs and data transformation (DT) techniques to solve the outlier problem in the cost database. The cost estimation of airframe structure manufacturing was used to evaluate the feasibility of this novel combining mechanism. Finally, a more accurate, available, and generalisable cost estimation model is presented. This research can serve as a useful reference for product cost planning and control in industries.