Cost model development using virtual manufacturing and data mining: part II—comparison of data mining algorithms

Abstract

Listen

Translate article

Cost models of manufacturing processes are an important tool enabling enterprises to make reasonable predictions and forecasts in relation to the production costs for existing and new products. Accurate and robust cost models can help to provide significant competitive advantage for manufacturing organisations. Advanced computational methods such as virtual manufacturing and data mining have been identified as potentially powerful techniques for generating cost models that bypass the problems associated with traditional cost modelling processes. Part I, of this two-part paper, described the development of a cost model development methodology that makes use of virtual manufacturing models and data mining techniques and used case study data to validate this methodology. A critical part of this methodology is the selection and use of effective data analysis techniques that can identify accurate and robust cost estimating relationships. Part II now examines in detail the effectiveness of alternative data mining algorithms in terms of their ability to develop relationships that are (1) representative of the real causal relationships that exist and (2) able to provide a high level of estimating accuracy. More specifically, it focuses on the data generated by virtual manufacturing models and how the size and complexity of the generated data sets impact the accuracy of the cost estimating relationships.