Skoči na glavni sadržaj

Izvorni znanstveni članak

DATA MINING METHODOLOGY FOR DETERMINING THE OPTIMAL MODEL OF COST PREDICTION IN SHIP INTERIM PRODUCT ASSEMBLY

Damir Kolich orcid id orcid.org/0000-0002-3287-682X ; University of Rijeka-Faculty of Engineering, Vukovarska 58, 51000 Rijeka, Croatia
Nikša Fafandjel ; University of Rijeka-Faculty of Engineering, Vukovarska 58, 51000 Rijeka, Croatia
Y. Lawrence Yao ; Columbia University, 220 Mudd building, MC 473, 500 West 120th Street, New York, NY 10027 USA


Puni tekst: engleski pdf 1.226 Kb

str. 1-18

preuzimanja: 573

citiraj


Sažetak

In order to accurately predict costs of the thousands of interim products that are assembled in shipyards, it is necessary to use skilled engineers to develop detailed Gantt charts for each interim product separately which takes many hours. It is helpful to develop a prediction tool to estimate the cost of interim products accurately and quickly without the need for skilled engineers. This will drive down shipyard costs and improve competitiveness. Data mining is used extensively for developing prediction models in other industries. Since ships consist of thousands of interim products, it is logical to develop a data mining methodology for a shipyard or any other manufacturing industry where interim products are produced. The methodology involves analysis of existing interim products and data collection. Pre-processing and principal component analysis is done to make the data “user-friendly” for later prediction processing and the development of both accurate and robust models. The support vector machine is demonstrated as the better model when there are a lower number of tuples. However as the number of tuples is increased to over 10000, then the artificial neural network model is recommended.

Ključne riječi

data mining; pre-processing; principal component analysis; support vector machine regression; artificial neural network; shipbuilding

Hrčak ID:

154713

URI

https://hrcak.srce.hr/154713

Datum izdavanja:

31.3.2016.

Posjeta: 1.472 *