Prediction of Design Parameters in Ship Designing Based on Data Mining Technique by Combining Genetic Programming with Self Organizing Map

Engineering data contains the experiences and knowhow of experts. Data mining technique is useful to extract knowledge or information from the accumulated existing data. Although Korean shipyards have accumulated a great amount of data, they do not have appropriate tools to utilize the data in practical works. Most of shipyards utilize empirical formulas for the prediction of design parameters in preliminary design stage. But these formulas generated from past existing ship data are old-fashioned, so it is not adequate to apply the formulas to new ship design for the prediction of parameters. This paper presents a machine learning method based on genetic programming (GP), which can be one of the components for the realization of data mining. Differing from neural network model which is black box, user cannot perceive the predicted formula, GP can be seen. Users can use the generated formula in real design system. In practical cases, we don’t have enough learning samples, but there are many design input parameters. Therefore, reducing the number of input parameters is essential. In order to reducing the number of input parameters, self organizing map (SOM) is adopted. By applying SOM to accumulated data, the influence of input parameters on design outputs can be found. The developed data mining system by combining GP with SOM can be powerful tool for the generation of empirical formulas to predict design parameters in ship design.