Data Mining Approach to Dual Response Optimization

Abstract

In manufacturing process optimization, analyzing a large volume of operational data is getting attention due to the development of data processing techniques. One of important issues in the process optimization is a simultaneous optimization of mean and variance of a response variable. It is called dual response optimization (DRO). Traditional DRO methods build statistical models for the mean and variance of the response variable by fitting the models to experimental data. Then, an optimal setting of input variables is obtained by analyzing the fitted models. This model based approach assumes that the statistical model is fitted well to the data. However, it is often difficult to satisfy this assumption when dealing with a large volume of operational data from manufacturing line. In such a case, data mining approach is an attractive alternative. We proposes a particular data mining method by modifying patient rule induction method for DRO. The proposed method obtains an optimal setting of the input variables directly from the operational data where mean and variance are optimized. We explain a detailed procedure of the proposed method with case examples.