Estimation of Qualities and Inference of Operating Conditions for Optimization of Wafer Fabrication Using Artificial Intelligent Methods

After selecting the method, data acquired from the target process is used in data mining. The collected data should be sufficient in number and clean enough to perform the data mining. Due to insufficient data for wafer’s quality in this research quality evaluation was performed according to sampling inspection, not total inspection. The lack of data causes an over-fitted model and incorrect rules. To overcome these problems, an appropriate data preprocessing of the bootstrap method was used to generate additional data sufficient for a total inspection. Improvement of model performance was observed from the results. In the following, the target process of the ingot fabrication process, the proposed road map for data mining, the applied data mining techniques, and the application to the experimental data are presented.