Main Branch Decision Tree Algorithm for Yield Enhancement with Class Imbalance
Continuous yield enhancement is crucial for semiconductor companies that are driven by Moore’s Law for technology migration. This study aims to develop a manufacturing intelligence approach for trouble shooting and low yield correlation with imbalanced classes. In particular, Engineering Data Analysis (EDA) system is an off-line analysis system that can be used for trouble-shooting and yield enhancement. While making trouble shooting in semiconductor industry, engineers can not only use the expert knowledge on physics or electronics to answer the problem because of numerous relevant analysis factors. However, little research has been done on analyzing class imbalance problem to extract useful rules for yield enhancement. This study proposed a main branch decision tree (MBDT) algorithm that modifies the criteria of tree growth rather than using accuracy-based methods. An empirical study was conducted in a semiconductor company to evaluate validity. The results provide references for engineers to quickly identify the assignable causes and diagnose the problem of low yield.
Keywordsmanufacturing intelligence yield enhancement main branch decision tree algorithm class imbalance semiconductor manufacturing
Unable to display preview. Download preview PDF.
- 1.Berry, M., Linoff, G.: Data Mining Techniques for Marketing, Sales and Customer Support, 2nd edn. John Wiley and Sons, New York (2004)Google Scholar
- 3.Chiang, C., Kawa, J.: Design for Manufacturability and Yield for Nano-scale CMOS. Springer, The Netherlands (2007)Google Scholar
- 7.Chien, C.-F., Hsu, C., Hsiao, C.: Manufacturing intelligence to forecast and reduce semiconductor cycle time. J. Intell. Manuf. (2011), doi:10.1007/s10845-011-0572-yGoogle Scholar
- 11.Kotsiantis, S.B., Pintelas, P.E.: Mixture of expert agents for handling imbalanced data sets. Ann. Math Comput. Teleinformatics 1(1), 46–55 (2003)Google Scholar