Yield forecasting is a critical task to every semiconductor manufacturer. However, the existing methods for yield forecasting often deal with the logarithmic or log-sigmoid value, rather than the original value, of yield. To resolve this problem, in this study, the fuzzy collaborative intelligence (FCI) method proposed by Chen and Lin (2008) is modified, so that it can consider the original value of yield directly. The modified FCI method is called the direct-solution (DS)-FCI approach. The effectiveness of the DS-FCI approach was validated with a real case. The experimental results showed that the DS-FCI approach outperformed Chen and Lin's FCI method in improving the forecasting accuracy and precision.
展开▼
