An Agent-Based Fuzzy Collaborative Intelligence Approach for Precise and Accurate Semiconductor Yield Forecasting

Abstract

Yield forecasting is an important task for the manufacturer of semiconductors. Owing to the uncertainty in yield learning, it is, however, often difficult to make precise and accurate yield forecasts. To solve this problem, we propose an agent-based fuzzy collaborative intelligence approach that is modified from the fuzzy linear regression and back propagation network approach. In the proposed methodology, software agents rather than domain experts are used to improve the efficiency of collaboration. In addition, an agent decides the adjustable parameters by referencing to others so that the overall prediction performance can be improved in an effective way. In addition, we proposed a simple and effective way to aggregate the fuzzy forecasts by agents. Compared with the fuzzy linear regression and back propagation network approach, the proposed methodology reduced the average range and mean absolute percentage error by 18% and 99%, respectively.