关键词: 加工周期预测, 半导体制造, 多层数据分析框架, 机器学习, 集成学习

Abstract: To predict cycle time accurately of semiconductor processing based on manufacturing data and to cover the shortage of traditional model's generalization,a multi-layer data analysis framework was proposed.A cycle time prediction algorithm based on this framework was realized.Based on the data of a semiconductor line,a prediction model was built,and the effectiveness of this method was validated by comparing with several common methods.Experiments demonstrated that the proposed multi-layer data analysis framework based cycle time prediction method could effectively improve the accuracy and the generalization.

Key words: cycle time prediction, semiconductor manufacturing, multi-layer data analysis framework, machine learning, ensemble learning