Abstract
Demand fluctuation has great influence on demand management in semiconductor foundry. It is important for demand prediction to be the foundation of utilization and material planning. Due to the migration properties of technological generations, decision makers are difficult to assess the long-term customer requirements via domain knowledge and experience rules only. This study aims to construct a manufacturing intelligence framework for developing a two-phase mask demand model based on long-term trend detection and short-term consists of general indicatory function acquisition and parametric fluctuation fine-tuning to facilitate decision makers for well demand management planning. Therefore, the special pattern and regular fluctuation of semiconductor mask demand variation can be detected through proposed twophase forecasting model respectively. An empirical study was conducted to validate this approach. The proposed model has smaller average forecast error rate than the existing method in two technology parts. The results demonstrated the practical viability of this approach.