Abstract
In semiconductor industry, demand fluctuation has great influence on demand management. Demand prediction is necessary to be the foundation of utilization and material planning. Demand variation may be largely caused by prosperity and changed production strategy. Indeed, more complexity of the number of transistors on a square centimeter of silicon and high speed of new technology development because of advanced products requirements increased. It leads to long-term technology life cycle with increasing customer orders. This research aims to develop a demand forecasting framework that consists of general indicatory function acquisition and parametric fluctuation fine-tuning in semiconductor foundry to forecast mask demand. It consists of a probability model for modeling technology life cycle and time domain approach for demand variation detection. An empirical study was conducted to validate developed demand forecasting framework. The result showed proposed forecasting framework can promote forecasting accuracy and provide a general mask demand forecasting model under the same problem definition to reduce decision uncertainty. It can provide valuable information for managers to support their decisions for demand forecasting and capacity planning.