A dual-forecast scheme for production output with paired/unpaired WIP data

Data concerning of the IC work in process (WIP) can be classified into paired and unpaired, according to whether the corresponding input and output times of the WIP are correlated with each other. To handle these different classes of data, this study presents a production output forecast scheme that can predict the distributions in the next-period and simulate the production outputs of a semiconductor supply chain. This scheme yields the forecast support of the proposed dual-forecast models, such as empirical and grey predictive models, using a reliance index. Industrial cases studies show that the average accuracies of the scheme in forecasting unpaired WIP and paired WIP are 81% and 86%, respectively. Based on the scheme, a virtual production control system enables a fabless company to monitor its outsourcing, effectively.