A hybrid due-date fulfilled forecasting based on clustering and decision trees

In wafer fabrication, due date fulfilling is very important. How to timely delivering the goods to customer is very important. It will enhance customer service and competitive advantage. But the complexities of production system and re-entrant processes in wafer fabrication have raised the challenge for due date management. How to predict due date related performance is very important. In this research, a predict model of due date related performance based on decision tree has been modeled. Due date fulfilling performance will be classified to EE (extremely earliness), NE (Normal earliness), OT (On time), NL (Normal lateness), EL (Extremely lateness) based on the decision tree trained by a great quantity of historical data. Furthermore, different combination of order releasing rules (UNIF, TB, and WR) and dispatching rules (FIFO, EDD, CR, SRPT, and COVERT) have been considered in the simulation tests. In the result of simulation and statistics, average rate of correctly classifying is more than 90%. Especially in the combination of TB*SRPT, UNIF*COVERT, and WR*COVERT, the rate are larger than 95%. Managers of the wafer fabrication plants can improve the decision quality for due date management and shop floor control by prediction the due date related performance.