Data mining based dynamic scheduling approach for semiconductor manufacturing system

This paper presents a data mining based dynamic scheduling approach which responses to changing system status for semiconductor manufacturing system. The proposed approach, based on the historical data, applies genetic algorithm as feature selection tool to eliminate the redundant data and K-nearest neighbor algorithm as a classifier to select the scheduling rule. Finally, a scheduling strategy selection model is built to make real-time scheduling decision for semiconductor manufacturing system. Here, efficacy function and information entropy are used for the multi-objective scheduling strategy evaluation. This ensures the selected scheduling strategy optimizes various performance criteria at the same time. At last, an actual semiconductor production line is used to test the practicability and effectiveness of the proposed dynamic approach.