An Improved AdaBoost Tree-Based Method for Defective Products Identification in Wafer Test

Wafer test is one of the most significant parts of semiconductor fabrication. Effective data mining technologies will improve wafer prediction performance, which will contribute to production cycle time reductions, yield enhancement and final product quality improvement. This paper presents an improved AdaBoost Tree-based method to decrease the false fail rate of defective products in wafer test. Particularly, it focuses on a discussion about the low classification accuracy for the minority class of samples due to the imbalance of wafer test data set. In addition, the novel method is proposed to identify defective products in wafer test, which achieves higher separability between the majority class and the minority class as well as shorter test time. Finally, experimental studies on wafer test data set show that our algorithm has better performance than traditional methods.