Systematic Analysis of Feature and Classifier Configurations for Wafer Defect Pattern Recognition
IEEE Access · 2026 · 14 · 114286-114299
DOI: 10.1109/access.2026.3717267 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Taehyeon Kim und 2 weiteren Autor:innen. Die Arbeit nennt 46 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Institute of Information and Communications Technology Planning and Evaluation (IITP) through the Information Technology Research Center, Korean Government.
Abstract (Originalsprache)
Wafer defect pattern recognition is an important task for semiconductor manufacturing yield management. This paper presents a systematic evaluation of how feature domains, normalization strategies, and classifier choices impact classification performance, based on 42 experimental configurations. Regional, geometric, and Radon-based features are examined using support vector machine (SVM), decision tree (DT), and random forest (RF), with and without normalization. The experimental analysis shows that performance variations are primarily driven by feature domain selection and its interaction with the underlying classifier. Combining multiple feature domains improves classification accuracy by up to 38% compared with single-feature configurations, while RF achieves the highest overall accuracy of 0.91. Normalization substantially improves SVM performance, increasing the F1-score of the near-full defect class by up to 0.78, whereas DT and RF show only minor changes. These findings demonstrate that classification performance depends strongly on the interaction between feature configurations and classifier characteristics. Our study provides practical guidance for selecting effective feature-classifier configurations in wafer DPR systems.
Förderung
- Institute of Information and Communications Technology Planning and Evaluation (IITP) through the Information Technology Research Center
- Korean Government
Zitieren
APA
Kim, T., Go, B. C., & Kim, H. (2026). Systematic Analysis of Feature and Classifier Configurations for Wafer Defect Pattern Recognition. IEEE Access, 14, 114286-114299. https://doi.org/10.1109/access.2026.3717267
Vancouver
Kim T, Go BC, Kim H. Systematic Analysis of Feature and Classifier Configurations for Wafer Defect Pattern Recognition. IEEE Access. 2026;14:114286-114299. doi:10.1109/access.2026.3717267
BibTeX
@article{Kim2026,
author = {Kim, Taehyeon and Go, Byeong Chan and Kim, Hwajung},
title = {Systematic Analysis of Feature and Classifier Configurations for Wafer Defect Pattern Recognition},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {114286--114299},
doi = {10.1109/access.2026.3717267}
}
EUVIMED