닥나무 인피섬유의 제조 지역 식별을 위한 적외선 스펙트럼 데이터 전처리 및 기계학습 모델링

The objective of this study was exploring the impact of spectral data preprocessing techniques on the performance of machine learning models for classifying the origin of mulberry bast fibers. The findings indicated that a selective spectral region (1800-1200 cm-1) significantly improves classification model performance. Among the classifiers tested, Partial Least Squares Discriminant Analysis (PLS-DA) and Support Vector Machines (SVM) demonstrated the highest accuracy. Additionally, A spectral preprocessing with the Norris-Williams algorithm effectively improved model performance within the same classifier for this dataset. These results suggest that applying machine learning modeling with spectral preprocessing can enable the origin classification of mulberry bast fibers and provide a chemical basis for classification rules beyond simple categorization. KCI Citation Count: 0