Geographical Origin Identification of Panax notoginseng Using a Modified K-Nearest Neighbors Model With Near-Infrared Spectroscopy

This study introduces a novel method for identifying the geographical origins of Panax notoginseng using near-infrared spectroscopy and a modified K-nearest neighbors algorithm. The proposed model integrates distance and cosine similarity metrics to enhance classification performance, particularly for imbalanced datasets. Principal component analysis is employed to reduce dimensionality, significantly improving computational efficiency without sacrificing accuracy. Comparative analyses reveal that the modified K-nearest neighbors surpasses classic K-nearest neighbors and other models, achieving up to 96.90% accuracy on balanced datasets and 95.79% on imbalanced datasets. These results demonstrate the potential of combining spectral data processing with advanced machine learning techniques for efficient and accurate geographical origin identification, providing a robust tool for quality assurance and traceability in traditional Chinese medicine.