Scholarlink

<p>Accurate gemstone classification is critical in gemology for authentication and identification. This study presents a novel one-class classifier using self-organizing maps (SOMs), combined with Monte Carlo permutation, for classifying gemstones based on Raman spectroscopy. The model was optimized through the map size and iterations, and its classification reliability was assessed using cumulative distribution values from Monte Carlo permutation. Raman spectra from the RRUFF database and experimental garnet gemstones were used to evaluate the model performances. The classifier demonstrated high accuracy in distinguishing gemstone types and identifying external validation set with minimal false positives. The one-class SOM achieved superior classification with 100% accuracy in training and strong performance in validation. The proposed method enhances prediction robustness and is especially suited to scenarios with limited reference data. This combination of one-class SOMs with Monte Carlo permutation testing represents a unique approach, enhancing prediction reliability and robustness in gemstone classification.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

One class self-organizing maps with Monte Carlo permutation for gemstone classification using Raman spectroscopy

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月4日

查看全文
  • Nontawat Sricharoen,
  • Sanong Ekgasit,
  • Pimthong Thongnopkun,
  • Kanet Wongravee

摘要

Accurate gemstone classification is critical in gemology for authentication and identification. This study presents a novel one-class classifier using self-organizing maps (SOMs), combined with Monte Carlo permutation, for classifying gemstones based on Raman spectroscopy. The model was optimized through the map size and iterations, and its classification reliability was assessed using cumulative distribution values from Monte Carlo permutation. Raman spectra from the RRUFF database and experimental garnet gemstones were used to evaluate the model performances. The classifier demonstrated high accuracy in distinguishing gemstone types and identifying external validation set with minimal false positives. The one-class SOM achieved superior classification with 100% accuracy in training and strong performance in validation. The proposed method enhances prediction robustness and is especially suited to scenarios with limited reference data. This combination of one-class SOMs with Monte Carlo permutation testing represents a unique approach, enhancing prediction reliability and robustness in gemstone classification.