基于近红外光谱结合机器学习的鳕鱼品种二分类方法研究

Research on cod species binary classification method based on near infrared spectroscopy and machine learning

中文关键词:  鳕鱼品种分类  近红外光谱  机器学习  支持向量机

英文关键词:classification of cod species  near infrared spectroscopy  machine learning  support vector machine

中文摘要:

      目的 探索适合分析鳕鱼近红外光谱数据的机器学习模型, 实现鳕鱼品种的快速二分类。方法 选取挪威大西洋真鳕、冰岛黑线鳕等8种鳕鱼, 对其研磨物进行傅里叶变换近红外光谱测试, 并采用最小-最大标准(min-max, Min-Max)归一化和独立成分分析法对近红外光谱数据进行预处理和降维, 进一步分别使用9种机器学习模型进行二分类, 通过6项指标对比各个模型的预测效果, 从中选出最适合鳕鱼二分类的模型。结果 本研究提出的独立成分分析法结合支持向量机的鳕鱼品种二分类模型的预测准确率可达到97.2%, F1分数可达到97.3%, 召回率达到99.4%。结论 本研究可实现较为准确的大西洋鳕鱼和非大西洋鳕鱼品种的分类, 为鳕鱼品种鉴别提供了方法依据。

英文摘要:

      Objective To rapidly carry out the binary classification of cod species, and explore the most suitable machine learning model for analyzing cod near-infrared spectroscopy data. Methods Eight kinds of cod, including Norwegian Atlantic cod and Icelandic haddock were selected for Fourier transform near infrared spectroscopy test on their ground products, and the near infrared spectroscopy data was preprocessed by Min-Max normalization and independent component analysis, and dimensionality reduction, and further used 9 kinds of machine learning models for binary classification, compared the prediction effects of each model through 6 indicators, and select the most suitable model for the two-classification of model. Results The prediction accuracy of the two-classification model of cod fish species proposed by the independent composition analysis method combined with support vector machine can reach 97.2%, the F1 score could reach 97.3%, and the recall rate could reach 99.4%. Conclusion This study can achieve a more accurate classification of Atlantic cod and non-Atlantic cod species, which provides a method basis for the identification of cod species.