CARS-SVM预测哈密瓜可溶性固形物含量

采用近红外光谱技术结合数据降维的方法，建立了哈密瓜可溶性固形物含量的预测模型：对比多种光谱预处理方法，确定二阶求导用于处理原始光谱；经预处理的光谱数据分别结合特征选择竞争性自适应重加权采样法（CARS）、蒙特卡罗无信息变量消除法（MC-UVE）提取特征波长，以及利用主成分分析进行降维；再使用特征选择和特征提取的光谱数据作为模型的输入变量，建立哈密瓜可溶性固形物含量预测模型。结果显示，CARS+SVM建立的预测模型最优，模型的校正集相关系数为0.981 4，预测集相关系数为0.900 2，模型能够准确预测哈密瓜可溶性固形物含量。

Prediction of soluble solids in Hami melon by CARS-SVM

Soluble solid content is one of the important indexes for the internal quality analysis of Hami melon. In this study, the prediction model of soluble solid content of Hami melon was established by using near infrared spectroscopy combined with data dimension reduction method. Compared with a variety of spectral preprocessing methods, the second-order derivative was used to process the original spectrum; the preprocessed spectral data were combined with CARS and MC-UVE to extract the characteristic wavelength, and the principal component analysis was used to reduce the dimension; Finally, the spectral data of feature selection and feature extraction were used as the input variables of support vector machine to establish the prediction model of soluble solid content of Hami melon. The results showed that the prediction model established by CARS + SVM was the best, with the correlation coefficient of the model calibration of 0.981 4, and the correlation coefficient of the prediction set was 0.900 2. This model could be used to accurately predict the soluble solids of Hami melon.