基于介电频谱灵武长枣维生素C含量预测方法的研究

发布时间：2019/11/4 15:45:22     浏览次数：1129

基于介电频谱灵武长枣维生素C含量预测方法的研究
  
DOI：
中文关键词: 灵武长枣  维生素C  介电频谱  特征频率  预测模型

英文关键词: Lingwu long jujube  vitamin C  dielectric spectrum  characteristic frequency  predictive model

   为探究基于介电频谱预测灵武长枣维生素C含量的可行性，寻找最佳预测模型，利用LCR测试仪测试1 kHz～1 MHz条件下采后10 d内长枣介电谱变化规律；以CARS法、GA法、SPA法、UVE法筛选特征频率点并建立PLS、LSSVM长枣VC含量预测模型。结果表明，UVE算法为最佳频谱筛选方法，筛选特征频率点38个；UVE-PLS为最佳预测模型，其Rc、RMSEC、Rp、RMSEP值分别为0.9871、3.9322、0.9460、4.0400，验证模型预测值与实测值无显著差异（P＞0.05），说明基于介电频谱预测灵武长枣VC含量的方法是可行的。

 In order to explore the feasibility of predicting vitamin C content of Lingwu long jujube based on dielectric spectrum and find the best predictive model. Dielectric spectrum of Changzao were tested and analysed under of 1kHz~1MHz using the LCR tester. The characteristic frequency points were screened out by Competitive Adaptive Reweighed Sampling， Genetic Algorithm， Successive Projection Aalgorithm， and Uninformative Variable Elimination and the predictive model of Vitamin C content Least square support vector machine and Partial Least Squares were established combining the characteristic frequencies and VC value. The results show that the UVE algorithm is the best method to screen spectrum and 38 characteristic frequency points were screened. UVE-PLS is the best predictive model， the Rc， RMSEC， Rp， RMSEP values were 0.9871， 3.9322， 0.9460， 4.0400 and no significant difference （P>0.05） between measured and predicted values of validation model. It is feasible to predict the content of VC in long jujube based on dielectric spectrum.