Based on the Model Population Analysis concept,a near-infrared spectroscopy prediction model for wheat protein content was established using the Variable Proportional Selection Adaptive Iteration(VSPAI)method combined with Partial Least Squares(PLS).VSPAI employs the Monte Carlo sampling method to randomly select sample subsets and uses PLS to build regression models for these subsets.The mean and standard deviation of the regression coefficients of each variable are calculated to obtain their initial weights,which are then combined with the Weighted Bootstrap Sampling method to select the optimal feature variables.The results show that the prediction accuracy of the VSPAI-PLS model improved by 39.9%,5.9%,25.4%,and 46.9%compared to the Competitive Adaptive Reweighted Sampling combined with PLS,Variable Combination Population Analysis combined with PLS,Variable Combination Population Analysis Iteratively Retained Information Variables combined with PLS,and Bootstrapping Soft Shrinkage combined with PLS models,respec-tively.In conclusion,the application of the VSPAI-PLS model for the non-destructive testing of wheat protein using near-infrared spectroscopy is feasible.
关键词
小麦蛋白质/近红外光谱分析/模型集群分析/变量选择比自适应迭代
Key words
wheat protein/near-infrared spectroscopy analysis/model population analysis/variable proportional selection adaptive iteration
维普
万方数据