为优选小麦粉蛋白质近红外建模校正集,在传统K/S(Kennard/Stone)方法划分的初始校正集基础上采用二进制蜻蜓算法(binary dragonfly algorithm,BDA)挑选代表性样品,建立小麦粉蛋白质含量偏最小二乘回归(partial least square regression,...为优选小麦粉蛋白质近红外建模校正集,在传统K/S(Kennard/Stone)方法划分的初始校正集基础上采用二进制蜻蜓算法(binary dragonfly algorithm,BDA)挑选代表性样品,建立小麦粉蛋白质含量偏最小二乘回归(partial least square regression,PLSR)模型,并用预测集检验评估模型的稳定性及预测性能。结果表明:BDA挑选出的最佳校正集样品数量为30个,所建模型的预测决定系数(R_(p)^(2))为0.9564,预测标准偏差(root mean square errors of prediction,RMSEP)为0.2781,与传统K/S划分的100个初始校正集的建模效果(R_(p)^(2):0.9388,RMSEP:0.3294)相比,R_(p)^(2)提高了1.87%,RMSEP降低了15.57%。10次BDA实验优选出校正集的平均数量为30.2个,且所建10个模型蛋白质含量预测效果均优于初始校正集建模。综上,BDA算法可以优选出数量少、具有代表性的校正集样品,建立的小麦粉蛋白质PLSR模型稳定性好、预测精度高,可为小麦粉品质近红外检测分析提供一种高效的校正集优选方法。展开更多
With 94 spring wheat cultivars as experimental materials, the correlations between the content of resistant starch （RS） in uncooked flour and cooked flour, and the apparent amylose content （AAC）, protein, lipid we...With 94 spring wheat cultivars as experimental materials, the correlations between the content of resistant starch （RS） in uncooked flour and cooked flour, and the apparent amylose content （AAC）, protein, lipid were investigated. The results showed that RS contents in both the uncooked flour and cooked flour assumed significantly positive correlation with AAC, and significantly negative cor- relation with protein content; and they were proved to be not significantly correlated with lipid content. RS content in uncooked flour was significantly correlated with that in cooked flour. These results provided references for the genetic improvement of wheat cultivars.展开更多
文摘为优选小麦粉蛋白质近红外建模校正集,在传统K/S(Kennard/Stone)方法划分的初始校正集基础上采用二进制蜻蜓算法(binary dragonfly algorithm,BDA)挑选代表性样品,建立小麦粉蛋白质含量偏最小二乘回归(partial least square regression,PLSR)模型,并用预测集检验评估模型的稳定性及预测性能。结果表明:BDA挑选出的最佳校正集样品数量为30个,所建模型的预测决定系数(R_(p)^(2))为0.9564,预测标准偏差(root mean square errors of prediction,RMSEP)为0.2781,与传统K/S划分的100个初始校正集的建模效果(R_(p)^(2):0.9388,RMSEP:0.3294)相比,R_(p)^(2)提高了1.87%,RMSEP降低了15.57%。10次BDA实验优选出校正集的平均数量为30.2个,且所建10个模型蛋白质含量预测效果均优于初始校正集建模。综上,BDA算法可以优选出数量少、具有代表性的校正集样品,建立的小麦粉蛋白质PLSR模型稳定性好、预测精度高,可为小麦粉品质近红外检测分析提供一种高效的校正集优选方法。
文摘With 94 spring wheat cultivars as experimental materials, the correlations between the content of resistant starch （RS） in uncooked flour and cooked flour, and the apparent amylose content （AAC）, protein, lipid were investigated. The results showed that RS contents in both the uncooked flour and cooked flour assumed significantly positive correlation with AAC, and significantly negative cor- relation with protein content; and they were proved to be not significantly correlated with lipid content. RS content in uncooked flour was significantly correlated with that in cooked flour. These results provided references for the genetic improvement of wheat cultivars.
