免疫算法在近红外光谱奇异样本识别中的应用

为了提高近红外光谱数据建模后的准确性,文中提出基于免疫算法的近红外光谱奇异样本的识别方法.通过免疫算法与遗传算法对同一近红外光谱数据集分别进行奇异样本识别并比较,删除奇异样本后,免疫算法较遗传算法分别将水分、脂肪、蛋白质的PLS模型的预测误差平方和分别降低了25.8%、32.1%、21.7%.实验表明,免疫算法适用于近红外光谱奇异样本的识别,提高了模型预测精准度和稳健性.

Immune Algorithm for Identification of Singular Sample with Near Inf rared Spectroscopy

In order to improve the accuracy of the modeling of near infrared spectral data ,this paper presents a method for identifing the singular sample with near infrared spectroscopy by the immune algorithm .The immune algorithm and genetic algorithm are used respectively to identify the singular sample in the same NIR spectral data sets .The comparison between the results obtained by the two methods shows that ,with the singular sample removed ,the immune algorithm increases PRZSS of the PLS models of water ,fat and protein by 25 .8% ,32 .1% and 21 .7% respectively .The experimental results show that ,the artificial immune algorithm is not only suitable for the identification of the singular sample with near infrared spectra ,but also can improve the prediction accuracy and robustness .