基于LASSO算法的光谱变量选择方法研究

Spectral Variable Selection Methods Based on LASSO Algorithm

光谱分析技术由于具有简单、快速、无损等优势,在复杂体系的定性和定量分析中得到了广泛应用。然而光谱中往往包含成百上千的波长点,有些波长点与研究的目标性质并不相关,加大了计算量并降低了模型的预测准确度。因此,在建立模型前需要进行变量选择。最小绝对收缩与选择算子(LASSO)可将回归系数收缩为0,进而达到变量选择的目的。该研究将LASSO用于三元调和油样品近红外光谱和生物样品拉曼光谱的变量选择,基于偏最小二乘(PLS)和多元线性回归(MLR)模型,分别对香油和肌氨酸的含量进行定量分析,并与无信息变量消除-PLS(UVE-PLS)、蒙特卡罗结合UVE-PLS(MCUVE-PLS)和随机检验-PLS(RT-PLS)3种变量选择方法进行比较。结果表明,基于LASSO的变量选择方法保留的变量数最少,运算速度最快。对三元调和油样品,LASSO-PLS预测的准确度最高;对生物样品,LASSO-MLR预测的准确度最高。因此,基于LASSO的变量选择算法有望在光谱分析领域中得到良好应用。

Spectral analysis technology has been widely used in the qualitative and quantitative analysis of complex systems due to its simple,fast,speed and non-destructive.However,spectra often contain hundreds or thousands of wavelengths(variables).Some of them may not be related to the research object,which will lead to a large amount of calculation and reduce the prediction accuracy of model.Therefore,it is necessary to select the variables before establishing the model.The least absolute shrinkage and selection operator(LASSO)could be used to shrink the regression coefficients to zero,so as to achieve the purpose of variable selection.In this study,LASSO was used to select the variables for near infrared(NIR)spectra of ternary blend oil and Raman spectra of bio-fluid samples.The contents of sesame oil and sarcosine were quantitatively analyzed by building partial least squares(PLS)and multiple linear regression(MLR)models.The methods are compared with the three variable selection methods,i.e.uninformative variable elimination-PLS(UVE-PLS),Monte Carlo uninformative variable elimination-PLS(MCUVE-PLS)and random test-PLS(RT-PLS).The result shows that the variable selection methods based on LASSO retain the least variable numbers and the fastest calculation speed.LASSO-PLS shows the highest prediction accuracy for ternary blend oil samples,while LASSO-MLR shows the highest prediction accuracy for bio-fluid samples.Therefore, the variable selection methods based on LASSO are expected to be well applied in the field of spectral analysis.