CAJ | 학술논문

在总结了现有的偏最小二乘回归方法的基础上，利用神经网络的非线性映射逼近能力，构造了一种新的定量分析算法。采用CO的IR光谱数据集对提出的算法进行了验证，同时与PLS、KPLS、PLSS算法进行了分析和比较。实验结果表明，该方法预测能力最优，而且在模型的复杂度和拟合优度方法，达到甚至优于现有算法，是一种有效的定量分析方法。
재총결료현유적편최소이승회귀방법적기출상，이용신경망락적비선성영사핍근능력，구조료일충신적정량분석산법。채용CO적IR광보수거집대제출적산법진행료험증，동시여PLS、KPLS、PLSS산법진행료분석화비교。실험결과표명，해방법예측능력최우，이차재모형적복잡도화의합우도방법，체도심지우우현유산법，시일충유효적정량분석방법。
Summarizing the existing partial least squares regression method, by using nonlinear mapping of neural network, a new quantitative analysis algorithms is constructed. IR spectral data sets of CO is used to verify the proposed algorithm, at the same time with the PLS, KPLS, PLSS algorithm analysis and comparison. The experimental results show that the new quantitative analysis algorithm is the best predictive capability, and in model complexity and goodness of fit, it reach or even better than existing algorithms, it is an effective method of quantitative analysis.