摘要
本文通过讨论了无信息变量消除法(uninformative variables elimination,UVE)的原理,并用此算法对玉米的近红外光谱数据进行波长变量选择,再使用偏最小二乘法(partial least squares,PLS)建立模型。结果表明,与使用全谱数据建立的模型相比较,筛选变量后建立的校正模型不仅简化了,而且增强了预测能力。
In the article, the theory of uninformative variables elimination (UVE) is presented and is used to select the wavelength variables of near-infrared spectroscopy (NIR) of corn samples, then the calibration model and the validation model are built. The result shows that, comparing with the model built with the whole spectra data, the model built with the selected variables not only simplifies and optimizes the calibration model but also has stronger predicted ability.
关键词
无信息变量消除法
偏最小二乘法
变量筛选
玉米
uninformative variables elimination （UVE）
partial least squares （PLS）
variables selection
corn
