摘要
偏最小二乘是一个在近红外光谱解析中常用的计量学算法,结合变量筛选方法既可以提高模型的预测能力,也可以大大降低建模的难度。本文将前向区间偏最小二乘用于烟煤水分近红外光谱解析。提取出的区间数为2,变量个数从1557减少到54个。所提取的波长区间主要位于O-H一级泛频吸收带。预测平均绝对百分误差从0.0865降低到0.0818。研究结果表明,前向区间偏最小二乘可以显著减少变量数并提高预测准确度。
Partial least squares method is a widely used method in near- infrared spectra analysis. When combined with feature selection technique,it can highly improve the predictive ability of the model and reduce its complexity.Interval partial least squares were applied forward to determine bituminous coal moisture with near- infrared spectra. Two intervals were selected which were lied in O- H first universal frequency absorption band. Furthermore,the number of variables reduced from 1557 to 54. Prediction mean absolute percent error reduced from 0. 0856 to 0. 0818.
出处
《广州化工》
CAS
2016年第4期26-28,共3页
GuangZhou Chemical Industry
基金
云南省省级大学生创新创业训练计划项目(编号:201310664003)
云南省教育厅一般项目(编号:2012Y414)
曲靖师范学院招标项目(编号:2011ZB006)
关键词
烟煤
水分
偏最小二乘
区间选择
bituminous coal
moisture
partial least squares
interval selection
