农机化研究Issue(4)：46-48,56,4.
基于 NIR 的小麦粉异常样本剔除方法研究
Study on Elimination of Abnormal Wheat Powder Samples Based on NIR
摘要
Abstract
As abnormal samples ’ disturbances have great impact on building analysis model in process of NIR quantita-tive analysis , identification and treatment of abnormal samples is an important step in improving the predictive ability of the model .This research selects 150 flour samples from GuChuan Flourmill , using Mahalanobis distance method and the Monte Carlo Cross-validation method ( MCCV ) to distinguish and exclude abnormal samples , then using RMSECV and RMSEP as evaluation indexes to evaluate quantitative analysis models .The results indicate that prediction accuracy has improved after excluding abnormal samples by Mahalanobis distance method and MCCV .Meanwhile , we selected a more suitable method to improve the accuracy and reliability of quantitative analysis model by comparing two methods .关键词
小麦粉/异常样本/近红外光谱/马氏距离法/蒙特卡罗交叉验证法Key words
wheat/abnormal samples/NIR/Mahalanobis distance method/Monte Carlo Cross-validation method ( MC-CV )分类
农业科技引用本文复制引用
刘翠玲,孙晓荣,吴静珠,吴胜男,苗雨晴..基于 NIR 的小麦粉异常样本剔除方法研究[J].农机化研究,2014,(4):46-48,56,4.基金项目
北京市教委科研计划重点项目（ KZ201310011012）；北京市优秀人才资助项目（2012D005003000007）；北京市自然科学基金项目 （）
