Outlier Sample Analysis on Near Infrared Spectroscopy Determination for Flour Ash

Author

Wu Shengnan ; Liu Cuiling ; Wu Jingzhu ; Sun Xiaorong ; Dong Xiuli

Author_Institution

Beijing Technol. & Bus. Univ., Beijing, China

fYear

2013

fDate

16-17 Jan. 2013

Firstpage

781

Lastpage

784

Abstract

The flour ash was determined by near infrared (NIR) apparatus. Mahalanobis distance method and Monte Carlo cross validation (MCCV) method for abnormal samples removed and using partial least squares (PLS) modeling, Eventually, Mahalanobis distance method for abnormal samples removed, when the weight factor of 1.5, excluding the number of samples is 3 to get the best results, the correlation coefficient (R2) is 92.67, cross validation mean square error (RMSECV) is 0.0485, MCCV method for abnormal samples removed, excluding the number of samples 3 to get the best results, R2 is 94.64, RMSECV is 0.0411. Mahalanobis distance method and MCCV method for abnormal samples removed have improved the accuracy of the model, and MCCV method for flour ash model is more stable and more to improve the accuracy.

Keywords

Monte Carlo methods; ash; food technology; infrared spectroscopy; mean square error methods; milling; spectrophotometers; testing; MCCV method; Mahalanobis distance method; Monte Carlo cross validation method; NIR apparatus; PLS modeling; RMSECV; abnormal samples; flour ash model; near infrared apparatus; near infrared spectroscopy determination; outlier sample analysis; partial least squares modeling; Accuracy; Analytical models; Ash; Calibration; Chemical analysis; Predictive models; Spectroscopy; MCCV; Mahalanobis distance; flour ash; near infrared spectroscopy; unusual samples;

fLanguage

English

Publisher

ieee

Conference_Titel

Measuring Technology and Mechatronics Automation (ICMTMA), 2013 Fifth International Conference on

Conference_Location

Hong Kong

Print_ISBN

978-1-4673-5652-7

Type

conf

DOI

10.1109/ICMTMA.2013.194

Filename

6493845