스펙트럼 분석을 통한 구기자 원산지 분류모형 연구
A Study on Classification Model of Boxthron Origin
0
조회
0
다운로드
1
인용
0
참고문헌
키워드
초록
Objectives: The origin of medicinal herbs is crucial to quality control. We aimed to find statistical model to discriminate the origin of boxthron. Methods: We used three types of methods for choosing meaningful variables that to predict the origin of boxthron:standardization, binning, variable selection. Firstly we used LRPC(logistic regression model with principal components)and PLSR(partial least square regression) model to discriminate the origin of spectrum pattern, and then figured chemical substance out. Results: We compared prediction power of alternative models according to three factors: variable selection, standardization and modeling. Variable selection method was worse than modeling using all variables. An error rate of the first type of standardization showed the best results among three standardization types. PLSR lower rate of error compared to LRPC. Optimized model, through LOOCV(leave and out cross validation), predicted the time that find out Boxthron origin, and also differentiated the orogin of Boxthron by Glycerol (14.3 seconds) and malic acid (23.18 seconds). Conclusions: We identified and suggested statistical model to discriminate the origin of boxthron by simulation. However,further research is needed to enhance the rate of prediction.
Abstract
Objectives: The origin of medicinal herbs is crucial to quality control. We aimed to find statistical model to discriminate the origin of boxthron. Methods: We used three types of methods for choosing meaningful variables that to predict the origin of boxthron:standardization, binning, variable selection. Firstly we used LRPC(logistic regression model with principal components)and PLSR(partial least square regression) model to discriminate the origin of spectrum pattern, and then figured chemical substance out. Results: We compared prediction power of alternative models according to three factors: variable selection, standardization and modeling. Variable selection method was worse than modeling using all variables. An error rate of the first type of standardization showed the best results among three standardization types. PLSR lower rate of error compared to LRPC. Optimized model, through LOOCV(leave and out cross validation), predicted the time that find out Boxthron origin, and also differentiated the orogin of Boxthron by Glycerol (14.3 seconds) and malic acid (23.18 seconds). Conclusions: We identified and suggested statistical model to discriminate the origin of boxthron by simulation. However,further research is needed to enhance the rate of prediction.
참고문헌 (19)
- Kwon SW, "A Study on Discriminating Methods of Medicinal Herbs by Applying the Plant Metabolomics", ~ (2009)
- Kho MS, "Introductory Instrumental Analysis" (2010), Yuhan Yu Han Publishing Co
- MS, "A Multivariate Model for Identification of Bacteria Using Pyrolysis -GC", pp.2001, Urbina ND
- Wold, S. Rube, "The collinearity problem in linear regression, the partial least squares (PLS) approach to generalized iInverses", SIAM Journal on Scientific and Statistical Computing, 15, pp.735~743 (1984)
- Otto M, "Spectrophotometric Multicomponoent Analysis Applied to Trace Element Determinations", Analytical Chemistry, 57, pp.63~69 (1985)
- Kim SW, "Case study papers: Partial least squares regression theory and application in spectroscopic diagnosis of total hemoglobin in whole blood", The Korean Journal of Applied Statistics, 10, pp.227~239 (1997)
- Rosipal R, "Overview and Recent Advances in Partial Least Squares", Subspace, Latent Structure and Feature Selection Techniques, 34~51 (2006)
- Beattie JR, "Effect of signal intensity normalization on the multivariate analysis of spectral data in complex ‘real-world’ datasets", Journal of Raman Spectroscopy, 40(4), pp.429~435 (2009)
- Bio-Rad Laboratories, Inc, "Search Strategies for IR Spectra- Normalization and Euclidean Distance vs. First Derivative Algorithm", ~ (2008)
- Hastie T, "The Elements of Statistical Learning" (2008), Springer
- Cho HW, "Genetic algorithm-based feature selection in high-resolution NMR spectra", EXPERT SYSTEMS WITH APPLICATIONS, 35(3), pp.967~975 (2008), PERGAMON-ELSEVIER SCIENCE LTD
- "Controlling the false discovery rate for feature selection in high-resolution NMR Spectra"[URL]
- Hemmateenejad B, "A comparative study between PCR and PLS in simultaneous spectrophotometric determination on of diphenylamine, aniline, and phenol: Effect of wavelength selection", Spectrochimica Acta Part A, 67, pp.958~965 (2007)
- Wentzell PD, "Comparison of principal components regression and partial least squares regression through generic simulations of complex mixtures", Chemometrics and Intelligent Laboratory Systems, 65, pp.257~279 (2003)
- Van Der Voet H, "Comparing the predictive accuracy of models using a simple randomization test", Chemometrics and Intelligent Laboratory Systems, 25, pp.313~323 (1994)
- 정우호, "단계별 순서를 응용한 첫 일자리에서의 조기퇴직에 대한 영향력 분석 -2009년 대졸자 이동경로조사로부터", 한국데이터정보과학회지, 21(6), pp.1191~1201 (2010), 한국데이터정보과학회
- Shtatland ES, "Stepwise methods in using SAS PROC LOGISTIC and SAS ENTERPRISE MINER for prediction", SUGI’28 Proceeding (2003), SAS Institute Inc.
- Faber K, "Propagation of measurement errors for the validation of predictions obtained by principal component regression and partial least squares", Journal of Chemometrics, 11, pp.181~238 (1997)
- Naes T, "Comparison of prediction methods for multicollinear data", Commun. Stat., Simul. Comput, 14, pp.545~576 (1985)