. 2013 Oct 14;13(10):13820–13834. doi: 10.3390/s131013820

Table 2.

Results of regression models for the quantification of Notoginseng powder (NP) adulterated by sophora flavescens powder (SFP), corn flour (CF), and the mixture of two adulterants using least-squares support vector machines (LS-SVM) algorithm based on the data of visible spectra, short-wave near infrared spectra (SNIR), and long-wave near infrared spectra (LNIR), respectively.

Adulterant

Spectral Range

Modeling Method

LVs

Calibration

Prediction

r_C

R_{C}^{2}

RMSEC (%)

r_P

R_{P}^{2}

RMSEP (%)

RPD

SFP

Visible

LS-SVM

0.971

0.943

1.693

0.932

0.846

2.778

2.670

PLSR

0.939

0.882

2.427

0.911

0.787

3.261

2.314

SNIR

LS-SVM

0.961

0.922

1.972

0.921

0.841

2.815

2.559

PLSR

0.867

0.751

3.528

0.874

0.762

3.451

2.055

LNIR

LS-SVM

0.984

0.967

1.276

0.917

0.840

2.829

2.501

PLSR

0.924

0.853

2.709

0.912

0.821

2.994

2.362

Visible

LS-SVM

0.950

0.897

2.269

0.845

0.710

3.809

1.858

PLSR

0.936

0.876

2.489

0.792

0.602

4.461

1.623

SNIR

LS-SVM

0.994

0.987

0.796

0.959

0.918

2.029

3.514

PLSR

0.851

0.724

3.716

0.821

0.670

4.062

1.746

LNIR

LS-SVM

0.986

0.973

1.166

0.930

0.864

2.609

2.724

PLSR

0.957

0.916

2.047

0.946

0.893

2.308

3.071

SFP&CF

Visible

LS-SVM

0.834

0.685

3.241

0.688

0.471

4.198

1.376

PLSR

0.548

0.300

4.830

0.574

0.327

4.735

1.220

SNIR

LS-SVM

0.996

0.991

0.555

0.786

0.560

3.830

1.577

PLSR

0.576

0.332

4.720

0.571

0.325

4.743

1.218

LNIR

LS-SVM

0.892

0.794

2.621

0.898

0.789

2.652

2.183

PLSR

0.887

0.787

2.665

0.871

0.754

2.862

2.033

LVs: Number of latent variables.