Table 1 Calibration, cross-validation and prediction results of the PPC, TVB-N values and sensory score of rainbow-trout samples by hyperspectral imaging system.

From: Prediction of various freshness indicators in fish fillets by one multispectral imaging system

Model

n

LVs

Calibration

Cross-validation

Prediction

RDP

R2C(adj)

RSMEC

R2CV(adj)

RSMECV

R2P(adj)

RSMEP

Bias

PPC (Log 10 CFU/g)

PLSR

9

6

0.923

0.504

0.901

0.548

0.899

0.579

0.077

3.17

MLR

9

—

0.923

0.522

0.908

0.551

0.918

0.533

0.106

3.43

LS-SVM

9

—

0.920

0.515

0.904

0.505

0.917

0.517

0.103

3.55

BP-ANN

9

—

0.922

0.508

909

0.550

0.921

0.504

0.128

3.64

TVB-N (mg N/100 g)

PLSR

9

5

0.889

3.084

0.874

3.304

0.857

3.585

0.257

2.645

MLR

9

—

0.892

3.164

0.870

3.355

0.855

3.593

0.174

2.640

LS-SVM

9

—

0.889

3.101

0.872

3.327

0.862

3.542

−0.208

2.678

BP-ANN

9

—

0.881

3.307

0.862

3.526

0.853

3.643

0.073

2.603

Sensory score (6–30)

PLSR

9

4

0.927

1.572

0.919

1.604

0.902

1.987

−1.009

3.024

MLR

9

—

0.930

1.544

0.918

1.609

0.909

1.991

−0.983

3.018

LS-SVM

9

—

0.928

1.523

0.921

1.599

0.912

1.802

−0.996

3.335

BP-ANN

9

—

0.920

1.597

0.913

1.664

0.910

1.848

−1.001

3.251

  1. LV: latent variable; R2C(adj): adjusted determination coefficient of calibration; R2CV (adj): adjusted determination coefficient of cross-validation; R2P(adj): adjusted determination coefficient of prediction; RMSEC: root-mean-square errors estimated by calibration; RMSECV: root-mean-square errors estimated by cross-validation; RMSEP: root-mean-square errors estimated by prediction; MLR: Multi-linear regression; PLSR: partial least squares regression; LS-SVM: least squares support vector machine BP-ANN: back-propagation artificial neural network.
Back to article page