Improving the optimization efficiency and precision of least squares support vector regression (LSSVR) for pear property prediction

Yong Hao; Yande Liu; Hailiang Zhang; Xuemei Liu; Yuanyuan Pan

doi:10.1117/12.865737

11 October 2010 Improving the optimization efficiency and precision of least squares support vector regression (LSSVR) for pear property prediction

Yong Hao, Yande Liu, Hailiang Zhang, Xuemei Liu, Yuanyuan Pan

Author Affiliations +

Proceedings Volume 7656, 5th International Symposium on Advanced Optical Manufacturing and Testing Technologies: Optical Test and Measurement Technology and Equipment; 76566X (2010) https://doi.org/10.1117/12.865737
Event: 5th International Symposium on Advanced Optical Manufacturing and Testing Technologies, 2010, Dalian, China

Abstract

In this study, Visible/near-infrared (Vis/NIR) diffuse reflectance spectroscopy at 530-1560 nm region was investigated for the analysis of the soluble solids content (SSC) and color of pear. Least squares support vector regression (LSSVR) has been proven to be a powerful tool for modeling complex samples through the use of adapted kernel functions. However, one of the major drawbacks of LSSVR is that the optimization of the regularization and kernel meta-parameters is time-consuming during training the model, and the modeling results are sensitive to spectral noise. Wavelet compression pretreatment is an effective method for spectral information extraction and noise elimination. The calibration set was composed of 75 pear samples and 32 pear samples were used as the validation set. The raw and pretreated spectra by wavelet compression were modeled using LSSVR, It was shown that wavelet compression procedure not only shortened the modeling time, but also improved the predictive precision. The correlation coefficient (r) was improved from 0.78 to 0.93 for SSC, and from 0.95 to 0.96 for color, respectively. The root mean square error of prediction (RMSEP), optimization time and calibration variables were reduced from 0.68, 0.33s and 1031 to 0.41, 0.03s and 24 for SSC, while from 1.10, 0.33s and 1031 to 1.07, 0.03s and 40 for color. The results indicated that Vis/NIR spectroscopy combined with wavelet compression procedure and LSSVR is a reliable approach for predicting the SSC and color of pear.

Citation Download Citation

Yong Hao, Yande Liu, Hailiang Zhang, Xuemei Liu, and Yuanyuan Pan "Improving the optimization efficiency and precision of least squares support vector regression (LSSVR) for pear property prediction", Proc. SPIE 7656, 5th International Symposium on Advanced Optical Manufacturing and Testing Technologies: Optical Test and Measurement Technology and Equipment, 76566X (11 October 2010); https://doi.org/10.1117/12.865737

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available