Research Article: An Efficient Elastic Net with Regression Coefficients Method for Variable Selection of Spectrum Data

Date Published: February 2, 2017

Publisher: Public Library of Science

Author(s): Wenya Liu, Qi Li, Fengfeng Zhou.

http://doi.org/10.1371/journal.pone.0171122

Abstract

Using the spectrum data for quality prediction always suffers from noise and colinearity, so variable selection method plays an important role to deal with spectrum data. An efficient elastic net with regression coefficients method (Enet-BETA) is proposed to select the significant variables of the spectrum data in this paper. The proposed Enet-BETA method can not only select important variables to make the quality easy to interpret, but also can improve the stability and feasibility of the built model. Enet-BETA method is not prone to overfitting because of the reduction of redundant variables realized by elastic net method. Hypothesis testing is used to further simplify the model and provide a better insight into the nature of process. The experimental results prove that the proposed Enet-BETA method outperforms the other methods in terms of prediction performance and model interpretation.

Partial Text

Spectrum data is always used for quality prediction of important product or prediction of solution concentrations which are hard to measure in real industry process, especially in chemical processes. Near-infrared (NIR) spectroscopy, as a non-destructive, rapid and reliable analytical technique, has been widely used in many industry processes. However, NIR spectrum data always suffers from background variation, noise and colinearity[1]. A mass of data with hundreds of predictors is collected with many redundant variables contained, and those redundant variables contain more noise than quality-related information. Adding too many redundant variables into the regression model can lower the prediction accuracy, so variable selection plays an important role to deal with spectrum data. By identifying the key variables, variable selection can improve the prediction performance of the built model, reduce the model complexity and computation load, and provide a better insight into the nature of the process.

In this section, four variable selection methods are briefly introduced as follows.

In this paper, an Enet-BETA method has been proposed to build a stable and accuracy regression model via variable selection. This method can not only select important variables to make the response easy to interpret, but also can improve the stability and feasibility of the built model. Then two case studies are given to demonstrate the effectiveness of proposed method by comparing with the other four variable selection methods. Meanwhile, Enet-BETA method reflects the advantage of shrinkage methods.

 

Source:

http://doi.org/10.1371/journal.pone.0171122

 

0 0 vote
Article Rating
Subscribe
Notify of
guest
0 Comments
Inline Feedbacks
View all comments