Uninformative variable elimination

When the number of noisy (noninformative) variables is too large, PLS models may also supply rather poor predictive performance. In order to overcome such a matter, a number of techniques for the elimination of noisy variables or the selection of useful predictors have been deployed, such as iterative stepwise elimination (ISE), iterative predictor weighting (IPW), uninformative variable elimination (UVE), and Martens uncertainty test (MUT) (Forina et ah, 2007). 

More multivariate methods of variable selection, especially suited for PLS applied to spectral data, are currently available. Among them, we can cite Interactive Variable Selection, Uninformative Variable Elimination, Iterative Predictor Weighting PLS, and Interval PLS. ... 

Uninformative variable elimination (UVE) is a widely used method for variable selection in chemometrics [26]. Its extended version, Monte Carlo UVE (MCUVE), was recently proposed [27, 31]. Mimicking the principle of "survival of the fittest" in Darwin s evolution theory, we developed a variable selection method in our previous work, called competitive adaptive reweighted sampling (CARS) [8, 28, 32, 33], which was shown to have the potential to identify an optimal subset of variables that show high predictive performances. The source codes of CARS are freely available at [34, 35]. 

From table 2 we can see that PLSR method do get satisfied prediction results. However, PLSR method using full bands of the spectra for developing calibration model are time-consuming while running the computer program. Some variables in the full bands of samples spectra are effective while some are not. As such determining effective spectra from the fuU band spectra is very essential. A special algorithm, namely uninformative variable elimination (UVE) was used in this research to find out the effective variables. The variables with useless information were eliminated. 

Chen, B., and D. Chen. 2005. The ap>plication of uninformative variables elimination in near-infrared spectroscopy. Spectronic Instruments and Analysis. 04 26-30. 

After elimination of uninformative variables, a PLS model for the y and Xnew data is constructed, using the leave-one-out cross validation procedure to estimate its complexity. 

Elimination of Uninformative Variables for Multivariate Calibration, Anal. Chem., 68 (1996) 3851. 

---