A graphical user interface for variable selection employing the Successive Projections Algorithm

Henrique Mohallem Paiva

doi:10.1016/j.chemolab.2012.05.014

A graphical user interface for variable selection employing the Successive Projections Algorithm

2012-08-01

46

PRE

AI

H

Henrique Mohallem Paiva

S

Sófacles Figueredo Carreiro Soares

R

Roberto Kawakami Harrop Galvão

M

Mário César Ugulino de Araújo *

DOI：10.1016/j.chemolab.2012.05.014

Original

Original request for help

Save

Abstract

En 中文

The present work describes a MATLAB graphical user interface for the Successive Projections Algorithm (SPA), which is a variable selection technique aimed at reducing collinearity problems in Multiple Linear Regression (MLR) modelling. The interface also offers the possibility of pre-processing the data using Savitzky-Golay smoothing/differentiation and/or wavelet denoising. Sample selection routines for dividing the samples into calibration and validation sets are also implemented. At the end, prediction statistics (PRESS, RMSEP, SDV, WAS and correlation coefficient) are calculated to evaluate the performance of the resulting MLR model. All these operations can be carried out by the user without the need for MATLAB programming skills. For illustration, an example involving near-infrared spectrometric determination of moisture in corn samples is presented. The software is freely available at www.ele.ita.br/similar to kawakami/spa. (C) 2012 Elsevier B.V. All rights reserved.

Keywords:

Multivariate calibration

Variable selection

Successive Projections Algorithm

Graphical user interface

MATLAB

A graphical user interface for variable selection employing the Successive Projections Algorithm

Abstract

Journal

Organization