Selection of variables for interpreting multivariate gas sensor data

In this work, methods to select relevant variables from a large set of avai lable gas sensor parameters were examined. Two data sets, containing a larg e number of variables, were studied. The objective was to find the best des criptors, which could predict interesting properties of the measurements. U sing a forward selection procedure, applying the root mean square error fro m a multilinear regression model as the selection criterion, it was possibl e to get good prediction accuracy from a backpropagation neural network (AN N). This procedure was fast, compared to the usual trial and error variable selection, objective, and can be made fully automated. In addition, the us e of principal component analysis (PCA) and partial least squares (PLS) sco re vectors used as descriptors were examined. The ANN models constructed wi th either the PCA or the PLS score vectors as input gave for the first, rat her smooth, data set, errors of the same size or smaller than for the forwa rd selected parameters. For the second data set, containing many noisy vari ables, the forward selected parameters outperformed the other two data redu ction methods. (C) 1999 Elsevier Science B.V. All rights reserved.