Robust linear discriminant analysis for chemical pattern recognition

Linear discriminant analysis (LDA) is an effective tool in multivariate mul tigroup data analysis. A standard technique for LDA is to project the data from a high-dimensional space onto a perceivable subspace such that the dat a can be separated by visual inspection. The criterion of LDA, unfortunatel y, is extremely susceptible to outliers which commonly occur because of ins trument drift and gross errors. This paper proposes a robust discriminant c riterion, and based on that criterion, a high-breakdown method for LDA is d eveloped. In an effort to circumvent the local optima trapping, a real gene tic algorithm (RGA) was used for the optimization of the criterion. The RGA is capable of locating the global optimal solution with high probability a nd acceptable computational burden. Classification of one simulated data se t and two real chemical ones shows that the developed robust LDA (RLDA) met hod provides much superior performance to the standard method for outlier-c ontaminated data and behaves comparably well with the standard one for data without outliers. Copyright (C) 1999 John Wiley & Sons, Ltd.