Analysis of Overlapping Voltammograms of Nitrophenols Combining Genetic Algorithms and Support Vector Machines

Abstract

This paper suggests a novel method named GA-LSSVM, combines genetic algorithms (GA) and least squares support vector machines (LS-SVM) techniques to provide a powerful model for improving the regression quality and to enhance the ability to extract characteristic information. Simultaneous differential pulse voltammetric multi-component determination of o-nitro phenol, m-nitro phenol and pnitrophenol was conducted for the first time by using the proposed method. The LS-SVM technique broadens the application of SVM by reducing the computational complexity since only the solution of a set of linear equations is required instead of a quadratic programming problem. Thus, LS-SVM has the capability of solving linear and nonlinear multivariate calibrations in a relatively fast way. Genetic algorithms (GA) introduced are probabilistic optimization techniques based on natural evolution and genetics and Darwin's theory of survival of the best. The GA-LS-SVM method is proven to be successful even when severe overlap of voltammograms existed.