Comparison of Retention Modeling in Ion Chromatography by Using Multiple Linear Regression and Artificial Neural Networks

The aim of this work is comparison of the prediction power of multiple linear regression and artificial neural networks retention models for inorganic anions (fluoride, chloride, nitrite, sulfate, bromide, nitrate, and phosphate) in suppressed ion chromatography with isocratic elution. Relations between ion chromatographic parameters (eluent flow rate and concentration of OH − in eluent) and retention time of particular anion are described with unique mathematical function obtained by multiple linear regression and with a three-layers feed-forward artificial neural network. The artificial neural network was trained with a Levenberg-Marquardt batch error back propagation algorithm. It is shown that the multiple linear regression retention model has lower, but still very satisfactory, predictive ability. Due to its complexity, the artificial neural network must still be regarded as a more complicated technique. That indicates multiple linear regression as a method of choice for retention modeling in the case of ion chromatographic analysis with isocratic elution.