The knowledge of migration behaviour is important in order to find the optimal experimental conditions for analyte determination by electromigration techniques, e.g. capillary zone electrophoresis (CZE). The metal ion analysis by means of CZE was possible as a consequence of application metal complexes with various organic and inorganic ligands. The metal complexes are responsible for metal ion speciation in solution and their mobility in environment. Recently, the systematic analysis of relationships between migration parameters and charge and size characteristics of metal complexes was done [1]. This quantitative structure-mobility relationships (QSMR's) are based on a function of electrophoretic mobility using enlarged Stokes-Einstein diffusion model with structural descriptors (metal atom electronegativity or effective charge) as well as the formal charge and ligand number [1]. The migration parameters are also dependent on experimental conditions (ionic strength, temperature, etc.). In this contribution, the method of migration behaviour prediction for different metal organic and inorganic complexes using "soft" modelling with artificial neural networks (ANN's) was examined on data taken from literature [1]. The achieved results were compared with values obtained by "hard" modelling (QSMR's). The proposed methodology allows to predict migration parameters and is not dependent on applied relationship. This alternative model-free approach can be used in practice for optimisation metal ion analysis by means of CZE.