Quantitative structure-properties relationship (QSPR) has been applied to modeling and predicting the electrophoretic mobilities of a series of benzoic acid derivatives in different carrier electrolyte composition. Descriptors that were selected by stepwise multiple linear regression (MLR) technique are radial distribution function-lag8 (RDF-8), unweighted R-maximal autocorrelation geometry, topology and atomic weight assembly-lag4 (R-GETAWAY-4), geometrical descriptor lag-26 (GEO-26), and the overall dielectric constant of the carrier electrolyte. These descriptors were used as inputs for generated 4-7-1 artificial neural network (ANN). The results obtained using ANN and MLR were compared as well as with the experimental values and showed the superiority of ANN over MLR model. Also the appearance of these descriptors in QSPR models reveals the role of electronic and steric interactions in solutes mobility in capillary electrophoresis due to the dielectric and hydrodynamic friction forces.