The multivariate modeling of adsorptive removal of arsenic from aqueous solution by a calcined Mg-Fe-(CO3) layer double hydroxide, synthesized by a co-precipitation method at a low supersaturation, was conducted by an artificial neural network (ANN). The major influencing parameters of the adsorption process, i.e., adsorbent dose (0.25–4 g L−1), reaction time (2–240 min), pH of the solution (3–12) and agitation rate (80–220 rpm) were varied through a ‘one variable at a time’ (OVAT) experiment to assess their individual effect on the arsenic removal efficiency. The OVAT experimental data were used for multivariate modeling through a feed forward ANN network with back propagation algorithm. The optimized network showed a correlation coefficient for the training, validation, testing and overall process above 0.99 and the mean square of error as 0.996. The analysis of variance conducted on the predicted values from the model and the actual experimental value exhibited a high F value of and low p value less than 0.001, which showed the applicability of ANN model in delineating the adsorption process of arsenic removal.