A phenomenological model was used to describe sequentially the three steps (flowthrough, washing and elution) of expanded bed adsorption chromatography for recovery of chitosanases from Bacillus cereus. Additionally, a hybrid strategy for model parameter estimation was carried out using particle swarm optimization and Gauss-Newton algorithms. The model was validated with independent experimental data and the statistical criteria (χ2 and mean squared error tests) showed that the hybrid strategy was more promising than just the heuristic method. With the calibrated model, surface response methodology was applied to obtain the optimal operational conditions, and experiments were performed to confirm these results. Overall, a value of 41.08% for yield was obtained using 700mM NaCl during elution. In summary, all approach employed in this work was relevant for maximizing the yield of the chromatographic process.