When designing nanoparticles for drug delivery, many variables such as size, loading efficiency, and cytotoxicity should be considered. Usually, smaller particles are preferred in drug delivery because of longer blood circulation time and their ability to escape from immune system, whereas smaller nanoparticles often show increased toxicity. Determination of parameters which affect size of particles and factors such as loading efficiency and cytotoxicity could be very helpful in designing drug delivery systems. In this work, albumin (as a protein drug model)-loaded chitosan nanoparticles were prepared by polyelectrolyte complexation method. Simultaneously, effects of 4 independent variables including chitosan and albumin concentrations, pH, and reaction time were determined on 3 dependent variables (i.e., size, loading efficiency, and cytotoxicity) by artificial neural networks. Results showed that concentrations of initial materials are the most important factors which may affect the dependent variables. A drop in the concentrations decreases the size directly, but they simultaneously decrease loading efficiency and increase cytotoxicity. Therefore, an optimization of the independent variables is required to obtain the most useful preparation.