The growth of the poultry industry in Colombia, and the change in current environmental regulations, leads to an improvement in the wastewater treatment system, through alternate techniques between coagulants and flocculants. The cost of these techniques require optimal dosing of the main products involved therein. In this work, we used an artificial neural model based on feedforward-backpropagation multilayer neural networks, to estimate the value of output turbidity in the treatment of wastewater. The trained neural networks subsequently used to propose optimal dosages of the products and improve the operating conditions that allow obtaining clarified wastewater, so we develop optimization charts. We used the R linear correlation factor as a performance indicator, for the evaluation of the performance of the neural model. The correlation results between the estimated a real values of the output turbidity show its reliability in the application as a prediction tool.