The present case study work is carried out on a typical steel industry in selection of the optimum amount of chemicals dosage used in DAF unit to remove oil and grease. Neural models are developed between DAF unit variables, i.e., the amount of chemical, the pH, and the amount of oil in the effluent. The amount of alum and polyelectrolyte are considered as independent whereas pH and amount of oil present in outlet stream are accounted as dependent variables. These input and output data are analyzed using ANN and GMDH method, which reduces the cost of the chemical/day-shift by 10.5% and 21.85%, respectively.