Application of Artificial Neural Network Modeling in Prediction of The Extraction Yield of Copper-Morin Complex from Aqueous Media Utilizing a Molecularly Imprinted Polymer Coated Stir Bar

In this research, a new modeling method based on three-layer artificial neural network (ANN) technique was applied to predict the extraction yield of copper-morin complex from aqueous samples by means of molecularly imprinted stir bar sorptive extraction. Input variables of the model were pH of the solution, absorption and desorption times, stirring rate, temperature, and amount of morin ligand; while the output was extraction yield of copper ions. It was found that a network with 12 hidden neurons is highly accurate in predicting extraction recovery of copper-morin complex. The mean squared error and correlation coefficient between the experimental data and the ANN predictions were achieved as 0.0009 and 0.9999 for training, 0.0032 and 0.976 for validation and 0.0030 and 0.96666 for testing data sets. Under the optimum conditions, the linear range found to be in the range of 5-1000 μg L-1 with the detection limit of 0.38 μg L-1. The relative standard deviation was obtained to be below 5.3%. The method was successfully applied for preconcentration and determination of Cu in a few real samples.