Abstrait

Estimating of gold recovery by using back propagation neural network and multiple linear regression methods in cyanide leaching process

Asghar Azizi, Seyyed Zioddin Shafaei, Reza Rooki, Ahmad Hasanzadeh, Mostafa Paymard

In this study, two techniques – back propagation neural network (BPNN) and multiple linear regression (MLR) were applied to estimate gold recovery in cyanide leaching process. The designed neural network has three layers including input layer (seven neurons), hidden layer (ten neurons) with tansing activation function and output layer (one neuron) with linear activation function. The comparison between the estimated recoveries and the measured data resulted in the correlation coefficients, R, 0.952 and 0.884 for training and test data using BPNN model. However, the R values were 0.786 and 0.767 for training and test data respectively, byMLRmethod. In addition, the root mean square (RMS) error obtained 1.08 and 1.22 for BPNN andMLRmethods, respectively. Finally, the results indicate that the BPNN can be used as a viable method to rapidly and cost-effectively estimate gold recovery in cyanide leaching solution.