APPLICATION NEURAL NETWORK APPROACH FOR THE ESTIMATION OF HEAVY METAL CONCENTRATIONS IN THE INAOUEN WATERSHED

This paper describes how the multilayer perceptron neural network (MLPNN) trained by Broyden-Fletcher-Goldfarb-Shanno (BFGS) quasi-newton back-propagation approach was used to estimate heavy metal concentrations: Aluminum (Al), Lead (Pb), Copper (Cu), and Iron (Fe), in province of Taza using sixteen physicochemical factors measured from 100 samples collected surface water sources our team, according protocol national office (ONE). We chose a with only one hidden layer identify architecture employ. The number neurons varied, as were types transfer activation functions, BFGS learning method used. following statistical metrics evaluate performance network’s stochastic models: Examining adjustment graphs residue, well Error Sum Squares (SSE); mean bias error (MBE) determination coefficient (R²). results reveal that predictive models created artificial (ANN) are quite efficient, thanks algorithm’s efficiency speed convergence. An architectural [16-8-1] (16: variables input layer, 8: 1: output layer) produced best results,{R²: Al(0.954), Pb(0.942), Cu(0.921), Fe(0.968)}, {SSE: Al(0.396), Pb(0.0059), Cu(0.252), Fe(4.29)} {MBE: Al(–0.033), Pb(0.008), Cu(–0.004), Fe(0.091)}, which is developed so each model responsible for estimating concentration single metal. result demonstrates there non-linear relationship between physical-chemical properties evaluated content province.