hybrid artificial neural network with imperialist competitive algorithm approach for prediction of adsorption efficiency of ni(ii) and cd(ii) from wastewater by perlite nanoparticles

contamination of water by heavy metals is a global problem. nowadays everybody knows that heavy metal ions consist of iron, lead, manganese, zinc, copper, cadmium, and nickel and so on they are common contaminants in wastewater and known to be toxic and carcinogenic that lead to many problems for human and water environment. in this research, experiments have been performed in the batch system to obtain equilibrium data of the individual adsorption of cadmium and nickel ions by perlite nanoparticles. the experiments have been carried out for the chosen temperature of 25 and operational conditions such as constant agitation and ph 4 and 6 for cadmium and nickel respectively. the results of isotherm show that the langmuir isotherm showed better correlation with the experimental data. also,in this paper, the model based on a multilayer perceptron artificial neural network (mlp-ann) optimized by imperialist competitive algorithm (ica) to predict of heavy metals removal process (cd2+ and ni2+) is proposed. ica is used to decide the initial weights of the neural network. the ica–ann model is trained using an experimental data set to approximate the relation between c0 and ce are the equilibrium concentration of the cadmium and nickel in solution, and time contact as inputs and adsorption efficiency as output. the performance of the ann-ica model is compared with multiple linear regressions (mlr). coefficient of determination (r2) and mean square error (mse) were calculated for the models to compare the results obtained. for the ann-ica model to predict of heavy metals, r2 and mse are equal to (0.9297 and 0.0141 for ni2+) and (0.9539 and 0.012 for cd2+). the results demonstrate the effectiveness of the ann-ica model