Use of ANFIS/Genetic Algorithm and Neural Network to Predict Inorganic Indicators of Water Quality

The present research used novel hybrid computational intelligence (CI) models to predict inorganic indicators of water quality. Two CI models i.e. artificial neural network (ANN) and a hybrid adaptive neuro-fuzzy inference system (ANFIS) trained by genetic algorithm (GA) were used to predict inorganic indicators of water quality including total dissolved solids (TDS), total hardness (TH), total alkalinity (TAlk), and electrical conductivity (σ). The study was conducted on samples collected from water wells of Kermanshah province through analyzing water parameters including pH, temperature (T), and the sum of mill equivalents of cations (SC) and anions (SA). A multilayer perceptron (MLP) structure was used to forecast inorganic indicators of water quality using the ANN approach. A MATLAB code was used for the proposed ANFIS model to adjust and optimize the ANFIS parameters during the training process using GA. The accuracy of the generated models was described using various evaluation techniques such as mean absolute error (MAE), correlation factor (R), and mean relative error percentage (MRE%). The results showed that both methods were suitable for predicting inorganic indicators of water quality. Moreover, the comparison of the two methods showed that the predicted values obtained from the ANFIS/GA model were better than those obtained from the ANN approach.