Development and comparison of Extreme Learning machine and multi-layer perceptron neural network models for predicting optimum coagulant dosage for water treatment

Abstract

Artificial neural networks have been extensively used in modelling the coagulation process in water treatment. Multi-layer Perceptron neural networks (MLP) are commonly used for coagulation modelling. However, a major drawback of MLPs is the high computational effort required due to its iterative nature. The Extreme learning machine neural network has a prediction accuracy comparable to MLPs and consumes far less computational effort. In this study an ELM-single layer feedforward neural network (ELM-SLFN), ELM-radial basis neural network (ELM-RBF) and a MLP was developed to predict the optimum coagulant dosage. All neural networks performed well with correlation coefficients exceeding 0.97. However, the ELM-RBF network performed better than the MLP model with higher prediction accuracy. Thus ELM-RBF neural network is a more efficient model for prediction of coagulant dosage for water treatment.