Estimating Water Quality Parameters Using a Hybrid Extreme Learning Machine Method and Wavelet Theory

Each of the various uses of water, such as agriculture, drinking, industry, etc., require water with a specific quality that is characterized by repeated sampling, testing, and analysis of the results. However, the cost of sampling surface water, measuring quality parameters in the laboratory environment, human errors are the most important problems in estimating the concentration of water qualitative parameters. For this purpose, there are several methods for modeling the water quality parameters. In this regard, the data mining methods have been considered by the researchers in recent decades. Therefore, in this research, the main purpose is to estimating and modeling water quality parameters using modern data mining methods and improve the performance of data mining methods with the aim of wavelet theory and compare them with other commonly used data mining methods. In other words, extreme learning machines (ELM) and multi-layer perceptron (MLP) method will be used to model water quality parameters. The evaluation of these two models was performed by statistical criteria Correlation Coefficient (R), root mean square error (RMSE) and mean absolute error (MAE) and relative standard error (RSE) for statistical data of 20 years. According to the results, it was found that the ELM method has been able to averagely provide a correlation coefficient of 0.97. Although both models yielded acceptable results, the results showed that the ELM model has higher accuracy than the MLP model for prediction of water quality parameters.