Hybrid Machine Learning Approaches for Accurate Forecasting of Total Dissolved Solids: Case Study of a Chinese Rive

Abstract

Surface water management is one of the important factors of water quality in industrial,
agricultural and urban basins. One of the most important quality indictors is Total
Dissolved Solids (TDS). Also, sampling for TDS is expensive, so metaheuristic methods
are very suitable and cost-effective. In this study, SVM-IWO and SVM-TLBO
metaheuristic models were used to simulate TDS in the Kim-Tin River in China. The
monthly measured temperature, pH , Salinity turbidity and Total Dissolved Solids (TDS)
from 2000-2023 data were used . The evaluation criteria of the coefficient of
determination and root mean square error were used to compare the results. The results
showed that the SVM-IWO metaheuristic method provided a better simulation in the
accuracy section than the SVM-TLBO method (R2
= 0.74 RMSE=63 mg/l). In general,
there is a little difference between the Total Dissolved Solids (TDS) simulation of these
two metaheuristic models. Either model can also be used to simulate TDS in the river.