predicting the electrical conductivity of water using the hicking ceemd-lssvm optimization algorithm

Accurate prediction of electrical conductivity (ec) concentrations in river water is essential for effective water quality management and environmental protection. this study develops a novel hybrid model, named hoa-ceemd-lssvm, that integrates the hiking optimization algorithm (hoa), complementary ensemble empirical mode decomposition (ceemd), and least square support vector machine (lssvm) to forecast daily ec concentrations in the aidoghmoush river, iran. hoa simultaneously optimizes key parameters of ceemd and lssvm to enhance prediction accuracy. ceemd decomposes complex time series into intrinsic mode functions (imfs) with more predictable patterns, which serve as inputs to the lssvm predictor. the model’s performance is evaluated through multiple metrics, demonstrating significant improvements over benchmark models in terms of r² and kling-gupta efficiency (kge). the proposed model enhances the r2 and kge values of other prediction models by 1%-10 % and 3.17%-17%, respectively. our findings show that the hao-ceemd-lssvm model can precisely forecast ec concentration. this approach provides a robust framework for capturing the nonlinear, nonstationary characteristics of ec time series data. the model is applicable in water resource planning, pollution control, and river ecosystem management. while showing high forecasting accuracy, its computational complexity and black-box nature present limitations. future work should explore parallel computing and explainable artificial intelligence techniques to enhance efficiency and interpretability. accurate prediction of electrical conductivity (ec) concentrations in river water is essential for effective water quality management and environmental protection. this study develops a novel hybrid model, named hoa-ceemd-lssvm, that integrates the hiking optimization algorithm (hoa), complementary ensemble empirical mode decomposition (ceemd), and least square support vector machine (lssvm) to forecast daily ec concentrations in the aidoghmoush river, iran. hoa simultaneously optimizes key parameters of ceemd and lssvm to enhance prediction accuracy. ceemd decomposes complex time series into intrinsic mode functions (imfs) with more predictable patterns, which serve as inputs to the lssvm predictor. the model’s performance is evaluated through multiple metrics, demonstrating significant improvements over benchmark models in terms of r² and kling-gupta efficiency (kge). the proposed model enhances the r2 and kge values of other prediction models by 1%-10 % and 3.17%-17%, respectively. our findings show that the hao-ceemd-lssvm model can precisely forecast ec concentration. this approach provides a robust framework for capturing the nonlinear, nonstationary characteristics of ec time series data. the model is applicable in water resource planning, pollution control, and river ecosystem management. while showing high forecasting accuracy, its computational complexity and black-box nature present limitations. future work should explore parallel computing and explainable artificial intelligence techniques to enhance efficiency and interpretability.