Research on Lake Pollutant Prediction Based on Osprey Optimization Algorithm (OOA)

Abstract

To address the complexities and real-time requirements of water quality parameter measurement, this study proposes an Osprey Optimization Algorithm (OOA)-enhanced method for predicting lake pollutants. Initially, six water quality parameters—including COD, total phosphorus, and total nitrogen - are measured in lake samples using a spectrophotometry-based multi-parameter water quality analyzer. Given the complex nonlinear relationships between these parameters and pollutant concentrations, the Gaussian Process Regression (GPR) model is employed to predict concentrations of three target pollutants. To mitigate suboptimal prediction accuracy observed in the initial GPR model, the Osprey Optimization Algorithm (OOA) is introduced to optimize and refine its hyperparameters, thereby enhancing the model’s adaptability to dataset characteristics. Comparative analysis is conducted between baseline and optimized models. Experimental results demonstrate that the OOA-enhanced model achieved correlation coefficients (R²) exceeding 0.9 for both training and testing sets, with MAE, MAPE, and RMSE metrics approaching zero. This research provides an effective methodological refinement for lake pollutant forecasting.