Yapay nöron ağı (ANN) kullanarak su kalitesi endeksi tahmini

Forecasting is vital to water quality management and is essential when working with a scarce resource like water in this polluted era. Therefore, generating consistent and precise machine learning models to predict the water quality index (WQI) based on significant water quality parameters and climate factors is essential for providing early warnings to prevent environmental disasters In view of this, Yesilirmak river, within the provincial borders of Çorum, turkey, was selected as a case study to predict the water quality index eleven physicochemical parameters (biochemical oxygen demand (BOD), PH, electrical conductivity (EC), dissolved oxygen (DO), chlorine (Cl-1 ), calcium(Ca+2 ), magnesium (Mg+2 ), nitrite (NO2−1 ), sodium (Na+1), sulfate (SO4-2) and total dissolved solids (TDS)) were assigned to calculate the water quality index (WQI). The result varied from good to poor, except for one point (June in 2000) that was unsuitable for drinking. Also, to accomplish this determination, three metaheuristic optimization algorithms (the constraint coefficient-based particle swarm optimization and chaotic gravitational search algorithm CPSOCGSA, the marine predator's optimization algorithm MPA, and particle swarm optimization PSO) were combined with an artificial neural network (ANN) with a Levenberg–Marquardt three-layer backpropagation algorithm architecture to build three hybrid models (CPSOCGSA-ANN, MPA-ANN, and PSO-ANN).In addition, three features of data pre-processing techniques (natural logarithm technique, singular spectrum analysis (SSA), and tolerance) have been used to normalize, enhance the original dataset and identify the optimal model input scenario (respectively). Furthermore, these models' accuracy was verified based on various statistical criteria (MAE, RMSE, and R2). The result revealed that the performance of the CPSOCGSA-ANN model was superior to other models (MPA-ANN and PSO-ANN) by achieving the highest accuracy (R² = 0.965, MAE = 0.01627, and RMSE = 0.0187this window of the thesis that opens to the outside.