Gradient descent–based linear regression: A novel framework for Kayseri Organized Industrial Zone wastewater treatment effluent parameters estimation

In this study, a method is proposed to develop a model for the estimation of chemical oxygen demand (effluent chemical oxygen demand [e‐COD]), suspended solids (effluent suspended solids [e‐SS]), and pH (effluent pH [e‐pH]) parameters from the discharge water parameters using various parameters of the wastewater treatment plant of the Kayseri Organized Industrial Zone (KOIZ‐WWTP). A gradient descent algorithm (GDA) based on machine learning, which is widely used to detect linear regression parameters, is proposed. In the first stage of the two‐stage study, the relationships between each parameter were determined, and correlational selection was applied to the input parameters accordingly. Thus, the relationships of 11 input parameters with the target parameters were determined and the parameters with high correlation between them were determined and selected. A median filter was applied to all datasets used, thus smoothing out potential outliers and reducing noise. In the second stage the batch gradient descent variant of the GDA machine learning algorithm was used to determine model parameter values. The model was applied to the training dataset and its accuracies were obtained on the test dataset. Root mean square error (e‐COD: 0.2771, e‐SS: 0.2792, e‐pH: 0.3240), variance accounting factor (e‐COD: 91.9362, e‐SS: 92.2587, e‐pH: 88.6390), and R 2 adj (e‐COD: 0.910, e‐SS: 0.913, e‐pH: 0.873) were obtained as performance metrics. Thanks to this study, the model parameters were determined fairly accurately iteratively without overfitting. With this study, it is planned to reduce the labor force of this and similar facilities in terms of consumables, equipment, and most importantly time, by accurately estimating the effluent parameters of KOIZ‐WWTP. Furthermore, this framework directly contributes to environmental sustainability and operational efficiency by offering potential energy savings in aeration processes and reducing the dependency on extensive laboratory analyses.