Modelling of COD concentrations in mixed industrial wastewater using artificial neural network optimized with NRBO and CPO algorithms

6. INTERNATIONAL ENVIRONMENTAL CHEMISTRY CONGRESS, Trabzon, Türkiye, 5 - 08 Kasım 2024, ss.25, (Özet Bildiri)

This study aims to investigate the usability of classical regression analysis (CRA), multivariate adaptive

regression splines (MARS) and artificial neural network (ANN) optimized with NRBO (Newton-Raphson-Based

Optimizer) and CPO (Crested Porcupine Optimizer) algorithms (opANN) models for prediction of chemical

oxygen demand (COD) concentration in mixed industrial wastewater.1-2 In the models, suspended solids,

color, pH, total nitrogen and temperature were used as input parameters while COD concentration was

defined as output parameter. In the development of CRA and MARS models and in the training of opANN

model, 80% of the data set with 88 data was used and the remaining 20% was used for testing the models.

The CRA approach with seven different regression functions (quadratic (QF), power (PF), exponential (EF),

inverse (IF), ln (LnF), linear (LF), and s (SF)) was applied for the prediction of COD concentration. Some ANN

prediction metrics (root mean square error, relative root mean square error, mean absolute error, Nash-

Sutcliffe efficiency, performance index, Pearson correlation coefficient and A10-index) were used to evaluate

the performance of these models. opANN method showed the best results among the applied models.