Evaluation of photocatalytic activity of immobilized titania nanoparticles by support vector machine and artificial neural network

In this study, TiO2 nanoparticles immobilized on sackcloth fibre were used for the photodegradation of acid dye, and the efficiency of heterogeneous photocatalysts was predicted using the support vector machines model and artificial neural network model. Acid Red 73 was applied as a model compound. The experimental results were determined as the function of key factors such as initial H2O2 concentration, dye concentration, dissolved anions, pH, and time. The obtained results were used for training the models. To find the most suitable and reliable network, different algorithms and transfer functions were tested. The trial and error method was used to find the optimum number of neurons and layers. The root mean squared of error (RMSE), the sum of square error (SSE), and R-2 for the models were calculated. Results show that support vector machines and neural network models can effectively learn and model the aforementioned process and predict the efficiency of photodegradation of coloured wastewater.