Symbolic regression with feature selection of dye biosorption from an aqueous solution using pumpkin seed husk using evolutionary computation-based automatic programming methods

Industrial waste pollution is a serious and systematic problem that harms the environment and people. The development of cheap, simple, and efficient techniques to solve this problem is important for sustainability. In this study, both experimental and evolutionary computation (EC)-based automatic programming (AP) methods were used to investigate the biosorption process for water treatment. In the experiments, titan yellow (TY), an anionic dye, was biosorbed from an aqueous solution containing pumpkin seed husk (PSH). The structure of PSH was examined using a Fourier transform infrared spectroscopy (FTIR) and a scanning electron microscope (SEM). The result of the experimental studies was that TY biosorption of PSH reached a biosorption efficiency of 95% after 120 min of contact time. The maximum biosorption capacity (qmax) was calculated to be 181.8 mg/g. It was found that the biosorption of TY better followed the Dubinin–Radushkevich isotherm (R2=0.98) and pseudo second-order reaction kinetics (R2=0.99). Based on the thermodynamic data, the biosorption process was exothermic and spontaneous. After the experiments, the process was modeled using pH, biosorbent concentration, initial dye concentration, contact time, and temperature as inputs and biosorption efficiency (%) as output for the methods. Moreover, the success of these AP methods was compared with a newly proposed evolutionary method. The simulation results indicate that AP methods generate best models (Rtrain2=0.99 and Rtest2=0.97). At the same time, the most important parameter of the process in the feature analysis is contact time. This study shows that EC-based AP methods can effectively model applications such as the biosorption process.