Genetic algorithms assisted machine learning algorithms to optimize nano-phytoremediation of cadmium designed by response surface methodology.

Abstract

Advancements in nanotechnology and artificial intelligence can enhance phytoremediation efficacy, particularly in removing hazardous contaminants like cadmium (Cd). Experiment was conducted by using different concentrations of Cd and titanium dioxide (TiO₂) NPs for different time periods, designed by design of experiment of with a total of 20 combinations. Response Surface Regression Analysis was used for data analysis to identify optimal input factors. Results revealed that TiO₂ nanoparticles significantly improved the efficiency of phytoremediation by increasing Cd uptake. Cd absorption rates were predicted using machine learning models, and their performance was evaluated using R² and MSE metrics. Moreover, the Genetic Algorithm (GA) was employed to minimize MSE between predicted and actual Cd absorption values. Ceratophyllum demersum showed an absorption capacity of 99.58%, with a remaining Cd concentration as low as 0.0199 mg/L. The Gaussian Process Regressor (GPR) was the most accurate predictive model with an R² of 0.99 and MSE of 0.07. The Genetic Algorithm (GA) further optimized the process, identifying optimal NP concentration, Cd concentration, and treatment time. It was concluded that computational models exhibited enhanced Cd absorption due to a synergetic relationship between Cd concentration and treatment time, and absorption efficiency was further enhanced by the supplementation of TiO₂ nanoparticles.