modeling of preparation of chitosan/tripolyphosphate nanoparticles using machine-learning techniques

In recent years, there has been extensive research on the synthesis of chitosan-tripolyphosphate (cs/tpp) nanoparticles. however, the influence of different parameters and their interactions on nanoparticles is not well known yet. the purpose of the present study is to use machine-learning techniques, random forests (rf), and artificial neural networks (ann), to estimate the size and zeta potential of nanoparticles based on chitosan and tripolyphosphate concentrations, chitosan molecular weight, degree of deacetylation, ph, temperature, stirring rate, and their interactions. machine-learning algorithms successfully estimated the size and zeta potential of nanoparticles spanning a size range from 50 to 1000 nm. the random forest algorithm had better overall r-squared accuracy of 0.95211 and 0.93978 in comparison to ann model with r-squared accuracy of 0.94744 and 0.93643 for particle size and zeta potential, respectively. results depicted that temperature and stirring rate did not have significant effects on the size and zeta potential. on the other hand, cs concentration and ph were the most important parameters on the size and zeta potential, respectively. furthermore, the results indicated that the primary cause for the unexplained variation in prior studies stemmed from the interactions among parameters. moreover, a correlation exists between size and zeta potential, attributed to the interaction of attractive and repulsive electrostatic charges, ionic interactions, and the quantity of cs molecules.