Phenolic Compound–Loaded Nanosystems: Artificial Neural Network Modeling to Predict Particle Size, Polydispersity Index, and Encapsulation Efficiency

Abstract

Artificial neural networks (ANNs) are a useful tool for the prediction of the particle size (PS), polydispersity index (PDI), and encapsulation efficiency (EE) of phenolic compounds (PC) in nanosystems because they consider the effects of all independent variables. This is very important for the prediction of the PS, PDI, and EE of nanosystems produced by ultrasound because the number of variables involved in the encapsulation process makes this prediction very complex. In this research work, three mathematical models for predicting the PS, PDI, and EE of PC in nanosystems produced by ultrasounds using ANN were developed. A database of scientific literature was used. These models allow the PS, PDI, and EE to be correlated mathematically with the PC mass; encapsulating copolymer concentration, ratio and mass; solvent volume; surfactant concentration; emulsion volume; and ultrasound time and power. The optimal configuration of the ANN consisted of a hidden layer with three, four, and two neurons in the hidden layer for PS, PDI, and EE, respectively. The models allowed us to predict PS, PDI, and EE for a wide range of factors. Mean square errors of 0.0538, 0.0337, and 0.0198 and correlation coefficients of 0.9139, 0.9064, and 0.8472 for PS, PDI, and EE, respectively, were obtained during training. Furthermore, mean square errors of 0.0408, 0.0224, and 0.0117 and correlation coefficients of 0.9138, 0.9115, and 0.8955 for PS, PDI, and EE, respectively, were achieved during verification.