Optimization of antioxidant extraction through artificial neural networks

The increasing demand in modern society for natural sources of bioactive compounds with potential applications in preventive medicine has driven the development of innovative approaches to optimize extraction processes. While Response Surface Methodology (RSM) has been extensively utilized for prediction and optimization in extraction studies, Artificial Neural Networks (ANNs) provide a powerful alternative by enabling the development of nonlinear computational models. These models are capable of self-learning and training to address practical challenges, without relying on predefined mathematical equations, by mimicking the structure and functionality of biological neural networks. The application of ANNs offers numerous advantages, including the ability to interpret complex datasets, scale results through optimization and parallelization, and model complex, nonlinear relationships. Furthermore, ANNs allow for the handling of large datasets and generalization across systems, without the limitations of predefined models or specific experimental designs. This systematic review compiles studies on the application of ANNs in the optimization of extraction processes of antioxidants present in natural matrices (e.g., edible plants, vegetables, and fruits), providing an objective evaluation of their potential for the sustainable development of industrial products enriched with these compounds. Special attention is given to how ANNs outperform traditional techniques, such as RSM, in predicting yields, enhancing extraction efficiency, and minimizing resource consumption. By exploring key studies and methodologies, this review aims to highlight the role of ANNs in advancing green and sustainable technologies, offering novel insights into their applicability in designing industrial processes that incorporate natural antioxidants into food, cosmetics, and pharmaceutical products.