An ANN-Based Information System for Predictive Modeling of Green Antioxidant Extraction from Medicinal Plants

Abstract This paper presents an Artificial Neural Network (ANN)-based information system for predictive modeling of eco-friendly antioxidant extraction from local medicinal plants. The study combines laboratory extraction experiments with a data-driven machine-learning workflow in order to align the contribution with Artificial Intelligence, Computer Science Applications, and Information Systems. Four medicinal plant materials from the Sidi Bel Abbes region of Algeria were investigated: olive leaves, rosemary, pomegranate peels, and Aloysia citriodora. Experimentally, the work compared maceration, decoction, ultrasound-assisted extraction, and ultrasound-assisted extraction coupled with mechanical agitation. Phytochemical assays confirmed that ultrasound plus agitation produced the strongest overall extraction performance, especially for total phenolic content and antioxidant activity. To move beyond descriptive chemistry and toward intelligent decision support, a global ANN model was developed to predict DPPH inhibition using plant type, extraction technique, and extract concentration as input descriptors. The model was trained and assessed using 5-fold cross-validation on 180 replicate-level observations. The selected multilayer perceptron architecture, with two hidden layers (16 and 8 neurons), achieved strong predictive performance with R² = 0.975, MAE = 2.28, and RMSE = 4.97. These results show that the proposed ANN framework can accurately capture the nonlinear interaction between botanical matrix, extraction strategy, and concentration-dependent antioxidant response. From an information systems perspective, the proposed workflow acts as a compact decision-support layer that can guide green extraction screening and reduce trial-and-error experimentation. The study therefore contributes an ANN-oriented computational framing of medicinal-plant extraction that is more suitable for an AI-centered journal while preserving the environmental and phytochemical relevance of the original research.