A Unified Web Platform for Predictive Modeling and Optimization of Agri-Food Systems Using ANN and GA

Abstract

Objective: To develop a modular, no-code web framework that enables efficient machine-learning-based modeling and optimization of agricultural and food processes as an accessible alternative to traditional coding-dependent environments. Methodology: The framework was built in Python and incorporates a fully integrated workflow covering design of experiments, automated data preprocessing, normalization, ANN-based predictive modeling, and GA-driven optimization. The ANN module includes customizable architectures, real-time training feedback, and performance diagnostics, while the GA module enables multi-parameter optimization with user-defined constraints. The system automates data handling, generates visual dashboards for model evaluation, and allows export of optimized solutions. All computational steps are executed directly through the browser, enabling non-programmers to perform advanced modeling and optimization tasks. Results & Conclusion: Benchmarking against MATLAB (R2024b) showed that the proposed platform eliminated the need for custom scripts, reducing workflow preparation and optimization time significantly. A complete dataset-to-optimization cycle required less than 5 minutes, compared to nearly 15 minutes in MATLAB under equivalent conditions. The ANN models achieved R² = 0.962 ± 0.031 and RMSE < 4%, confirming predictive accuracy comparable to conventional coded implementations. The results demonstrate that this no-code, modular platform provides an efficient, accessible, and reproducible solution for agri-food process modeling and optimization. Its integrated visual analytics and automation capabilities make it well-suited for researchers, practitioners, and industry users seeking rapid, reliable machine learning workflows.