AIoTU: An open-source IoT and machine learning platform for smart agriculture and yield prediction

Abstract

Agriculture in developing regions faces persistent challenges in productivity, sustainability, and resource efficiency due to limited access to affordable digital technologies and data-driven decision-support tools. Most existing smart agriculture solutions are proprietary, costly, or poorly adapted to research reproducibility and low-connectivity environments. To address these limitations, this work proposes AIoTU (Artificial Intelligence, Internet of Things, and Unstructured Supplementary Service Data), an open-source intelligent platform designed to support smart and inclusive agriculture. AIoTU automates the collection of environmental data through Internet of Things (IoT) sensors, integrates external weather information, and provides advanced analytical tools and machine-learning-based yield prediction. Built on Django with a modular architecture, the platform offers real-time monitoring, geospatial visualization, and a pre-trained Random Forest Regressor (RFR) model for yield forecasting. The system is lightweight, scalable, and deployable on both Central Processing Unit (CPU) and Graphics Processing Unit (GPU) environments, ensuring accessibility in resource-constrained settings. AIoTU stands out for its open-source nature, high modularity, and research-oriented design, making it adaptable to diverse agricultural contexts and reproducible scientific experiments. Furthermore, the integration of Unstructured Supplementary Service Data (USSD) enables farmers without Internet connectivity to access AI-driven decision support, contributing to the reduction of the digital divide in rural communities.