<![CDATA[The implementation of Machine Learning (ML) in precision agriculture often stops at probabilistic outputs without producing directly executable (actionable) agronomic guidance. In response to this functional gap, this study created an IoT proof of Concept (PoC) system architecture design that integrates Machine Learning with the Mamdani Fuzzy Expert System. To validate the software before entering the implementation phase, the system was trained and tested using 5000 synthetic datasets. This synthetic data was generated using a Multimodal Gaussian distribution to represent real-world conditions in the field (TDS, pH, humidity, and temperature). The evaluation results obtained proved that the Naive Bayes classification was very efficient for use in system architectures with minimal resources, with an accuracy of 96.65% and a latency of 0.054. In the final stage of the system, the Fuzzy Logic integration successfully produced output in the form of solutions and recommendations based on the values generated by the Machine Learning system. The integration of these three systems successfully became a unified, complete architectural system. This system has proven to be able to bridge the gap between the complexity of the system and the needs of farmers, and is ready to be implemented in the field using real-time data.]]>
Copyrights © 2026
