A Microservice_Based Iot Framework with Machine Learning Approaches for Drip irrigation scheduling in Rwanda’s Cyohoha Sud Region.

Abstract

Smart irrigation scheduling is vital to tackle water scarcity and improve agricultural productivity, particularly in Rwanda's East Province, where the climate is predominantly sunny. Farmers in the Lake Cyohoha Sud region, frequently encounter irrigation challenges caused by the critical need for efficient water management. This study presents a microservices-based Internet of Things (IoT) and Machine Learning (ML) system for scheduling drip irrigation, a water-efficient technique ideal for the region’s conditions. Microservices are utilized instead of the monolithic paradigm due to their scalability, modularity, and ability to handle the dynamic and distributed workloads of IoT and ML. Each microservice independently manages IoT data acquisition, ML processing, and web application integration, ensuring fault tolerance and efficient resource allocation. The system employs IoT devices to monitor soil moisture, weather conditions, and crop growth in real-time, while ML models process the data to optimize irrigation schedules. Key dataset features include soil moisture, temperature, humidity, rainfall, crop type, and growth stage, with an estimated dataset size of 50,000 records collected over two growing seasons. Among the analysis ML models tested, Random Forest (RF) demonstrated superior performance with an R² of 92%, RMSE of 0.15, and accuracy of 93%, surpassing Support Vector Machines (SVM) and Convolutional Neural Networks (CNN). Evaluation metrics such as R², RMSE, precision, and recall validate the system's performance. At the same time, drip irrigation significantly enhances water usage efficiency and crop yields, demonstrating its practical impact on sustainable agriculture in Rwanda.