Adaptive sugarcane monitoring in Mojokerto using a hybrid powered IoT multi-sensor system and machine learning

Abstract

This study develops a hybrid-powered IoT multi-sensor system integrated with machine learning for sugarcane monitoring in Mojokerto. Four sensors—soil moisture, pH, LM35 temperature, and LDR light—are connected to an Arduino UNO R4 WiFi microcontroller. A hybrid power supply (mains electricity and solar panels) and dual data storage (real-time transmission to Google Sheets and local SD backup) ensure resilience and reliability under field conditions. Sensor data are normalized and smoothed prior to analysis using K-Means clustering to map environmental states and a Random Forest classifier to predict crop health. Field validation demonstrates soil moisture as the most influential parameter, followed by temperature, pH, and light intensity. The Random Forest model achieved 93.01% accuracy, 93.88% precision, 99.02% recall, and a 96.38% F1-score on held-out data. By combining hybrid power, multi-sensor integration, dual storage, and machine learning, the system provides robust, data-informed monitoring that supports timely irrigation and management decisions in sugarcane cultivation.