IoT-Enabled Machine Learning Framework for Strawberry Yield Prediction with Mobile Decision Support

Smart agriculture increasingly relies on real-time monitoring and data-driven decision support to improve crop productivity. Strawberry cultivation, in particular, is sensitive to soil and environmental variations that must be monitored accurately. Traditional soil testing methods are slow, costly, and infrequent, limiting farmers’ ability to respond effectively to changing field conditions. Existing IoT-based solutions often lack integrated prediction capabilities or realtime decision support. This study presents an integrated IoT and machine learning framework for real-time soil monitoring and strawberry yield prediction. To ensure real-time data transmission and accessibility we employ 4502C pH, SKU-SN-0070 moisture, and RS-485 Ec sensors, with Arduino-nano microcontroller which serves as a hub for monitoring the various soil characteristics efficiently and relaying this information to the mobile application. A network of low-cost sensors measures pH, moisture, and electrical conductivity, while a mobile application provides instant access to field data and alerts. Soil data collected from strawberry fields in Jhelum were used to train and evaluate multiple machine learning algorithms for yield classification and analysis. The Random Forest model achieved the highest classification accuracy of 99.95%, outperforming Sequential Minimal Optimization (SMO), Naïve Bayes, and Bagging. The availability of multiple algorithms and the possibility of working with larger datasets in the future provide a means of creating significant comparisons. Targeting various fruit varieties would also be beneficial.