MACHINE LEARNING-BASED CROP RECOMMENDATION SYSTEM INTEGRATING SOIL PROPERTIES AND WEATHER CONDITIONS WITH IOT-DRIVEN DATA COLLECTION

Abstract

Agriculture forms the backbone of human civilization by ensuring food security, economic growth, and rural development. However, farmers face significant challenges in selecting suitable crops due to variability in soil nutrients, pH levels, and unpredictable weather conditions, often leading to reduced productivity and soil degradation. Indian farmers experience seasonal yield losses due to inappropriate crop selection and limited scientific guidance. Existing crop recommendation systems largely rely on static datasets or conventional machine learning models with limited integration of real-time data, resulting in moderate accuracy levels of 80–90 % and limited adaptability to field variability. To overcome these limitations, the proposed system integrates soil properties and weather conditions using Internet of Things (IoT)-driven data collection. Soil and weather sensors connected through a Long Range (LoRa) gateway collect real-time environmental data, which are processed using the XGBoost algorithm in a cloud environment for accurate crop prediction. The developed system achieved 99 % accuracy, outperforming Decision Tree, Random Forest, and Artificial Neural Network (ANN) models, and provides a reliable, scalable, and sustainable decision-support tool for data-driven precision agriculture.