Publication

Abstract : In recent years, the agricultural sector has faced challenges in achieving optimal crop yields due to complex environmental factors and limited access to data-driven decision-making tools. An innovative IoT-based approach to predict crop yields is presented, aiming to improve crop productivity for farmers. The system utilizes a network of IoT sensors, including DHT22 and rainfall sensors, to gather real-time data on temperature, humidity, and rainfall. Soil-integrated sensors are also used to monitor specific soil properties such as nitrogen, phosphorus, potassium, pH, electrical conductivity, humidity, and temperature. This data is sent to a web interface for live display and stored for predictive modeling. Several regression models—decision tree, random forest, ridge regression, and linear regression—were implemented to predict crop yields, with the random forest regression model achieving the highest accuracy and an error rate of 6.46 percent. The stacking model was also analyzed, but the random forest model’s MAE and R- metrics proved superior, leading to its selection for deployment. A user-friendly web-based interface has been developed to enable farmers to interact with the system and predict crop yields effectively. This implementation smoothly integrates IoT technology with advanced data analytics, providing valuable insights to optimize agricultural practices, which in turn enhances food security and improves livelihoods.