Leveraging IoT and Climate Insights for Soil Property Prediction in Precision Agriculture

Abstract

In addition to the spectrum range, apparatus, and data analysis techniques used to relate reflectance data to soil parameters, the performance of soil spectroscopy varies from site to site based on the physical composition and chemical qualities of the soil. Understanding soil characteristics enables farmers to implement effective and efficient agricultural practices, resulting in increased crop yields with reduced resource consumption. This study endeavors to predict soil characteristics utilizing Machine Learning (ML) methodologies. The principal constituents of soil prediction are calcium, phosphorus, pH, soil organic carbon, and sand. These characteristics significantly affect crop yield. Four established ML models which are XGBOOST, SVM, Decision Tree, and Random Forest are utilized for predicting these soil qualities. On the basis of the Africa Soil Property Prediction dataset, the performance of these models is assessed. The experimental findings indicate that XGBOOST outperforms other models in terms of the coefficient of determination. Knowing the characteristics of the soil in their particular terrain will be useful to the farmers.