Prediction of Soil Organic Carbon Based on Soil Color Using Random Forest

Abstract

This study aims to predict the soil organic carbon (C-organic) content based on soil color using the Random Forest algorithm. This prediction is essential as C-organic is a key indicator of soil fertility. The method used is regression with a machine learning approach. The dataset was obtained from soil color images and actual C-organic laboratory results. The model was evaluated using metrics such as Mean Squared Error (MSE), R-squared (R²), and accuracy. Additionally, a classification was performed to categorize the fertility level of the soil to support the prediction interpretation. The results showed excellent performance of the Random Forest regression model, with an R² of 0.9988 and accuracy of 99.88%. The fertility classification showed perfect precision and recall in all classes. These findings demonstrate that soil color can be effectively used to predict C-organic content and support data-driven agricultural decisions.