Machine Learning for Soil Fertility Assessment: A Comprehensive Study

Geographical, environmental, and meteorological factors have a significant impact on soil, which is referred to as Earth's skin. Mineral and nutrient content is crucial for controlling the ecosystem's core dynamics. Crop yield is aided by soil fertilization activities. However, a smaller crop output volume may result in the amount of soil composition (fertilizer) failing to be regulated and kept constant. Advancements to help avoid lower crop quality and production to control the amount of soil fertilizer, in addition to improving fertilization and plant growth and soil nutrient monitoring, are crucial. Instructive soil parameters to ascertain soil fertility, calcium, phosphorus, and pH are therefore among the parameters that are frequently measured to monitor soil fertility. The grading of soil and prediction of crops that are suitable for specific land areas can be achieved using a machine learning-based method for the examination of key soil attributes. The high demand for laboratory-based analyses of soil fertility has led us to devise a quick, easy, and affordable method for assessing soil fertility. Portable X-ray fluorescence (pXRF) spectrometry determines the total elemental concentration in soils quickly, easily, and without producing hazardous waste, but its use for predicting soil fertility properties in tropical conditions is still in its infancy. It uses supervised machine learning algorithms like decision trees, K-Nearest Neighbor (KNN), and Support Vector Machine (SVM) to predict soil fertility based on the macronutrient and micronutrient status information found in datasets. R Tool is used to create supervised machine learning algorithms, which are then tested on the test dataset and applied to the training dataset. Several assessment criteria, such as accuracy, cross-validation, and mean absolute error, are used to analyze the performance of these algorithms. Analysis of the results reveals that decision tree has the lowest mean square error (MSE) rate and the highest accuracy of 99%. With less waste production and lower expenses, this environmentally friendly approach can be applied to the evaluation of soil fertility characteristics in a variety of tropical and subtropical soil types.