Machine Learning-Based Soil Classification

Abstract

A crucial component of agriculture is soil. There are several varieties of dirt. Different properties may be found in each kind of soil, and various crops can be grown on various types of soils. To understand which crops do better in different soil types, we need to be aware of their features and traits. In this situation, machine learning approaches may be useful. It has made significant development in recent years. In the realm of agricultural data analysis, machine learning is still a young and difficult study area. In this study, we provide a model that predicts soil series with regard to land type and, in accordance with prediction, suggests appropriate crops. For soil classification, a number of machine learning techniques are utilised, including Decision Tree (CART), Multilayer Perceptron (MLP) and support vector machines (SVM) using a gaussian kernel. The suggested SVM-based technique outperforms several current methods, according to experimental data. The support vector machine technique is used to examine and compare the classification outcomes of various training sample numbers. The purpose of this work is to evaluate the viability of land cover classification using small samples using SVM and, in accordance with the machine learning algorithm, to investigate novel techniques for quick, non-destructive, and precise land cover classification.