Development of Accurate Soil Profile Prediction for the Agriculture Sector using Machine Learning

Abstract

India is one of the largest agricultural countries in the world. It contributes major production of pulses and spices and is considered one of the powerhouses of agriculture. However, the agriculture sector is facing the challenges such as soil degradation, low productivity, variations in the monsoons, lack of agriculture infrastructure, water scarcity and irrigation. There are several methods to find soil profile prediction such as field observation, soil sampling analysis, Pedogenesis  model, remote sensing, and machine learning algorithms. Among these model, machine learning algorithm is one of the emerging trends because of scalability, adaptability, versatility, robustness and continuous improvement. In this regard, an Improved Regression by Discretization Algorithm (IRDA) is used to predict the soil profile of an agriculture unit for the better cultivation.  The accurate value of PH, presence of essential nutrients , and the nature of soil are determined using IRDA algorithms. The performance of IRDA algorithm is compared with other popular machine learning algorithms using cover coefficient, mean absolute error, and root relative square errors.