Agronomic Classification of Crops: A Model-based Approach for Precision Agriculture

Precise prediction of crop yield is crucial for making decisions drawn from breeding programs. Although much progress has been made in crop yield prediction, current methods often encounter difficulties with integrating heterogeneous sensor data and achieving superior prediction accuracy under changing environment. This paper is mainly concerned with the principal crop, wheat. Wheat is one of the major staple food crops in the world, and its development largely depends on soil thriving conditions. For many farmers, it's difficult to know if their soil has the correct nutrients, moisture and pH balance necessary for producing healthy wheat. Conventional soil testing technologies are time-consuming, expensive and occasionally inaccurate because of human factors. Our study aims to address this issue by employing Machine Learning (ML) to identify the soil condition automatically and predict whether a soil sample is made for wheat planting or not. Here we consider soil features such as pH, nitrogen, phosphorus and potassium contents and moisture in addition to temperature. These features are fed into the classifiers e.g. SVM, Random Forest and CN model as input to obtain the prediction of soil suitability.