Predictive Modelling of Crop Yield using XGBoost: An Advanced Machine Learning Technique

Global agriculture is the foundation of the economy. Various phenomena, such as unpredictable weather patterns, differences in soil types, and different crop choices, affect the agricultural system and, in essence, keep it from performing its primary functions. The present work intends to address this situation by developing an efficient crop prediction system using state-of-the-art machine learning techniques. This system uses datasets on the medium scale and analyses soil content, weather, rainfall, temperature, and crop yield history to decide the best crops for any region. Different Algorithms-Support Vector Machines (SVM), Linear Regression, Multi-Layer Perceptron’s (MLP), Convolutional Neural Networks (CNN), and XGBoost- are compared in terms of their predictability. The XGBoost model has been selected as the best, due to high accuracy and good generalizability to various agricultural situations, contributing toward sustainable agriculture by enhancing productivity and risk reduction through data-driven decision-making. Augmenting future developments with IoT sensors and satellite imagery combined with real-time data may lead to higher accuracy and larger scale.