Publication

Abstract : Precise crop yield forecasting is essential for precision agriculture, maximizing resource utilization, and boosting output. However, the outdated yield forecast techniques are not always applicable or accurate. Traditional models are less expandable and are unable to reflect the intricate non-linear relationships among soil factors, climate, and spatial variability because these frequently rely on incomplete sampling data and simple statistics. In the study, a system that combines meteorological, soil, and satellite imagery with Extreme Gradient Boosting (XGBoost) geographical data is proposed. The technique uses high-dimensional, multi-source input to improve prediction robustness and accuracy. The proposed system in the paper performs far better than the existing systems, according to experiments. With a mean absolute value of 500 kg/ha and a root mean square error of 650 kg/ha, the model's R2 is 0.89, and its performance holds steady when used for various crops or geographical areas. A significant advancement above conventional methods of agricultural output prediction, the fusion-driven machine learning (ML) system will enhance farming operations' sustainability, scalability, and decision-making.