A Mixed Gaussian Distribution Approach using the Expectation-Maximization Algorithm for Topography Predictive Modelling

The incidence of sugarcane crop infestations at the migration stage, especially by the top borer, can lower yields substantially, which may translate to revenue losses of over 20% across many parts of the world. Traditional pest surveillance approaches tend to lack the accuracy required for timely intervention. This research introduces a new burden rate concept incorporated within a Gaussian Mixture Model (GMM), framed within a machine learning environment in order to enhance the precision of infestation pattern prediction. Through the utilization of the Expectation-Maximization (EM) algorithm, the model easily receives maximum likelihood estimates automatically, thus efficiently dealing with cluster distributions at low computational costs. A significant extension of this research is the inclusion of wind direction and topography as dynamic predictors. This allows for maximizing the model's potential in determining highly susceptible locations of infestation. The incorporation of remote sensing and drone data increases the precision of parameter estimation, leading to accurate predictive modeling. The EM-based clustering method reaches a high level of accuracy of 97.5%, which is greater compared to conventional pest monitoring methods. The result of this study provides a new analytical instrument for pest outbreak control and forecasting in precision agriculture. The tool provides real-time workforce management, selective pest eradication, and efficient resource management. Furthermore, the new synergy of clustering processes, topographic modeling, and remote sensing used in the study achieves a scalable data-driven approach to sustainable farm management that involves proactive crop loss minimization.