Abstract: The expected global population growth to 9.1 billion by 2050 is projected to intensify food security challenges. Termites contribute to food security challenges through the damages to crops, wooden infrastructure and water dams. In this study we employ the ant and termite behavioral data during raid from the Mpala Research Center and Machine Learning (ML) models to identify key features that enhance the effectiveness of Matabele ant raids on termites. We found that the number of ants in a raid, foraging distance, and raid timing are significant and manipulatable features in influencing raid effectiveness. We used an XGBoost multivariate regression model to simultaneously predict raid duration, average termites per ant, and the number of ants injured or killed. Raid effectiveness in the study is collectively defined by the three predicted variables. We developed an interactive Streamlit interface to allow users such as farmers and ecologists to simulate various scenarios and optimize outcomes for sustainable termite control. Th model was found to meet the PCS standards by accurately predicting values for our success metrics, deployed a usable model with standard computational resources and found the shaded regions to not be too wide indicating good stability.