Our project demonstrates the effectiveness of combining machine learning and geospatial analysis to study Urban Heat Island Intensity (UHII) in the Appalachian region. By integrating satellite data, topographic features, and socioeconomic indicators, we built predictive models that not only achieved strong accuracy but also revealed important spatial patterns. Gradient Boosting and Random Forest did quite similar prediction for UHII, while Geographically Weighted Regression provided meaningful local insights, especially in topographically complex areas. Overall, our findings underscore the value of hybrid approaches that balance predictive power with spatial interpretability—critical for environmental planning, public health, and climate resilience efforts in diverse landscapes.