This thesis proposes a hybrid spatiotemporal modeling framework to address the limitations of traditional machine learning methods in capturing localized “neighbor effects” during the process of Electric Vehicle (EV) diffusion. Although the rate of EV adoption in the United States has accelerated significantly, passing 10% in 2023, growth remains highly heterogeneous across states. Utilizing a national dataset spanning all 50 U.S. states from 2018 to 2023, this study comparatively evaluates the predictive performance of the XGBoost model against a hybrid model incorporating Sequential Gaussian Simulation (SGS) techniques. The standalone XGBoost model achieved an R2 value of 0.70 on the 2023 test set and identified infrastructure density, socioeconomic fundamentals, policy and political environments, and environmental psychological characteristics as the primary drivers of EV adoption. Furthermore, a Global Moran’s I analysis revealed that, beginning in 2021, spatial autocorrelation within the model residuals became statistically significant, indicating that there exist non-random regional factors resulting in the adoption variance that typical covariates cannot capture. To address this issue, the study introduces an SGS-based correction layer designed to model the spatial structures within the residuals that were not explicitly captured by the standalone XGBoost model. The resulting hybrid ML-SGS model significantly enhanced predictive accuracy, boosting the R^2 score to 0.80. Beyond predictive metrics, this study produces a spatial mean surface alongside a probabilistic mapping of adoption uncertainty, identifying stable leading markets and unstable adoption frontiers. As the 2026 Middle East conflict and subsequent fuel price volatility expose the systemic risks of gasoline dependency, this thesis provides the guidance of infrastructure planning and confirms electric vehicles as a strategic necessity for national energy resilience.