Global Spatiotemporal Dynamics of African Swine Fever: An Integrated Multi-Scale Spatial and Time-Series Analysis

African swine fever (ASF) poses a persistent and escalating threat to global swine production. To comprehensively characterize its global spatiotemporal dynamics from 1996 to 2025, we developed an integrated framework combining multi-distance spatial analysis and advanced time series forecasting, utilizing a dataset of 57,253 outbreak records. Our findings reveal a clear divergence in transmission patterns: wild boar accounted for approximately 70% of outbreaks and predominantly sustained transmission in Eastern Europe, whereas domestic pig outbreaks were largely concentrated in Southeast Asia. A pronounced epidemiological shift occurred between 2017 and 2020, during which ASF spread transitioned from a predominantly north–south axis linking Africa and the Caucasus to a broad east–west expansion across Eurasia, coinciding with rapid dissemination throughout Asia. In the Northern Hemisphere, ASF outbreaks exhibited a bimodal seasonal pattern, with peaks observed in January–March and July–August. Comparative forecasting analyses demonstrated that machine learning approaches consistently outperformed both traditional statistical and deep learning models. Among these, the random forest algorithm achieved the highest predictive accuracy, surpassing SARIMA, Prophet, XGBoost, and GRU. Collectively, these findings underscore the pivotal role of wild boar in maintaining global ASF transmission and highlight the necessity of integrated surveillance at wildlife–livestock interfaces. Furthermore, they support the application of machine learning-based approaches for improving early warning systems and enhancing the effectiveness of global ASF control strategies.