Dengue Fever (DF) has been identified by the World Health organization (WHO) as one of the most serious vector-borne infectious diseases in tropical and subtropical areas. During 2007, in particular, there were over 2,000 DF cases in Taiwan, which was the highest number of cases in the recorded history of Taiwan epidemics. Most DF studies have focused mainly on temporal DF patterns and its close association with climatic covariates, whereas they have understated spatial DF patterns (spatial dependence and clustering) and composite space-time effects. The present study proposes a spatio-temporal DF prediction approach based on stochastic Bayesian Maximum Entropy (BME) analysis. Core and site-specific knowledge bases are considered, including climate and health datasets under conditions of uncertainty, space-time dependence functions, and a Poisson regression model of climatic variables contributing to DF occurrences in southern Taiwan during 2007. The results show that the DF outbreaks in the study area are highly influenced by climatic conditions. Furthermore, the analysis can provide the required ''oneweek-ahead'' outbreak warnings based on spatio-temporal predictions of DF distributions. Therefore, the proposed approach can provide the Taiwan Disease Control Agency with a valuable tool to timely identify, control, and even efficiently prevent DF spreading across space-time.