This report analyses current methods of spatio-temporal prediction and summarises the results by establishing a criteria we deem to be of importance when developing a predictive methodology. Using these statements, we introduce a novel approach for identifying emerging spatio-temporal hotspots. Utilising Bayesian Statistics and a measure of similarity between prior and posterior probability distributions we adapt the method of Bayesian Surprise, from sensory processing, into a methodology that can use `surprising’ events as a way to detect emerging spatio-temporal trends. We analyse the effectiveness of the methodology, by varying a selection of parameters, on three different datasets using a specially created application. The application, by design, is capable of functioning with any correctly formatted spatio-temporal data with no restrictions on the data’s domain. We then conclude the report by summarising the results obtained, our conclusions on the methodologies effectiveness, and a selection of suggestions for future work.