The emerging spatiotemporal data (e.g. social event data like crimes, or meteorology science data) bring in both challenges and opportunities for understanding the complex temporal and/or geographic processes and phenomena from different fields. Learning from spatiotemporal data has been a challenging task, because the variations among spatial, temporal and multivariate spaces have created a huge amount of features and complex regularities within the data. In this study we presented three methods targeting at two analytical purposes of spatiotemporal events: visualization and prediction. In detail, we developed 1) Hotspot Optimization Tool (HOT) for spatial incident hotspot visualization with reasoning, 2)Spatial Cluster Optimization Tool (SCOT)for summarizing the spatial regularity of variable's temporal patterns and an instance-based algorithm, Spatial Cluster Pattern based Classifier (SPC) for predicting spatiotemporal events through classification. All the three algorithms were built based on the mining of contrast frequent pattern. We applied and evaluated our algorithms using two real-world spatiotemporal data set: residential burglary data collected from a northeast city in the United States for crime hotspot visualization and meteorological data for forecasting extreme rainfall events in the eastern Central Andes area.