Uncertain time series analysis has recently become an important research topic, particularly when searching for features of natural phenomena using similarity functions. Natural phenomena are often modeled as time series, such as in weather forecast, in which temperature variation is monitored through space and time. In such a context, different models for weather forecast produce variations on predictions that can be interpreted as predictions uncertainty. One important problem is to represent the variations presented in predictions along space and time. In order to address a solution to this problem, this paper defines a new type of series, here named uncertain spatio-temporal series, and proposes a computational strategy to manage uncertainty in probabilistic database. Using this new series some analytical queries can be performed, leading to the discovery of interesting observation patterns.
Clustering spatio-temporal series of confirmed COVID-19 deaths in Europe
