In everyday life, we can observe many phenomena occurring in space and time simultaneously. For example, the movements of a person associate spatial information (e.g. the departure and arrival coordinates) and temporal information (e.g. the departure and arrival dates). Other applications, with more complex dynamics, are much more difficult to analyze. It is the case of spread of infectious disease, which associates spatial and temporal information such as the number of patients, environmental or entomological data. Yuang describes this concept of dynamics as a “set of dynamic forces impacting the behavior of a system and components, individually and collectively”.
In this project, we focus on spatio-temporal data mining methods to better understand the dynamics of complex systems for epidemiological surveillance. In the case of dengue or malaria epidemics, public health experts know that the evolution of the disease depends on environmental factors (e.g. climate, areas with water points, mangroves…) and interactions between human and vector transmission (e.g. the mosquito that carries the disease). However, the impact of environmental factors and their interactions remain unclear.
To address these issues, spatio-temporal data mining provides highly relevant solutions through the identification of relationships among variables and events, characterized in space and time without a priori hypothesis. For example, in our context, we will discover combinations of changes in environmental factors that lead epidemic peaks in specific spatial configurations. However, we know that existing methods are not completely adapted to our problem. For this reason, we will propose, in this project, new algorithms to extract spatio-temporal patterns. These algorithms can be used for analysis by health care professionals, to better understand how environmental factors influence the development of epidemics.