Apart from working on the theoretical foundations of data analytics, our group actively engages in the transfer of our methods to practice. We develop pathpy, an OpenSource python package for the analysis of time-series data on graphs and networks. Examples for data that can be analysed with pathpy include time-stamped social networks, user click streams in information networks, biological pathways, citation networks, passenger itineraries in transportation systems, data from supply chain and logistic networks, or information cascades in social networks.
Taking a new perspective on the problem, pathpy unifies the modelling and analysis of such temporal data based on higher- and multi-order network models of causal paths in time series data. pathpy facilitates the analysis of temporal correlations in time-series data on networks. It uses model selection and statistical learning to generate optimal higher- and multi-order models that capture both topological and temporal characteristics. It can help to answer the important question when a network abstraction of complex systems is justified and when higher-order representations are needed instead.
The science behind pathpy is explained in the following video:
More information, tutorials, talks, and demos on pathpy can be found at http://www.pathpy.net.