|
Abstract:
|
In applications of dynamic network models, the network structure may not be directly observed. Methods to learn the network structure rely on other available data about the actors. In sports, with player tracking data, such data may include the location of the individuals through time. We present an extension to these models to gain inference on the dynamic network, where we are interested in the dynamics in the connectivity between the actors. For this extension we assume the states of connectivity follow a semi-Markov process, which leads to adding the length of time a pair of individuals are in a particular state to the model. We use player tracking data from two segments of a soccer game to illustrate the inference that can be obtained.
|
