Knowledge extraction

Spatio-temporal data mining methods must be developed to extract useful patterns out of trajectories. Spatio-temporal data mining is still in its infancy, and even the most basic questions in this field are still largely unanswered: what kinds of patterns can be extracted from trajectories? Which algorithms should be applied to extract them?

The following basic examples give a glimpse of the wide variety of patterns and their potential applications.

Clustering

Clustering, the discovery of group of similar trajectories, together with a summary of each group. Knowing which are the main routes (represented by clusters) followed by people during the day can represent a precious information for improving several different services to citizens. E.g., trajectory clusters may highlight the presence of important routes not adequately covered by the public transportation service.

Classification

Classification, the discovery of behavior rules, aimed at explaining the behavior of current users and predicting that of future ones. urban traffic simulations are a straightforward example application for this kind of knowledge, since a classification model can represent a sophisticated alternative to the simple ad hoc behavior rules, provided by domain experts, on which actual simulators are based

Frequent Patterns

Frequent patterns, the discovery of frequently followed (sub)paths. Such information can be useful in urban planning, e.g., by spotlighting frequently followed inefficient vehicle paths, which can be the result of a mistake in the road planning.

Whichever kind of pattern is extracted, a special attention has to be paid to the risks of privacy violation. Existing work on privacy protection in data mining has to revosed and extended to enforce measurable degrees of protection of identity in the extracted knowledge