Data mining of spatio-temporal objects

Peter Revesz

Date: 01/12/2014
University: University of Nebraska, Lincoln
Room : Α56
Time: 13:30

We consider spatio-temporal objects in a broad sense of anything whose shape can be represented in n-dimensional space at all times but varies over time. The most challenging problems regarding spatio-temporal objects are in the interpolation and extrapolation of their trajectories rather than a mere recording of their trajectories. In this talk, we give an algorithmic solution to the problem of inertial navigation, which is the estimation of the current location and direction of a moving object given past periodic measurements of its acceleration measured by an accelerometer. We also describe some recent data mining results on other spatio-temporal objects, including an algorithm to predict many years ahead the citation curves of individual researchers.

MaDgIK 2009-2018