MULTI-DIMENSIONAL TIME-SERIES DATASETS:

 
bullet Camera Mouse Words
Created using the camera mouse program. This dataset consists of 15 two-dimensional trajectories.
Each sequence represents a word, such as Athens, Berlin etc. There are 3 copies for each of the
five words. The dataset is ideal for clustering purposes and for testing the robustness to noise of
various distance measures (the sequences can be especially noisy at various parts). Support of
time warping is necessary for high clustering accuracy.


 
bullet Random Walk

 Two-dimensional random walk, constructed like this:

 x(i) = x(i-1) + randomNumber
 
bullet Marine Mammals

 Trajectories of mammals over time. Obtained from [], cleaned and interpolated. Generally it is
difficult to deduce any patterns or clusters from this dataset, since the trajectory lengths vary
considerably. 



 
bullet TrackMouse Words (donated by Marios Hadjieleftheriou)

 Obtained using a similar procedure as in the camera mouse dataset, by tracking the mouse position
while spelling different words.

Number of words = ???,
number of copies per word = ???.

  (if you wish to obtain any of the above datasets just send me an email)