Using Trajectory Subclustering to Improve Destination Prediction
Algorithm 2
inputs: , a set of clusters containing trajectories , the maximum number of components to use for each GMM , the maximum number of instances to select , the features to train the GMM with
outputs: , a set containing the trained GMMs for each cluster in
(1)
fordo
(2)
//get all features vectors for all trajectories within cluster
//e.g. the vector for sspd is lat, long
(3)
//pick random sample of instances from trajectories
(4)
//for number of components in 1 to
(5)
fordo
(6)
//use Bayesian Information Criterion, , to select model