Using Trajectory Subclustering to Improve Destination Prediction

Table 1

Notation used in this article.

Notation	Description
	A set of trajectories
	A set of trajectories in cluster
	A trajectory within , a strictly ordered sequence of instances
	An instance is a latitude and longitude position, , at time
	A dissimilarity matrix for the input set of trajectories
	A set of clusters extracted from trajectories in
	A set of GMMs trained for each cluster
	A pretrained GMM for the cluster
	A probability value for a trajectory, , fitting the most likely GMM in
	A sparse matrix containing the clustering parameters, with an implicit ordering
	A parameter value for current iteration of clustering
	The distance function for clustering, which is used to calculate
	The maximum number of components for a GMM
	The maximum number of instances for training a GMM
	The best (lowest) Bayesian information criterion value for a GMM
	The parameter value to use for the decision threshold, that is, a probability value in the static mode and a multiplier in the dynamic mode
	A Boolean flag indicating whether to use the dynamic (true) or static (false) mode for the decision threshold