Using Trajectory Subclustering to Improve Destination Prediction

Algorithm 4

, prediction stage of DPTS.

	inputs:, a unfolding trajectory to predict , a set containing the trained GMMs for each cluster in
	output:, the predicted cluster for the last instance in , along with the probability for the prediction
(1)
(2)
(3)	fordo
(4)
(5)
(6)	fordo
	//calculate likelihood for instance in GMM
(7)
	//convert likelihood into probability using softmax function
(8)
	//increment total likelihood and probability
(9)
(10)
(11)	end
	//calculate average probability
(12)
	//only predict for the latest instance in the unfolding trajectory
(13)	ifthen
(14)
(15)
(16)
(17)	end
(18)	end
(19)	return