Fig. 7: Predicting co-destination taxi rides in New York City.
From: Structure and inference in hypergraphs with node attributes
We report the AUC calculated running SOO hyperedge prediction on hypergraph test sets built considering taxi rides taking place in various time frames subsequent to the two used to train the algorithm. Test time frames with time window between 17.00-20.00 (A, B) and 00.00-03.00 (C, D), for the two training sets (left-right). Hatched bars denote the performance in the training sets; bars and errors are average and standard deviations over 10 random realizations of the SOO procedure.
