Identifying and predicting social lifestyles in people’s trajectories by neural networks

EPJ Data Science

Table 7 Performance measures averaged over 38 users in a leave-one-out experiment. For all the metrics, except the Hamming loss, higher is better. CK is for Cohen’s Kappa score; JS is for Jaccard similarity

	Accuracy	f1 score	Precision	Recall	CK score	JS score	Hamming loss
HMM	0.18	0.11	0.14	0.13	0.10	0.18	0.82
NB	0.21	0.08	0.10	0.12	0.09	0.21	0.79
LSVM	0.32	0.23	0.25	0.24	0.23	0.32	0.68
RSVM	0.20	0.07	0.07	0.11	0.07	0.20	0.80
PSVM	0.14	0.02	0.01	0.07	0.00	0.14	0.86
SVC	0.31	0.23	0.24	0.24	0.23	0.31	0.69
CNN	0.32	0.23	0.25	0.23	0.24	0.32	0.68
RNN	0.37	0.28	0.31	0.29	0.29	0.37	0.63
LSTM	0.49	0.40	0.42	0.40	0.43	0.49	0.51
BLSTM	0.49	0.40	0.42	0.41	0.42	0.49	0.51
CLSTM	0.52	0.42	0.44	0.42	0.45	0.52	0.48
CBLSTM	0.51	0.41	0.44	0.42	0.44	0.51	0.49