A Destination Prediction Network Based on Spatiotemporal Data for Bike-Sharing

Table 7

Results for different methods of destination prediction.

Methods	Recall	Precision	F1
Historical Count (HC)	28.96%	24.95%	26.81%
Naive Bayesian (NB)	32.14%	27.69%	29.75%
Our Methods (l means layers)
DPNst1: UBS
(2-l LSTM)	33.12%	27.58%	30.10%
DPNst2: UBS + PM
(2-l LSTM + 2-l CNN)	35.46%	30.55%	32.82%
DPNst3: UBS + PM + EF
(2-l LSTM + 2-l CNN + 2-l FCNN)	37.54%	32.34%	34.75%
DPNst4: UBS + PM + EF
(2-l LSTM + 2-l CNN + 5-l FCNN)	31.98%	59.56%	41.62%
DPNst5: UBS + PM + EF
(5-l LSTM + 2-l CNN + 5-l FCNN)	35.27	54.12	42.71