Traffic forecasting as a canonical task of multivariate time series forecasting has been a significant research topic in AI community. To address the spatio-temporal heterogeneity and non-stationarity implied in the traffic stream, in this study, we propose Spatio-Temporal Meta-Graph Learning as a novel Graph Structure Learning mechanism on spatio-temporal data. Specifically, we implement this idea into Meta-Graph Convolutional Recurrent Network (MegaCRN) by plugging the Meta-Graph Learner powered by a Meta-Node Bank into GCRN encoder-decoder. We conduct a comprehensive evaluation on two benchmark datasets (i.e., METR-LA and PEMS-BAY) and a new large-scale traffic speed dataset called EXPY-TKY that covers 1843 expressway road links in Tokyo. Our model outperformed the state-of-the-arts on all three datasets. Besides, through a series of qualitative evaluations, we demonstrate that our model can explicitly disentangle the road links and time slots with different patterns and be robustly adaptive to any anomalous traffic situations. Codes and datasets are available at https://github.com/deepkashiwa20/MegaCRN.

No abstract

Notifications of e-mail arrivals at inappropriate times disrupt workers engaged in intellectual activities and reduce their productivity. We propose an e-mail delivery mediation system to deal with this issue. The system suspends e-mail delivery and estimates users' interruptibility from their PC operation activities. Then, it delivers e-mail only when the estimated interruptibility matches certain delivery criteria, especially when a user switches applications. A long-term evaluation in a real office revealed that e-mail mediation based on the system's estimates of user interruptibility can shift a certain percentage of the notifications to task or subtask breakpoints or to times at which the user is less engaged in a task.

To relieve the cognitive impact of interruptions caused by email delivery notifications, we propose an automatic email delivery mediation system based on the user interruptibility, which is estimated from the PC operation activity. A prototype system has been developed to be compatible with existing email clients and to deliver emails at higher estimated interruptibility times, especially at application switching moments that are considered as breakpoints in PC tasks. Trial use by eight participants in ordinary working environments suggested that email notifications were delivered at moments with lower levels of operation activity and the feelings of hindrance were decreased. Further study is to be conducted to investigate the effect of automatic email delivery mediation on the cognitive cost and work efficiency.

No abstract