Abstract
The rapidly increasing number of sharing bikes has facilitated people’s daily commuting significantly. However, the number of available bikes in different stations may be imbalanced due to the free check-in and check-out of users. Therefore, predicting the bike demand in each station is an important task in a city to satisfy requests in different stations. Recent works mainly focus on demand prediction in settled stations, which ignore the realistic scenarios that bike stations may be deployed or removed. To predict station-level demands with evolving new stations, we face two main challenges:(1) How to effectively capture new interactions in time-evolving station networks; (2) How to learn spatial patterns for new stations due to the limited historical data. To tackle these challenges, we propose a novel Spatial Community-informed Evolving Graphs (SCEG) framework to predict station-level demands, which considers two different grained interactions. Specifically, we learn time-evolving representation from fine-grained interactions in evolving station networks using EvolveGCN. And we design a Bi-grained Graph Convolutional Network(B-GCN) to learn community-informed representation from coarse-grained interactions between communities of stations. Experimental results on real-world datasets demonstrate the effectiveness of SCEG on demand prediction for both new and settled stations. Our code is available at https://github.com/RoeyW/Bikes-SCEG