Name
Abstract | ||
|---|---|---|
Taxis are one of the representative modes of traffic systems. However, with the emergence of shared cars led by DiDi and Uber in recent years, the traditional taxi companies are facing unprecedented competitions. Without personalized data collected from the mobile devices, passenger flow prediction based on vehicle GPS records presents a unique solution that can improve taxis’ operating efficiency while preserving personal privacy. In this paper, we propose the Travel Behavioral Inertia (TBI) from taxi GPS records, which embodies Driver Inertia (DI) and Passenger Inertia (PI). Then we integrate TBI with other features to construct multi-dimensional features and predict taxi passenger flow based on a deep learning algorithm. We call the entire framework TBI2Flow. Extensive experiments demonstrate that TBI features has outstanding contribution to passenger flow prediction and TBI2Flow outperforms state-of-the-art methods including time series-based method and other deep learning-based methods on long-term taxi passenger flow prediction. |
Year | DOI | Venue |
|---|---|---|
2020 | 10.1007/s11280-019-00700-1 | World Wide Web |
Keywords | Field | DocType |
Travel behavior, Traffic flow prediction, Smart city, Taxi operation | Travel behavior,Data mining,Computer science,Flow (psychology),Taxis,Real-time computing,Mobile device,Global Positioning System,Artificial intelligence,Smart city,Inertia,Deep learning | Journal |
Volume | Issue | ISSN |
23 | 2 | 1386-145X |
Citations | PageRank | References |
1 | 0.35 | 13 |
Authors | ||
6 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Xiangjie Kong | 1 | 425 | 46.56 |
| Feng Xia | 2 | 2013 | 153.69 |
| Zhenhuan Fu | 3 | 1 | 0.35 |
| Xiaoran Yan | 4 | 2 | 0.73 |
| Amr Tolba | 5 | 177 | 29.10 |
| Zafer Al-Makhadmeh | 6 | 17 | 6.79 |