Spatial-Temporal Analysis on Spring Festival Travel Rush in China Based on Multisource Big Data

Jiwei Li, Qingqing Ye, Xuankai Deng, Yaolin Liu, Yanfang LiuSustainability 2016 · DOI

TL;DR

This paper analyzes the massive annual migration in China, known as "Chunyun" or the Spring Festival travel rush, using big data from Baidu, Tencent, and Qihoo. It reveals that economic and social factors are stronger drivers of travel than geography, and suggests a unified real-time traffic platform integrating mobile and official data to manage this phenomenon.

What Problem Does This Paper Address?

Every year, hundreds of millions of people in China undertake a massive migration to return home for the Spring Festival, a phenomenon called "Chunyun." This creates an enormous strain on transportation systems and leads to significant social and logistical challenges. While previous research often relied on limited data or qualitative methods, this study aims to provide a more in-depth, quantitative understanding of the spatial and temporal dynamics of this travel rush by leveraging multisource big data. The goal is to identify key characteristics, influencing factors, and potential solutions to mitigate the associated problems.

Methods

The researchers employed a multi-faceted approach:

  • Time Series Analysis: This was used to examine the overall trends of travel flows, identify peak and valley periods, and compare the temporal patterns across different data sources.
  • Complex Network Analysis: This powerful technique was used to model the travel network at the provincial level. They specifically investigated:
  • Small-world and Scale-free Properties: To understand if the network exhibits characteristics of efficient information spread or hubs.
  • Centrality Analysis: To identify key provinces (nodes) and travel routes (edges) that play a significant role in the network. They proposed a novel method using net travel flows to categorize provinces as immigration or emigration hubs.
  • Community Detection: To identify distinct clusters of provinces with high internal travel or migration.
  • Data Comparison and Validation: The study rigorously compared the three big data sources (Baidu, Tencent, Qihoo) and validated their findings against 2010 census migration data to assess accuracy and reliability.

Datasets

The study utilized three primary big data sources for the 2015 Spring Festival travel season:

  • Baidu LBS data: Location-based services data from Baidu, covering multiple transportation modes.
  • Tencent LBS data: Location-based services data from Tencent, also covering multiple transportation modes.
  • Qihoo automatic train ticket booking platform data: Data specifically from train ticket bookings, thus primarily reflecting long-distance travel.

Additionally, 2010 national census migration data was used as a reference for validation.

Key Insights and Ideas

  • Data Source Reliability: Location-based service (LBS) data from Baidu and Tencent appears to be more comprehensive and reliable for capturing the overall travel rush compared to Qihoo's train-specific data, which shows limitations in capturing diverse travel modes and potential data incompleteness.
  • Temporal Patterns: The travel rush exhibits distinct temporal patterns, with two major peaks occurring approximately five days before and six days after the Spring Festival, and a significant dip on the actual festival day.
  • Network Structure: The Spring Festival travel network at the provincial level is not a "small-world" or "scale-free" network. Instead, it's characterized by a multicenter structure (multiple hubs) and strong geographic clustering, suggesting that travel is concentrated within certain regions.
  • Driver of Migration: Economic and social factors (like GDP, industry, population, urbanization) are more influential in shaping travel patterns than purely geographical distance.
  • Unbalanced Development: The persistent challenges of the travel rush are deeply rooted in China's unbalanced urban-rural and regional development, which drives large-scale migration for economic opportunities.

Key Results

  • Baidu and Tencent LBS data showed high correlation with each other and with the census data, indicating their accuracy. Qihoo data showed a weaker correlation.
  • Two distinct travel peaks were observed: one around five days before the Spring Festival and another around six days after.
  • The provincial travel network did not exhibit small-world or scale-free characteristics.
  • The network displayed a multicenter structure, with eastern coastal and central/western populous provinces acting as key hubs.
  • Economic and social factors had a stronger correlation with network node strength than geographical location factors.
  • The study identified specific travel path chains (e.g., Guangzhou-Hubei-Henan-Beijing) that play a leading role in the network.

Limitations and Future Work

  • Data Incompleteness/Bias: The Qihoo data, being solely from train bookings, likely misses other travel modes and may suffer from incompleteness. Baidu's data showed a stronger pre-festival travel flow than post-festival, which deviates from theoretical expectations.
  • Time and Spatial Resolution: While big data offers improvements, differences in time and spatial resolution across datasets can still pose challenges.
  • Future Work: The authors strongly advocate for the establishment of a unified real-time traffic platform that integrates mobile big data with official transportation data. This platform could significantly improve the government's ability to manage resources and provide timely information to the public, thereby alleviating the problems associated with the Spring Festival travel rush. They also suggest that as China's urban-rural and regional development becomes more balanced, the severity of the abnormal travel rush may decrease over time.

Summary updated

Abstract

Spring Festival travel rush is a phenomenon in China that population travel intensively surges in a short time around Chinese Spring Festival. This phenomenon, which is a special one in the urbanization process of China, brings a large traffic burden and various kinds of social problems, thereby causing widespread public concern. This study investigates the spatial-temporal characteristics of Spring Festival travel rush in 2015 through time series analysis and complex network analysis based on multisource big travel data derived from Baidu, Tencent, and Qihoo. The main results are as follows: First, big travel data of Baidu and Tencent obtained from location-based services might be more accurate and scientific than that of Qihoo. Second, two travel peaks appeared at five days before and six days after the Spring Festival, respectively, and the travel valley appeared on the Spring Festival. The Spring Festival travel network at the provincial scale did not have small-world and scale-free characteristics. Instead, the travel network showed a multicenter characteristic and a significant geographic clustering characteristic. Moreover, some travel path chains played a leading role in the network. Third, economic and social factors had more influence on the travel network than geographical location factors. The problem of Spring Festival travel rush will not be effectively improved in a short time because of the unbalanced urban-rural development and the unbalanced regional development. However, the development of the modern high-speed transport system and the modern information and communication technology can alleviate problems brought by Spring Festival travel rush. We suggest that a unified real-time traffic platform for Spring Festival travel rush should be established through the government's integration of mobile big data and the official authority data of the transportation department.
BibTeX
@inproceedings{li2016spatial,
  author = {Jiwei Li and Qingqing Ye and Xuankai Deng and Yaolin Liu and Yanfang Liu},
  title = {{Spatial-Temporal Analysis on Spring Festival Travel Rush in China Based on Multisource Big Data}},
  booktitle = {Sustainability},
  year = {2016},
  doi = {10.3390/su8111184},
}
Figure