Inferring High Quality Co-Travel Networks

Youfang Lin, Xuguang Jia, Mingjie Lin, Steve Gregory, Huaiyu Wan, Zhihao WuarXiv 2013 · arXiv

Introduction

In the aviation industry, transactional data (OLTP) is abundant, yet explicit social connections between passengers—such as family ties or professional relationships—are rarely recorded. While "co-flight" counts offer a starting point for inferring these relationships, they are often contaminated by "passive ties" created by large tour groups or travel agencies. This paper introduces Co-Travel Networks, a framework designed to filter this noise and reveal authentic social structures by introducing a sophisticated "Co-Journey Times" metric.

The Challenge of Passive Ties

A primary hurdle in social network inference from aviation data is distinguishing between Active Ties (genuine social connections) and Passive Ties (strangers grouped by a third party). For instance, a travel agency booking 50 strangers into a single tour group creates thousands of artificial links. A naive algorithm would treat these as close connections, leading to a network dominated by "artificial hubs."

Distribution of PNR sizes showing the prevalence of small groups and the long tail of large tour groups.

Methodology: From Co-Flight to Co-Journey

The researchers propose a two-step approach using Passenger Name Records (PNR):

  1. Co-Journey Discovery: Instead of simple segment counts, the system identifies "complete journeys." A state machine tracks transitions between "Inside-of-Co-Travel" and "Outside-of-Co-Travel," accounting for layovers and round trips to define the boundaries of a single social event.
  2. Tie Strength Calculation: The core metric is Co-Journey Times. To mitigate noise, groups are categorized into Small Passenger Groups (SPG) (fewer than 10 people) and Large Passenger Groups (LPG) (10+ people). While a single shared trip in a small group suggests a strong tie, shared trips in large groups are treated as "uncertain" unless validated by subsequent travel in smaller configurations.

Data and Scale

The study leveraged a massive dataset from the Civil Aviation Administration of China (CAAC), encompassing two years of nationwide domestic PNR data. This included approximately 256 million PNRs and 100 million distinct passengers across 5 million flights. To ensure privacy, all passenger identities were encrypted and location data masked.

Key Insights: Behavioral Patterns

The analysis revealed distinct behavioral differences between group types. Notably, large travel groups exhibit a "5-Multiple Preference," where group sizes cluster around multiples of five (10, 15, 20, etc.), reflecting the logistics of organized tours.

Furthermore, there is a strong correlation between group size and booking complexity. Larger groups are significantly more likely to book return tickets within the same PNR, as shown in the distribution of flight segments.

Proportion of flight segments in PNRs for small (PsgrCnt < 10) versus large (PsgrCnt >= 10) passenger groups.

Results and Impact

The resulting network comprises 75 million passengers and 320 million co-travel relationships. By applying the Co-Journey method, the researchers produced a network with a smooth power-law degree distribution, indicating the successful removal of artificial hubs. Unlike digital-only social networks, co-travel networks represent high-cost, face-to-face interactions, making them exceptionally authentic proxies for real-world social circles. Specifically, the study found that at a threshold of just one or two co-journeys, they could identify the stable "backbone" of the network.

Conclusion and Future Work

While highly effective, the model currently relies primarily on PNR data. Future refinements could incorporate seat proximity, shared luggage tags, or check-in times. Beyond academic interest, these high-quality networks have practical applications in optimized seat allocation, personalized marketing, and the detection of ticket agency fraud.

Summary updated

Abstract

Social networks provide a new perspective for enterprises to better understand their customers and have attracted substantial attention in industry. However, inferring high quality customer social networks is a great challenge while there are no explicit customer relations in many traditional OLTP environments. In this paper, we study this issue in the field of passenger transport and introduce a new member to the family of social networks, which is named Co-Travel Networks, consisting of passengers connected by their co-travel behaviors. We propose a novel method to infer high quality co-travel networks of civil aviation passengers from their co-booking behaviors derived from the PNRs (Passenger Naming Records). In our method, to accurately evaluate the strength of ties, we present a measure of Co-Journey Times to count the co-travel times of complete journeys between passengers. We infer a high quality co-travel network based on a large encrypted PNR dataset and conduct a series of network analyses on it. The experimental results show the effectiveness of our inferring method, as well as some special characteristics of co-travel networks, such as the sparsity and high aggregation, compared with other kinds of social networks. It can be expected that such co-travel networks will greatly help the industry to better understand their passengers so as to improve their services. More importantly, we contribute a special kind of social networks with high strength of ties generated from very close and high cost travel behaviors, for further scientific researches on human travel behaviors, group travel patterns, high-end travel market evolution, etc., from the perspective of social networks.
BibTeX
@misc{lin2013inferring,
  author = {Youfang Lin and Xuguang Jia and Mingjie Lin and Steve Gregory and Huaiyu Wan and Zhihao Wu},
  title = {{Inferring High Quality Co-Travel Networks}},
  year = {2013},
  eprint = {1305.4429},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1305.4429},
}
Figure