Interpersonal and Intrapersonal Variabilities in Daily Activity-Travel Patterns: A Networked Spatiotemporal Analysis

Wenjia Zhang, Chunhan Ji, Hao Yu, Yi Zhao, Yanwei ChaiIsprs Int. J. geo inf 2021 · DOI

TL;DR

This paper introduces a novel approach to analyze daily activity-travel patterns by distinguishing between individual (interpersonal) and day-to-day (intrapersonal) variations. By applying network analysis and a multilevel logit model to GPS trajectory data, the study reveals that ignoring intrapersonal variability can lead to biased travel demand models and suggests optimal strategies for collecting activity-travel data.

What Problem Does This Paper Address?

Understanding human mobility patterns is crucial for effective transportation planning, especially as the focus shifts from building more infrastructure to managing travel demand. A key challenge is accounting for the inherent variability in people's daily routines. Traditional studies often focus on interpersonal variability, which describes how different individuals' travel behaviors differ based on their characteristics (e.g., age, income). However, they often overlook intrapersonal variability, the day-to-day fluctuations in an individual's own travel patterns. This oversight can lead to inaccurate models and predictions because assuming a "typical day" might not represent the full spectrum of an individual's behavior over a week or longer. This paper aims to bridge this gap by simultaneously analyzing both interpersonal and intrapersonal variabilities in activity-travel patterns.

Methods

The study employs a two-pronged methodological approach:

  1. Network Analysis for Pattern Mining:
    • The researchers adapt a method proposed by Zhang and Thill [25] to transform individual GPS-based activity-travel trajectories into a two-mode network. This network represents relationships between individuals and their activity-travel events.
  2. This is then converted into an individual-based one-mode network by constructing an adjacency matrix that quantifies the similarity between individuals based on shared activity-travel events (considering temporal overlap and spatial proximity).
  3. The Louvain method, a community detection algorithm, is applied to this network to cluster individuals into groups with similar activity-travel patterns. This effectively identifies distinct "activity-travel patterns" or "communities."
  4. Multilevel Multinomial Logit Model for Variability Analysis:
    • A two-level multinomial logit model is used to analyze the identified activity-travel patterns.
  5. Level 1 accounts for intrapersonal variability by incorporating day-of-the-week variables (e.g., Monday, Tuesday) as independent variables. This helps determine if the likelihood of belonging to a certain pattern varies across different days of the week.
  6. Level 2 accounts for interpersonal variability by including individual socioeconomic attributes (e.g., gender, age, education, income, hukou status) as predictors. This assesses how individual characteristics influence the probability of adopting a particular activity-travel pattern.
  7. The model's dependent variable is the class of detected communities (activity-travel patterns), which is a categorical variable.

Datasets

The study utilizes a one-week activity-travel diary survey conducted in the Shangdi-Qinghe area of Beijing from September to December 2012. This dataset includes:

  • GPS trajectory data: Collected via wearable GPS devices for seven consecutive days for each participant, capturing detailed spatial-temporal information.
  • Interactive website survey data: Collected residents' socioeconomic attributes and daily activity diaries.

The final dataset comprises 3793 person-days from 680 residents across 23 neighborhoods and 19 representative companies.

Key Insights and Ideas

  • Intrapersonal variability is significant: The study strongly suggests that people's daily routines are not static. Ignoring day-to-day fluctuations can lead to a distorted understanding of travel behavior.
  • "Typical day" assumption is problematic: Relying on a single "typical day" to represent an individual's weekly or longer-term travel behavior is likely to introduce bias into travel demand models.
  • Day of the week matters: Specific activity-travel patterns are more likely to occur on weekdays versus weekends, and even specific days within the week show variations. This has implications for data collection strategies.
  • Network analysis reveals distinct patterns: The application of network analysis effectively identifies and visualizes cohesive groups of individuals with similar daily routines, moving beyond simple trip-based analyses.
  • Socioeconomic factors shape patterns: Individual characteristics significantly influence the likelihood of adopting certain activity-travel patterns, highlighting the importance of interpersonal variability.

Key Results

  • Seven major activity-travel patterns were identified: These include "working pattern," "private-affair pattern," "leisure pattern," "working-housework pattern," "other-activities pattern," "school pattern," and "medical-activity pattern."
  • Significant day-to-day variability was found for most patterns: For instance, the "private-affair," "leisure," and "school-related" patterns are more prevalent on weekends, while the "working pattern" is more common on weekdays.
  • Weekday vs. Weekend differences: While most patterns show variations between weekdays and weekends, the study found no significant differences within the weekdays (Monday-Friday) for most patterns after controlling for socioeconomic factors.
  • Optimal data collection strategy: The findings suggest that to accurately capture activity-travel patterns, researchers should ideally collect data for at least three days a week, including one weekday and two weekend days.
  • Interpersonal variability is also significant: The multilevel model confirmed that individual socioeconomic attributes are strong predictors of activity-travel pattern adoption. For example, married individuals are more likely to engage in private-affair activities.

Limitations and Future Work

  • Limited temporal scope: The study relies on one-week of data. The authors acknowledge that this may not capture longer-term (e.g., seasonal) variations in activity-travel arrangements. Future work should aim to collect longer-term datasets, such as mobile phone signaling data.
  • Geographic specificity: The study focused on a typical suburban area in Beijing. The findings might be biased and may not generalize to other urban or rural areas. Future research should investigate and compare findings across different regions.
  • Data availability: While the data is available upon reasonable request, the authors note that some residents only recalled a few days of their activities, leading to some missing data.

Summary updated

BibTeX
@article{zhang2021interpersonal,
  author = {Wenjia Zhang and Chunhan Ji and Hao Yu and Yi Zhao and Yanwei Chai},
  title = {{Interpersonal and Intrapersonal Variabilities in Daily Activity-Travel Patterns: A Networked Spatiotemporal Analysis}},
  journal = {Isprs Int. J. geo inf},
  year = {2021},
  doi = {10.3390/ijgi10030148},
}
Figure