[Paper Review] Effective Metagraph-based Life Pattern Clustering with Big Human Mobility Data

Wenjing Li, Haoran Zhang+6 more|arXiv (Cornell University)|Apr 24, 2021

Human Mobility and Location-Based Analysis46 references4 citations

TL;DR

This paper proposes a metagraph-based framework for clustering life patterns from big human mobility data using GPS records, preserving individual features without aggregation. By integrating spatial-temporal similarity, semantics, and frequency into a metagraph structure and applying non-negative matrix factorization, the method efficiently identifies groups with similar mobility behaviors, outperforming traditional approaches in computational efficiency and robustness while revealing region- and period-specific lifestyle patterns.

ABSTRACT

Life pattern clustering is essential for abstracting the groups' characteristics of daily mobility patterns and activity regularity. Based on millions of GPS records, this paper proposed a framework on the life pattern clustering which can efficiently identify the groups have similar life pattern. The proposed method can retain original features of individual life pattern data without aggregation. Metagraph-based data structure is proposed for presenting the diverse life pattern. Spatial-temporal similarity includes significant places semantics, time sequential properties and frequency are integrated into this data structure, which captures the uncertainty of an individual and the diversities between individuals. Non-negative-factorization-based method was utilized for reducing the dimension. The results show that our proposed method can effectively identify the groups have similar life pattern and takes advantages in computation efficiency and robustness comparing with the traditional method. We revealed the representative life pattern groups and analyzed the group characteristics of human life patterns during different periods and different regions. We believe our work will help in future infrastructure planning, services improvement and policies making related to urban and transportation, thus promoting a humanized and sustainable city.

Motivation & Objective

• To identify groups with similar life patterns from large-scale human mobility data without aggregating individual features.
• To model diverse life patterns using a metagraph data structure that captures spatial, temporal, and frequency-based semantics.
• To improve computational efficiency and robustness in life pattern clustering compared to traditional methods.
• To reveal representative life pattern groups and analyze their characteristics across different times and regions.
• To support urban planning, service improvement, and sustainable city development through data-driven insights.

Proposed method

• The framework constructs a metagraph to represent individual life patterns, encoding spatial locations, temporal sequences, and activity frequencies.
• Spatial-temporal similarity is integrated into the metagraph to reflect place semantics, time ordering, and repetition patterns.
• Non-negative matrix factorization (NMF) is applied to reduce the dimensionality of the metagraph representation while preserving structural and semantic information.
• The method avoids data aggregation, maintaining individual-level features for more accurate pattern representation.
• Clustering is performed on the low-dimensional NMF-represented metagraphs to identify groups with similar life patterns.
• The approach is evaluated on millions of GPS records, demonstrating scalability and robustness.

Experimental results

Research questions

• RQ1 How can life patterns be effectively modeled from big human mobility data while preserving individual-level features?
• RQ2 What is the optimal way to integrate spatial, temporal, and frequency-based semantics into a unified data structure for life pattern representation?
• RQ3 How does the metagraph-based approach compare to traditional clustering methods in terms of computational efficiency and robustness?
• RQ4 What distinct life pattern groups emerge across different regions and time periods?
• RQ5 How can the identified life pattern clusters inform urban infrastructure planning and policy-making?

Key findings

• The metagraph-based method successfully identifies distinct life pattern groups with high accuracy and computational efficiency.
• The approach outperforms traditional clustering methods in both speed and robustness, especially in handling noisy or variable mobility data.
• Non-negative matrix factorization effectively reduces dimensionality while preserving meaningful spatial-temporal patterns in individual mobility data.
• The framework reveals region-specific and time-period-specific lifestyle clusters, such as commuting patterns in urban centers and leisure activities in suburban areas.
• The results demonstrate the method’s potential for supporting human-centered urban planning and sustainable city development.
• The study confirms that preserving individual features without aggregation leads to more nuanced and representative clustering outcomes.