Analyzing International Travelers' Profile with Self-Organizing Maps

Gang Li, Rob Law, Jinlong WangJournal of Travel & Tourism Marketing 2010 · DOI

TL;DR

This paper introduces a novel hybrid approach for segmenting international travelers using Self-Organizing Maps (SOMs) for clustering and association rule discovery for profile characterization. This method provides a more objective and robust way for tourism stakeholders to understand traveler segments, leading to better-informed business strategies and marketing efforts.

What Problem Does This Paper Address?

In the competitive tourism industry, understanding traveler profiles is crucial for effective business strategies and customer service. While market segmentation is a well-established practice, existing methods often rely on expert opinions or struggle with high-dimensional data. This paper addresses the challenge of gaining deep insights into the characteristics, motivations, and behaviors of international travelers, particularly for a major destination like Hong Kong, where a comprehensive understanding of diverse traveler segments is needed for strategic planning and resource allocation.

Methods

The core of this research lies in a two-stage hybrid data mining approach:

  1. Self-Organizing Maps (SOMs): SOMs, a type of unsupervised artificial neural network, are employed for dimensionality reduction and clustering. SOMs are particularly useful here because they can map high-dimensional data onto a lower-dimensional (typically 2D) grid while preserving the topological relationships of the input data. This allows for the visualization of complex relationships and the identification of inherent groups (segments) within the traveler data. The paper details the SOM learning process, including weight vector updates and neighborhood functions, to create ordered clusters.
  2. Association Rule Discovery: Once the SOM has identified the segments, an association rule discovery algorithm (specifically mentioning the Apriori algorithm or GRI) is used to automatically characterize the profile of each segment. This method identifies "if-then" rules that describe the distinct patterns and characteristics of travelers within each segment, based on variables not necessarily used in the SOM clustering itself. This provides an objective and semantic way to understand what defines each group.

This hybrid approach is presented as a systematic strategy that combines the visual and clustering capabilities of SOMs with the descriptive power of association rules.

Datasets

The study utilizes a dataset collected from a survey of 1,282 non-transit international visitors at the Hong Kong International Airport in October 2005. The data was gathered through face-to-face interviews using a questionnaire that included demographic information, trip information, and expenditure information. The paper lists 26 variables initially, which are then refined to 15 core variables for segmentation based on their discriminating ability.

Key Insights and Ideas

  • Visualizing Complex Data: SOMs offer a powerful way to visualize high-dimensional traveler data, making it easier for tourism managers to grasp complex relationships and identify distinct traveler groups.
  • Objective Profiling: The integration of association rule discovery provides an objective and data-driven method to characterize the profiles of these segments, moving beyond subjective interpretations.
  • Hybrid Approach Effectiveness: The combination of SOM for segmentation and association rules for profiling is shown to be a robust and systematic strategy for understanding international traveler behavior.
  • Segment-Specific Insights: The research successfully identified 10 distinct traveler segments, each with unique profiles in terms of demographics, motivations, and activities, offering granular insights for targeted marketing.

Key Results

  • 10 Distinct Segments: The SOM successfully segmented international travelers to Hong Kong into 10 distinct groups.
  • Significant Differences: Chi-square tests confirmed significant differences between these segments across various demographic and trip information variables, validating the segmentation's effectiveness.
  • Detailed Segment Profiles: The paper provides detailed profiles for each of the 10 segments, outlining their key characteristics, motivations (e.g., family, discovery, business), and activity preferences (e.g., sightseeing, shopping, visiting museums). For example, segment G9 showed a strong preference for discovering new things but a low interest in business travel, while segment G3 was characterized by full-packaged Chinese visitors with a strong interest in rest and relaxation.
  • Activity Patterns: Analysis revealed that while some activities (like sightseeing and shopping) were popular across most segments, others showed distinct preferences, with certain segments showing significantly different attitudes towards specific activities.
  • Dunn Index Validation: The SOM segmentation achieved a higher average Dunn index compared to K-means and EM algorithms, suggesting more compact and well-separated clusters.

Limitations and Future Work

  • Dataset Size: The authors acknowledge that the dataset size is limited, which restricts the ability to make generalized claims.
  • Generalizability: The study's focus on Hong Kong means that the findings might not be directly transferable to other destinations without further research.
  • Future Research Directions:
    • Applying the SOM method to other tourism sectors and destinations to assess its generalizability.
  • Conducting in-depth qualitative studies with industry practitioners to understand their perspectives on the identified clusters and their associated activities.
  • Performing longitudinal studies to track the consistency of traveler segment outcomes over time.
  • Exploring the use of association rules for more robust profile identification in SOM-based segmentation.

Summary updated

Abstract

It is generally agreed that knowledge is the most valuable asset to an organization. Knowledge enables a business to effectively compete with its competitors. In the tourism context, an in-depth knowledge of the profile of international travelers to a destination has become a crucial factor for decision makers to formulate their business strategies and better serve their customers. In this research, a self-organizing map (SOM) network was used for segmenting international travelers to Hong Kong, a major travel destination in Asia. An association rules discovery algorithm is then utilized to automatically characterize the profile of each segment. The resulting maps serve as a visual analysis tool for tourism managers to better understand the characteristics, motivations, and behaviors of international travelers.
BibTeX
@article{li2010analyzing,
  author = {Gang Li and Rob Law and Jinlong Wang},
  title = {{Analyzing International Travelers' Profile with Self-Organizing Maps}},
  journal = {Journal of Travel \& Tourism Marketing},
  year = {2010},
  doi = {10.1080/10548400903579647},
}
Figure