Incorporating Both Positive and Negative Association Rules into the Analysis of Outbound Tourism in Hong Kong

Gang Li, Rob Law, Jia Rong, Huy Quan VuJournal of Travel & Tourism Marketing 2010 · DOI

TL;DR

This paper introduces a novel data mining approach that goes beyond traditional association rule mining by incorporating negative association rules to analyze outbound tourism behavior. By considering what people don't do, alongside what they do, the authors provide richer insights for tourism marketers to develop more targeted and cost-effective strategies.

What Problem Does This Paper Address?

Understanding traveler behavior is crucial for the tourism industry. While traditional association rule mining can identify patterns like "people who buy X also buy Y," it often misses equally important negative associations, such as "people who buy X are unlikely to buy Z." This limitation can lead to missed opportunities or inefficient marketing efforts. The authors aim to address this by developing a method that can uncover both positive and negative associations, specifically for analyzing outbound tourism from Hong Kong, to provide more nuanced insights for practitioners.

Methods

The core of this paper is a new data mining method for discovering targeted positive and negative association rules.

  • Targeted Rules: Instead of finding all possible associations, the method focuses on rules related to specific "target" items (e.g., "traveling to Mainland China" or "traveling overseas"). This makes the analysis more efficient and relevant to specific marketing goals.
  • Incorporating Negative Rules: Traditional association rule mining primarily focuses on positive rules (A ⇒ B). This new method explicitly incorporates negative rules (e.g., A ⇒ ¬B, meaning if A occurs, B is unlikely to occur). This is achieved by considering infrequent item-sets and using a modified confidence measure.
  • Conditional-Probability Increment Ratio (CPIR): The authors introduce CPIR as a measure to evaluate the strength and direction (positive or negative) of associations. CPIR is an improvement over traditional confidence measures as it can distinguish between positive and negative dependencies.
  • Leverage Measure: To further refine the identification of promising item-sets, a leverage measure is used to estimate how often an item-set appears independently of the target attribute.

Datasets

The study utilizes three large-scale domestic tourism surveys of outbound pleasure travel in Hong Kong, collected in:

  • Early/mid-2007 (2,025 data samples)
  • 2008 (1,404 data samples)
  • 2009 (1,466 data samples)

The 2007 and 2008 datasets were used for generating rules, while the 2009 dataset was used for validation. The datasets contain demographic information (age, gender, education, income, household size) and travel behavior (travel experience, website usage, past travel to Macau, Guangdong, Mainland China, and overseas).

Key Insights and Ideas

  • The Power of "What Not To Do": The most significant insight is the value of negative association rules. Understanding why certain groups don't travel to specific destinations or engage in certain travel behaviors can be as, if not more, valuable than knowing what they do do. This allows for more precise targeting and avoids wasted marketing spend.
  • Targeted Analysis is Key: Focusing on specific target items (like "traveling overseas") significantly improves the efficiency and relevance of data mining in tourism research.
  • CPIR as a Superior Metric: The CPIR metric offers a more nuanced understanding of associations by quantifying both strength and direction, which is crucial for interpreting negative rules effectively.
  • Demographics and Travel Behavior Interplay: The study reveals complex relationships between demographic factors (like education, income, and travel experience) and outbound travel decisions, highlighting that these factors often interact in non-obvious ways.

Key Results

  • Prevalence of Negative Rules: The analysis found that negative association rules were often stronger (higher CPIR values) than positive ones, particularly in the overall dataset.
  • Key Indicators for Not Traveling:
    • Low education levels, lack of travel website experience, and no prior experience in Guangdong were strongly correlated with not traveling to Mainland China or Overseas.
  • Inexperienced travelers (low travel experience) showed a significantly lower likelihood of traveling to Mainland China or Overseas.
  • Factors Influencing Overseas Travel:
    • Travelers with low education and no website experience were more likely to forgo overseas travel plans, especially if they had low household income.
  • An increase in travel experience, even a small one, could significantly boost the chance of overseas travel.
  • Inexperienced males who hadn't visited Macau or Mainland China were less likely to travel overseas, suggesting a preference for shorter, closer trips before venturing further.
  • Validation: All generated rules were validated with the 2009 dataset, showing high accuracy (CPIR values over 50% in validation).

Limitations and Future Work

  • Single Dataset Applicability: The current method is designed for a single dataset. The authors suggest future work to extend the method to compare multiple datasets (e.g., from different countries) to identify differences or contrasts.
  • Expanding Datasets: Future research should focus on expanding the datasets to include more demographic and behavioral characteristics to test the generalizability of the method.
  • Cultural Factors: The authors acknowledge that cultural factors might influence findings and suggest testing the method in other tourist-generating regions.
  • Broader Applications: The targeted rules mining method could be generalized to other applications beyond tourism, such as demand analysis in various sectors.

Summary updated

Abstract

ABSTRACT This article presents a novel approach to data mining that incorporates both positive and negative association rules into the analysis of outbound travelers. Using datasets collected from three large-scale domestic tourism surveys on Hong Kong residents' outbound pleasure travel, different sets of targeted rules were generated to provide promising information that will allow practitioners and policy makers to better understand the important relationship between condition attributes and target attributes. This article will be of interest to readers who want to understand methods for integrating the latest data mining techniques into tourism research. It will also be of use to marketing managers in destinations to better formulate strategies for receiving outbound travelers from Hong Kong, and possibly elsewhere.
BibTeX
@article{li2010incorporating,
  author = {Gang Li and Rob Law and Jia Rong and Huy Quan Vu},
  title = {{Incorporating Both Positive and Negative Association Rules into the Analysis of Outbound Tourism in Hong Kong}},
  journal = {Journal of Travel \& Tourism Marketing},
  year = {2010},
  doi = {10.1080/10548408.2010.527248},
}

Loading related content...

Figure