A Heterogeneous Information Attentive Network for the Identification of Tourist Attraction Competitors

Abstract

Tourist attraction competition amongst tourist destinations is a crucial com-ponent of a sustainable growth of tourism destinations, and it still deserves appropriate studies to identify them as well as appropriate management solu-tions. Existing studies usually focus on mining tourism locations correla-tions using available statistical data or inference mechanisms applied to tex-tual and cartographical reports. However, a few works apply a combination of qualitative and quantitative approaches, based on multiple contextual characteristics, to infer tourism attraction patterns and competition patterns. Over the past few years, the emergence of social media and Location-Based Services (LBS) in the tourism sector such as geo-tagged reviews, photos, consuming behaviors, and itineraries, provides a new paradigm for extracting and understanding competition among attractions. This research introduces a Heterogenous Information Network (HIN) and Graph Neural Network-based model to capture the complex contextual features for and identifica-tion of attraction competitions. Specifically, three categories of LBS data are processed, extracted, and integrated into a unified HIN, including tourists’ journeys, online text, and spatial attributes. This supports the exploration of significant regularities of attraction competing contexts. The GNN-based model, so-called Competitor-GAT, extract spatial distribution properties and semantic correlations. The experiments applied on a real-world dataset demonstrate the effectiveness of our method.

This research was supported by the National Key Research and Development Program (2022YFB3904200).