Understanding attractions’ connection patterns based on intra-destination tourist mobility: A network motif approach

Tourism is a major economic driver globally, particularly urban tourism. The mobility patterns of contemporary tourists are flexible and dynamic, departing from rigid itineraries. Advances in information and communication technologies, specifically location-based services on mobile devices, provide rich data for understanding tourist mobility. Previous research has utilized methods like geographic information systems, time geography, and Markov chains to analyze tourist movement. A prevalent approach aggregates individual mobility data into networks to study the topological structure of attraction systems. Network analysis, a data mining technique, reveals connection patterns among attractions. However, most existing studies rely on descriptive measures, hindering the assessment of the reliability and validity of identified patterns. This study introduces the concept of network motifs—recurring and statistically significant subgraphs—to address this limitation. Network motifs reveal functional properties based on network structure, offering insights into destination connectivity, tourist movement between destinations, and the impact of tourism policies. This study aims to identify the types of motifs forming among attractions, their relationships with specific attractions, and their association with attraction attributes. Suzhou, China, serves as the case study area, leveraging social media data to link network nodes to specific attractions.

Network motifs are recurring patterns in complex networks that occur significantly more often than in randomized networks. They reveal functional properties of the network. Motif discovery methods are categorized as network-centric (enumerating all subgraphs) or motif-centric (searching for specific query graphs). In tourism research, network analysis examines tourist flow networks from inter-destination and intra-destination perspectives. Common metrics include network-based indicators (density, efficiency, etc.) and node-based metrics (degree centrality, closeness centrality, etc.). While network science is applied to tourism, the use of network motifs remains limited. Some studies have explored travel motifs, extending the concept to include temporal and semantic dimensions, but these focus on individual-level mobility rather than aggregated patterns. This study addresses the gap by using network motifs to examine aggregated individual-level tourist mobility.

The study area is Suzhou, China, a city with abundant tourism resources. Data were collected from Sina Weibo, a popular Chinese social media platform, using its API. Data from April 12, 2012, to October 31, 2016, were collected. The data included post ID, user ID, text, pictures, location, and post time. User profile information (registration location, gender, age, etc.) was also gathered. Data preprocessing involved filtering out local users (based on registered location) and users with stays exceeding one month. Only users posting within attractions listed by the Suzhou Tourism Bureau were considered tourists. This resulted in 234,049 microblogs from 54,712 tourists. Tourist trajectories were mapped to create a directed network of attractions. The Kavosh algorithm was used for motif discovery. This algorithm enumerates all k-size subgraphs, uses the NAUTY algorithm to group isomorphic subgraphs, generates random networks for comparison, and identifies significant motifs based on frequency, Z-score, and p-value. Motifs were considered significant if the p-value < 0.01, frequency > 4, and Z-score > 1, using 1000 randomized networks. The study extracted motifs with three and four nodes.

The analysis revealed nine significant motifs: three three-node motifs and six four-node motifs. These motifs were categorized into four classes: chain, mutual dyad, double-linked mutual dyad, and fully connected triad. Three-node motifs accounted for 37.61% of subgraphs, indicating that sequential visits to three attractions are dominant. Four-node motifs comprised 29.67%. Analysis of the highest-degree node in each motif identified Guanqian Street, Jinji Lake, and Pingjiang Road as central attractions. Zhouzhuang acted as a gateway attraction, while Hanshan Temple functioned as a convergence point. Analysis by attraction type (natural, cultural, commercial) and title (5A, 4A, others) revealed that high-degree nodes were predominantly well-known attractions (5A or 4A), frequently cultural or commercial. Chain motifs showed less distinction in attraction type among nodes. Mutual dyads displayed opposing attraction types in nodes B and C. Double-linked mutual dyads showed significant commercial influence in node A for motifs 4 and 5, but less so for motif 6. Fully connected triads exhibited consistent attraction types across nodes, with node A acting as a central hub.

The network motif analysis provides a novel approach to understanding intra-destination tourist mobility. The identified motifs reveal underlying patterns in tourist movement, highlighting the roles of core, transit, and gateway attractions. This granular understanding of attraction function is valuable for destination marketing and planning. Core attractions form the basis of planning, transit attractions benefit from improved transport links, and gateway attractions require enhanced services. The study underscores the importance of prominent attractions in shaping network structure, raising awareness of potential overtourism issues. The findings also help refine group user profiles, allowing for more targeted product development and alternative attraction design.

This study used network motif analysis to examine tourist mobility in Suzhou, identifying nine motifs categorized into four classes. Key attractions were identified, highlighting their roles within the network. The analysis of attraction types and titles further revealed the characteristics of these key attractions. This work offers a new methodological framework for understanding attraction connection patterns, informing destination management strategies. Future research could explore attraction selection mechanisms within the identified motifs and compare findings across different destinations.

The study acknowledges limitations associated with using social media data. Such data can be biased, potentially overrepresenting certain user populations or specific platforms. The data predominantly reflect spatial behavior within a single city. Generalizability might be enhanced by comparing tourism networks across multiple destinations with varying characteristics. Future research should address these limitations to further strengthen the generalizability of the findings.