Spatiotemporal Patterns of Tourist Flow in Beijing and Their Influencing Factors: An Investigation Using Digital Footprint

Abstract

Amid ongoing societal development, tourists’ travel behavior patterns have been undergoing substantial transformations, and understanding their evolution has emerged as a key area of scholarly interest. Taking Beijing as a case study, this research aims to uncover the spatiotemporal evolution patterns of tourist flows and their underlying driving mechanisms. Based on digital footprint relational data, a dual-perspective analytical framework—“tourist perception–tourist flow network”—is constructed. By integrating the center-of-gravity model, social network analysis, and regression models, the study systematically examines the dynamic spatial structure of tourist flows in Beijing from 2012 to 2024. The findings reveal that in the post-pandemic period, Beijing tourists place greater emphasis on the cultural connotation and experiential aspects of destinations. The gravitational center of tourist flows remains relatively stable, with core historical and cultural blocks retaining strong appeal, though a slight shift has occurred due to policy influences and emerging attractions. The evolution of the spatial network structure reveals that tourism flows have become more dispersed, while the influence of core scenic spots continues to intensify. Government policy orientation, tourism information retrieval, and the agglomeration of tourism resources significantly promote the structure of tourist flows, whereas the general level of tourism resources exerts no notable influence. These findings offer theoretical insights and practical guidance for the sustainable development and regional coordination of tourism in Beijing, and provide a valuable reference for the spatial restructuring of urban tourism in the post-COVID-19 era.

1. Introduction

As urban functions evolve and the tourism industry advances, analyzing the spatial patterns of urban tourism flows has gained increasing academic importance [1,2]. Concurrently, the global tourism sector is experiencing a paradigm shift driven by digital technologies. China’s 14th Five-Year Plan for Digital Economy Development and Smart Tourism Innovation and Development Action Plan [3,4] have notably accelerated the integration of digital technologies with the tourism sector.

In recent years, the widespread adoption of social media, online booking platforms, and location-based services (LBSs) has profoundly transformed tourists’ travel behavior and generated vast quantities of behavioral data [5,6,7]. These “digital footprints” not only carry spatiotemporal attributes but also convey emotional and cognitive content, thereby offering a novel perspective and a robust data foundation for the study of tourist flows. Among them, online travel diaries, as a form of digital footprint, contain both perceptual and trajectory information, making them especially suitable for exploring the interplay between tourist sentiment and mobility.

As a megacity of political, cultural, and economic significance, Beijing exhibits tourism behavior driven by diverse and multidimensional factors [8,9], including the substantial appeal of its historical heritage (e.g., over 19 million annual visitors to the Palace Museum), the symbolic status of Tiananmen Square, and the spatial transformations resulting from coordinated development in the Beijing–Tianjin–Hebei region. The city’s policy agenda to establish itself as an “International Consumption Center” provides further policy support and empirical significance for investigating its tourism spatial patterns.

Despite its importance, the structural characteristics and underlying mechanisms of tourist flows between different scenic spots within Beijing remain underexplored. Most existing studies focus primarily on individual attractions or tourist satisfaction, lacking a network-based approach that quantifies interaction patterns and dynamic evolution processes [10,11].

Therefore, this study aims to construct a dual-perspective analytical framework integrating “tourist perception” and “tourist flow networks.” Specifically, it combines gravity center models, SNA, and QAP regression analysis to examine the spatiotemporal evolution and influencing factors of tourist flows in Beijing from 2012 to 2024. Online travel diary data are used as the primary source of tourists’ digital footprints, enabling the identification of tourist sentiments and perceptual patterns. SNA is employed to reveal the spatial structure and network positioning of attractions, while QAP regression is used to uncover the driving mechanisms of the tourist flow network. By bridging individual-level perception with structural-level mobility patterns, this research advances theoretical understanding of urban tourism flows and offers actionable insights for constructing efficient, balanced, and sustainable tourism spatial systems in megacities.

2. Literature Review

2.1. Digital Footprints in Tourism Research

Digital footprints are defined as datasets with spatiotemporal identifiers generated by tourists through digital devices such as mobile phones, social media platforms, and navigation applications. These data include information on location, duration of stay, movement trajectories, and interaction content, enabling precise characterization of tourists’ dynamic behaviors in urban spaces, particularly in complex flow relationships [12,13]. Online travel diaries, a critical subset of digital footprints, are typically published voluntarily by users on social media or tourism platforms and comprise text, images, comments, and other content. These diaries capture both tourists’ perceptions and movement trajectories, thus overcoming the limitations of traditional textual data in sentiment analysis and spatial interpretation [14].

Compared to GPS traces or check-in data that provide only spatial coordinates, online travel diaries are more effective for examining tourists’ emotional orientations, travel preferences, and behavioral patterns [15,16]. Specifically, Mou et al. employed such data to depict the spatial distribution of tourist flows, demonstrating the analytical potential of digital footprints in tourism spatial studies [2].

2.2. Spatial Network Structure in Tourist Flow

Tourist flow is defined as the spatiotemporal movement of individuals from points of origin to destinations, emphasizing both dynamic processes and spatial interrelationships [17]. The core of tourist flow research centers on identifying the spatiotemporal evolution of tourist behaviors and their associated socioeconomic effects. While numerous studies have applied methods such as Markov chains, clustering algorithms, and correlation analysis to investigate factors influencing tourist flows [18,19,20], most have focused on the macro level and lack systematic modeling of intra-city attractions and their network structures [21].

Traditional statistical approaches exhibit notable limitations in capturing the structural and interactive dimensions of tourist flows, making it challenging to reveal the intensity, directionality, and temporal dynamics of attraction-to-attraction linkages [22,23]. Although graph-based and spatial visualization methods enhance interpretability, they fall short in capturing complex multivariate interactions [24]. In particular, the post-pandemic era demands new analytical perspectives and frameworks to reconfigure resilient and efficient urban tourism networks [2,25].

In recent years, the integration of social network analysis (SNA) with gravity center models has gained traction in tourism network research [26,27]. Rooted in sociological theory, SNA emphasizes relational structures among nodes and facilitates the quantification of connection strength, centrality, dependency, and structural changes among tourist attractions [28]. Compared with clustering or correlation-based methods, SNA better captures the heterogeneity and nonlinear evolution within tourist flow networks, thereby offering enhanced explanatory capacity for understanding urban tourism system organization [2].

Additionally, the Quadratic Assignment Procedure (QAP) regression model—a nonparametric statistical approach tailored to relational data—addresses estimation bias caused by spatial autocorrelation and nested network dependencies [29]. Extensively applied in fields such as organizational behavior, regional development, and transportation networks, QAP is particularly appropriate for uncovering the causal relationships between node attributes and structural features in tourism networks shaped by multifactorial influences [30]. Nevertheless, research leveraging SNA and QAP remains predominantly focused on national or regional scales, with limited attention to the fine-grained intra-city tourism flow networks structures of megacities [31,32].

3. Materials and Methods

3.1. Study Area

This study focuses on the entire administrative area of Beijing (comprising 16 district-level divisions with a total area of 16,410 km²) as the study region (Figure 1) and takes major tourist attractions within the city as the research objects. First, Beijing boasts diverse tourism resources. As of December 2024, the city had 8 world cultural heritage sites, 10 AAAAA-level scenic areas, and 5 national-level tourist and leisure blocks. Its rich cultural heritage, combined with a modern urban aesthetic, has made it a prominent global tourism destination [33]. The wide variety of scenic types also provides a solid basis for comparative analysis of the roles and strengths of different tourism nodes within the network. Second, Beijing is one of the top destinations in China in terms of tourist volume. In 2023, Beijing received approximately 329 million tourists, representing an 80.2% increase compared to the previous year [34]. This massive tourist flow offers a solid foundation for analyzing the spatiotemporal aggregation of tourism digital footprints. Finally, the “Opinions on the Implementation of Measures to Promote the High-Quality Development of Beijing’s Tourism Industry,” issued in October 2024, outlines the strategic direction for the sector’s future development and reflects Beijing’s commitment to positioning tourism as a new engine for economic growth. Therefore, an in-depth analysis of tourist flows can provide empirical support for the formulation of future tourism management strategies.

3.2. Dataset and Preprocessing

• Data collection

This study utilized online travel diaries obtained from Qunar (https://www.qunar.com/) as the primary source of digital footprint data. Qunar.com is China’s leading travel search engine and the largest platform for sharing online travel diaries. It has a diverse user group including celebrities, travel experts, and ordinary netizens. According to Qunar’s official website, it has more than 200 million transaction users and continues to grow at a rate of more than 10 million new transaction users per year. In addition, data from Qunar.com have been widely employed in research on destination image, tourist rating behavior, and tourist mobility [35,36]. The platform offers an intelligent travel itinerary editing tool. Users can designate points of interest (POIs) within their travel routes and generate visual travel routes using the location-based service (LBS) module [37]. However, this functionality is often overlooked in existing literature [2]. POIs are digitally encoded in the source code of the travel diaries. By matching the POI identifiers with Qunar’s attraction database, detailed information about the visited locations can be extracted, providing data support for studying the center of gravity of tourist flow, network structure, and traffic matrix. A total of 2893 travel diary entries—including user ID, diary ID, departure date, travel duration, and POI visitation sequences—shared by tourists on Qunar.com between 2012 and 2024 served as the initial dataset.

Additionally, data on the list of tourist attractions in Beijing, including their official ratings, were obtained from the official website of the Ministry of Culture and Tourism of China (https://www.mct.gov.cn/), while hotel POI data were sourced from the Resource and Environmental Science and Data Center of the Chinese Academy of Sciences (https://www.resdc.cn/). The hotel POI data were primarily used in Section 3.3.3 for calculating the tourism reception service indicator in the analysis of influencing factors. Considering policy orientations, market dynamics, and the phased impact of the COVID-19 pandemic, this study categorizes the data from 2012 to 2024 into three distinct periods for analysis: 2012–2015, 2016–2019, and 2020–2024.

2.

Data preprocessing

To eliminate errors and logical inconsistencies in the online travel diary data, this study adopts a four-step data cleaning process. The process includes regional clipping, scenic spot consolidation, data deduplication, and the removal of isolated points [2].

(1) Spatial filtering of the study area. Some travel diaries include destinations outside Beijing or record scenic spots beyond the defined study area due to user input during editing. Therefore, the first step involves removing records of scenic spots located outside the study area. This step ensured that all data were closely related to the research focus and minimized the influence of irrelevant records on the results. (2) Merging of subordinate attractions. Users often document smaller attractions nested within larger ones, such as the Seventeen-Arch Bridge within the Summer Palace. These smaller POIs were merged with their corresponding larger attractions based on established hierarchical relationships. This step simplified the data structure and reduced redundancy while preserving data integrity. (3) Removal of duplicate entries. If the same POI identifier appears consecutively within a single travel diary, it is assumed that the user did not travel between locations. Accordingly, redundant POI entries were removed to ensure the uniqueness and independence of the data. (4) Elimination of isolated points. To facilitate subsequent correlation analysis and tourism flow network construction, entries containing fewer than two scenic spots were excluded from the dataset. These entries typically lack sufficient attraction information to reflect actual movement patterns and were thus removed to improve data validity.

Based on the preprocessed data, we conducted a preliminary identification of the number of travel days and travel periods for tourists, as shown in

Table A1. Proportion of tourists by length of stay in Beijing.

Number of Travel Days	2012–2015	2016–2019	2020–2024
1	66%	62%	60%
2	16%	15%	12%
3	6%	7%	9%
4	4%	3%	5%
5	3%	4%	5%
6	1%	2%	3%
7	1%	2%	2%
>7	3%	5%	4%

and

Table A2. Proportion of tourists across different holidays in Beijing.

Travel Period	2012–2015	2016–2019	2020–2024
Weekend	76.86%	64.95%	52.32%
Labor Day	7.08%	9.94%	12.56%
National Day	11.43%	12.71%	16.17%
Spring Festival	0.12%	3.11%	7.45%
Qingming Festival	0.49%	2.60%	5.05%
Dragon Boat Festival	0.24%	3.29%	4.03%
New Year’s Day	1.98%	1.90%	1.61%
Others	1.80%	1.50%	0.81%

. The analysis results further support the reliability of the dataset.

3.3. Methods

Based on tourists’ digital footprint data, this study integrates traditional quantitative methods with social network analysis to comprehensively examine the spatial patterns of tourism flow in Beijing. First, tourists’ emotional tendencies and tourism perceptions are assessed using online travel diary data. Second, the gravity center model and flow frequency statistics are applied to analyze the flow characteristics among tourist attractions in Beijing. Third, SNA indices are employed to evaluate the spatial connectivity of Beijing’s tourism network and to uncover the functional roles and connectivity disparities among attractions. Finally, the underlying drivers of the tourism flow network structure in Beijing are examined using the QAP.

3.3.1. Center-of-Gravity Model

As a classical concept in physics, the “center of gravity” originally refers to the point at which the resultant gravitational force acts on an object [38]. With advancements in spatial analysis theory, this concept has been extended to disciplines such as economic geography and environmental science, becoming a critical tool for analyzing the spatial dynamics of complex systems. Through the development of “social–spatial” coupling analysis frameworks, scholars have introduced derivative concepts such as ecological, population, and economic centers of gravity, enabling the evaluation of spatial migration patterns of various elements during regional evolution [39].

To intuitively illustrate the dynamic characteristics of tourist flow distribution, this study examines the geometric center of gravity and spatial evolution of tourist flows in Beijing.

Step 1. Tourist attraction data from Beijing were matched with POI information extracted from Qunar.com online travel diary data to obtain tourist flow intensity and geographic coordinates; these data were subsequently preprocessed.

Step 2. A gravity model was constructed using ArcGIS 10.8 software, where the tourist flow of each tourist attractions served as a weight in calculating its weighted spatial coordinates.

Step 3. The weighted average method was used to calculate the annual centroid coordinates of all tourist attractions. The corresponding formula is provided in Equation (1). According to Liu et al. [26], the specific formula is as follows:

$$\overline{X}={\displaystyle \frac{{\displaystyle {\sum }_{i=1}^n{x}_i{w}_i}}{{\displaystyle {\sum }_{i=1}^n{w}_i}}},\overline{Y}={\displaystyle \frac{{\displaystyle {\sum }_{i=1}^n{y}_i{w}_i}}{{\displaystyle {\sum }_{i=1}^n{w}_i}}}$$

(1)

where $\overline{X}$ and $\overline{Y}$ denote the longitude and latitude of the center of gravity of tourist attractions in Beijing, respectively; x_i and y_i represent the coordinates of each attraction, respectively; i is the i-th tourist attraction; n is the total number of attractions; and w_i denotes the weight of tourist attraction i. The weight is reflected by the intensity of tourist flow, that is, the number of times the attraction appears in the online travel diary data.

Step 4. ArcGIS software was used to visualize the temporal trajectory of centroid shifts. By comparing centroid coordinates across different years, the spatiotemporal evolution of the tourism gravity was analyzed.

3.3.2. Social Network Analysis

SNA is a sociological method used to analyze the structural relationships among social actors. It serves as a standard tool for exploring interactions, structural characteristics, and element mobility among different nodes within a network [40]. SNA has been widely applied across various disciplines, including economics, management, and geography [28,30,41].

This study conceptualizes tourist attractions in Beijing as nodes within the tourism flow network. Based on a binary matrix of tourist flows between attractions, an association network is constructed to examine the relationships among them. Drawing on social network theory, the study emphasizes three dimensions of node-level structure: degree centrality, closeness centrality, and betweenness centrality, to develop a tourism flow network evaluation index system for Beijing’s tourist attractions. The calculation formulas and characteristics of each metric are presented in

Table 1. Evaluation indicators of tourist flow network in Beijing.

Indicator	Formula	Description	Source
Degree centrality	$DC(i)={\displaystyle {\sum }_j^n{r}_{ij}}$ rij: the flow from node i to node j	Measures the degree of direct correlation between nodes and other nodes; the higher the degree centrality is, the higher the direct correlation between the node and other nodes is, and the more important the position in the tourist flow network is.	[42]
Closeness centrality	$CC{(i)}^{-1}={\displaystyle {\sum }_j^n{g}_{ij}}$ gij: the shortest path distance between node i and node j	Measures the proximity of nodes to other nodes; the higher the closeness centrality is, the higher the closeness between the node and other nodes is, and the better the tourist flow accessibility between nodes is.	[28,41]
Betweenness centrality	$\begin{array}{l}BC(i)={\displaystyle {\sum }_j^n{\displaystyle {\sum }_k^n\frac{g_{jk}(i)}{g_{jk}}}},\\ j\ne k\ne i,j<k\end{array}$ gjk(i): the number of paths from node j to node k through node i; the number of paths from node j to node k	Measures the degree of control of nodes over other nodes; the higher the betweenness centrality, the higher the degree of control of the node over other nodes, and the stronger the role of the bridge in the tourist flow network.	[26,42]

.

3.3.3. Influencing Factor Analysis

(1)

Selection of Influencing Factors

The formation of tourism network structures results from the interplay of multiple factors. This study primarily investigates the influence of objective destination factors on the formation of tourism flow networks. Given the spatiotemporal evolution characteristics of tourism flow in Beijing, the study adheres to the principles of scientific rigor, systematic structure, and data accessibility in indicator selection. Drawing on a comprehensive review of both domestic and international literature [43,44,45], six core influencing factors were selected across four key dimensions: policy orientation, service efficiency, resource endowment, and information coupling (

Table 2. Influencing factors of tourism flow structure in Beijing.

Factors	Description	Reference
Policy orientation	Number of times each attraction appears in relevant policy documents on government websites	[46]
Tourist satisfaction	Ratings of each attraction on Qunar.com	[47]
Tourism reception	Number of hotels within 5 km of the attraction	[2]
Tourism resource level	Grade of tourist attraction (A−AAAAA)	[48]
Tourism resource concentration	Number of other tourist attractions within 5 km of the attraction	[48]
Tourism information search	The average value of the Baidu Index	[25]

).

Policy orientation shapes the spatial structure of tourism through administrative resource allocation [46]. The frequency with which each tourism node appears in official policy documents from the website of the Beijing Municipal Bureau of Culture and Tourism was calculated, and a corresponding difference matrix was subsequently constructed. Considering that in mature scenic areas such as the Forbidden City and the Summer Palace, historically accumulated tourist flows may, in turn, reinforce sustained policy attention (e.g., maintenance and security efforts), forming a positive “flow–policy” feedback loop, we classified policy documents into two categories—“planning-oriented” and “maintenance-oriented”—to mitigate the potential impact of reverse causality in the analysis.

Tourist satisfaction serves as a key indicator for evaluating tourism experiences, and online word of mouth significantly influences destination selection [47]. A difference matrix was constructed using user ratings for each tourism node on Qunar.com. This matrix captures both inter-node variation in tourist satisfaction and provides a broader understanding of tourist behavior and destination choice.

Tourism reception services play a critical role in enhancing destination competitiveness [2]. To quantify this factor, each tourism node was used as a center point to count the number of hotels within a 5 km radius. A difference matrix was then constructed to quantify disparities in reception service capacity.

The quality and classification of tourism resources significantly influence the spatial distribution and intensity of tourist flows [48]. In this study, the official tourist attraction rating system was used as the evaluation standard. Each tourist attraction was assigned a coefficient from 1 to 5 according to its rating (e.g., A-rated attractions were assigned a value of 1, AA-rated spots a value of 2, and so on up to AAAAA). Based on the assigned values, a resource attractiveness matrix was constructed.

Nodes with high tourism resource agglomeration are more likely to evolve into distribution hubs within the tourism flow network. This reflects not only their resource endowment but also their locational advantages [48]. In this study, each tourist attraction was taken as the center of a circle with a radius of 5 km. Using the buffer analysis tool in ArcGIS 10.8 software, a spatial scanning window was constructed. The number of A-level tourist attractions within each scanning window was then counted, based on which an agglomeration coefficient was calculated for each tourist attraction.

Information search behavior in the digital age has reshaped tourism decision-making patterns [25]. This study introduces the Baidu Index as a proxy variable for online attention and collects search frequency data for each scenic spot. The average Baidu Index from 2012 to 2024 is used as the indicator of information search intensity for each node.

(2)

Standardization

Since the selected influencing factors vary in dimension and scale, direct comparison is not feasible [49]. To ensure the reliability and validity of the calculation results, this study applies the range standardization method to normalize the original indicator data prior to constructing the difference matrix for each factor. The standardization formula is as follows:

$$X_{ij}^{\prime }={\displaystyle \frac{X_{ij}-X_{\min }}{X_{\max }-X_{\min }}}$$

(2)

where $X_{ij}^{\prime }$ is the dimensionless value; X_ij is the j-th indicator value for node i in year h; X_min and X_max are the minimum and maximum values of the j-th factor, respectively. Standardization unifies the value range of all variables to [0, 1], thereby enhancing the accuracy and comparability of subsequent analysis.

4. Results

4.1. Evolution of Tourists’ Perceived Characteristics

Online travel diaries are tourists’ reviews and summaries of the entire trip after the trip. They provide clear and objective descriptions of personal experiences and effectively convey emotional responses. This study employs the ROSTCM6 text analysis tool to extract semantic emotions from online travel diaries of tourists in Beijing between 2012 and 2024 and conducts fine-grained emotion measurement through a three-dimensional emotion intensity matrix (negative/neutral/positive × general/moderate/high) (

Table 3. Sentiment analysis results.

Emotional Type	Emotional Segmentation	2012–2015		2016–2019		2020–2024
		Frequency	Percentage	Frequency	Percentage	Frequency	Percentage
positive	general	29	5.38%	6	2.14%	4	2.29%
	moderate	29	5.38%	16	5.71%	3	1.71%
	high	350	64.94%	233	83.21%	158	90.29%
	total	408	75.7%	255	91.07%	164	94.29%
negative	general	18	3.34%	4	1.43%	2	1.14%
	moderate	13	2.41%	4	1.43%	2	1.14%
	high	28	5.19%	3	1.07%	3	1.71%
	total	59	10.94%	11	3.93%	7	3.99%
neutral	total	72	13.36%	14	5.00%	3	1.71%

). The results indicate a polarization in the intensity of tourists’ emotional tendencies.

Positive emotions exhibit an evolutionary pattern of “low-intensity contraction to high-intensity expansion.” Between 2012 and 2024, the proportions of general and moderate positive emotions declined from 5.38% each to 2.29% and 1.71%, respectively, while the proportion of highly positive emotions surged from 64.94% to 90.29% (Table 3). Negative emotions showed an overall contraction, declining from 10.94% to 3.99%. The sustained rise in highly positive emotions not only affirms the demonstration effect of Beijing but also offers practical insights for reconstructing the emotional value of urban tourism in the post-pandemic era.

High-frequency word analysis effectively reveals the dynamic relationship between tourism experience evolution and consumption upgrading (Figure 2). The study finds that from 2012 to 2024, tourism experiences have shifted from “natural perception” to “cultural decoding.” Between 2012 and 2015, tourism activities were predominantly nature-oriented. The prominence of keywords such as “self-driving,” “skiing,” and “food” reflects the rise of the leisure and vacation market driven by improvements in transportation networks. From 2016 to 2019, tourist perceptions became increasingly culture-oriented. The semantic density of “history” and “art” increased, accompanied by the emergence of symbolic consumption related to “thermae,” “life,” and “style,”. Between 2020 and 2024, the co-occurrence intensity of the “culture–tourism–history” triad increased. This semantic focus further underscores growing tourist interest in local cultural characteristics and artistic experiences.

4.2. Tourist Flow Characteristics

4.2.1. Gravity Center of Tourist Flows

Figure 3 illustrates that the gravity center of Beijing’s tourist flows has remained relatively stable over the years, concentrated within an area approximately 18 km in diameter, located northeast of the Beijing Olympic Park. Between 2012 and 2015, the gravity center shifted northward, primarily due to the development of the northern eco-tourism zone—comprising rural tourism and forest health tourism—and the optimization of Beijing’s transportation network. From 2016 to 2019, the gravity center remained fixed northeast of Olympic Forest Park. During this period, the cultural heritage in-depth tour (along the Forbidden City–Summer Palace axis) and the suburban leisure tour (Wenyu River Greenway) demonstrated functional complementarity. Between 2020 and 2024, the construction of the Universal Studios complex prompted the gravity center to move southward, confirming the reshaping impact of major cultural and tourism projects on the urban tourism system. Consequently, with shifts in the tourist flow gravity center, relevant authorities should promptly adjust and optimize tourism services and infrastructure to enhance the overall tourism experience.

4.2.2. Frequent Directions of Tourist Flows

The frequency of tourist flow directions reflects the connectivity between scenic spots. Using effective tourism digital footprint data from Beijing spanning 2012 to 2024, this study analyzes the high-frequency tourist flow direction network through threshold screening (Figure 4). The findings reveal that across all three periods, the Tiananmen–Palace Museum area consistently represents the peak tourist flow directions. A comparative analysis reveals that during 2012–2015 and 2016–2019, significant linkages were observed among historical and cultural blocks such as South Luogu Lane and Shichahai, underscoring tourists’ strong interest in Beijing’s hutongs and commercial areas. The frequency of tourist flows directions connecting Yuanmingyuan–Summer Palace, Tsinghua University–Peking University, and Badaling Great Wall–Ming Tombs also remained strong, indicating the appeal of royal gardens and top universities during this period, with surrounding tours serving as important diversion channels. In addition, compared to the 2012–2015 period, the number of connections involving the Forbidden City increased significantly during 2016–2019 and 2020–2024, reinforcing its status as a core attraction in Beijing’s tourism system. Notably, in the 2020–2024 period, major public health events prompted a spatial reorganization of tourist flows; however, top universities maintained strong tourist attraction, while the connectivity of traditional commercial blocks declined. Meanwhile, Universal Studios emerged as a new tourism node. During this period, the number of connections involving the Palace Museum and Tiananmen Square continued to show a steady upward trend.

4.3. Network Structure Analysis

4.3.1. Spatial Correlation Network of Tourist Flows

Based on online travel diary data, this study constructed a tourism flow matrix among scenic spots and quantified the flow relationships between tourists at each node. ArcGIS 10.8 software was employed to visualize Beijing’s tourism flow network. To facilitate comparison across different years, visit frequencies of attractions and the spatial connection strengths of tourist flows were uniformly categorized into three levels—low, medium, and high—using the natural breakpoint method (Figure 5). Overall, from 2012 to 2024, a general spatial connectivity existed between tourist attractions in Beijing, with a trend toward increasing closeness.

Between 2012 and 2015, connections between scenic spots were weak and concentrated primarily around the Palace Museum, Badaling Great Wall, Temple of Heaven, and Summer Palace. As a core attraction, the Palace Museum maintained the most intensive connections with other sites. From 2016 to 2019, the network structure among tourist attractions showed significant growth and strengthening of connections, especially within the central area. Between 2020 and 2024, influenced by the public health incident, the connections among Beijing’s tourist attractions exhibited a slight decline, although the overall network structure remained stable. Notably, the opening of Universal Studios enhanced the spatial connectivity of the tourism network in eastern Beijing.

4.3.2. Centrality Analysis of Tourist Flow Network

To facilitate both horizontal and vertical comparisons of the status and influence of each tourist node within the network, this study utilized UCINET software (version 6.661) to compute and analyze the structural characteristics of Beijing’s tourism network. By comprehensively assessing degree centrality, closeness centrality, and betweenness centrality for each attraction, the study identified 22 key nodes within the tourist flow network (

Table 4. Calculation results of the node centrality in Beijing for 2012–2024.

Tourist Attraction (Node)	2012–2015					2016–2019					2020–2024
	Out-Degree Centrality	In-Degree Centrality	Out- Closeness	In- Closeness	Betweenness Centrality	Out-Degree Centrality	In-Degree Centrality	Out- Closeness	In- Closeness	Betweenness Centrality	Out-Degree Centrality	In-Degree Centrality	Out- Closeness	In- CLOSENESS	Betweenness Centrality
Palace Museum	41	34	2.568	3.047	4.71	42	32	3.047	3.709	7.389	38	30	5.366	5.647	6.209
Tiananmen	22	17	2.559	3.024	3.575	24	21	3.024	3.715	3.664	36	27	5.371	5.664	7.382
Summer Palace	34	44	2.582	3.045	9.279	23	22	3.045	3.733	7.957	26	28	5.387	5.67	9.512
Temple of Heaven	26	21	2.578	3.042	9.69	16	17	3.042	3.706	3.391	22	25	5.382	5.641	8.026
Badaling Great Wall	30	21	2.572	3.044	7.199	19	22	3.044	3.724	5.601	17	15	5.35	5.635	5.537
Yuanmingyuan	22	21	2.56	3.026	2.72	14	19	3.026	3.702	3.855	16	17	5.345	5.601	5.082
Tsinghua University	16	20	2.557	3.019	2.13	17	11	3.019	3.659	7.842	12	15	5.283	5.55	0.913
National Museum	4	5	2.554	2.999	1.391	15	17	2.999	3.698	3.258	11	13	5.304	5.612	3.051
Universal Studios	-	-	-	-	-	-	-	-	-	-	10	12	5.335	5.601	3.802
Peking University	24	18	2.555	3.016	0.445	16	14	3.016	3.704	7.705	9	11	5.228	5.533	0.878
South Luogu Lane	36	33	2.574	3.04	3.642	27	30	3.04	3.729	5.537	9	8	5.309	5.55	0.846
Prince Gong’s Mansion	14	16	2.562	3.031	0.652	15	21	3.031	3.718	8.308	8	10	5.35	5.578	1.765
Shichahai	17	21	2.569	3.03	2.604	19	14	3.03	3.688	3.527	8	9	5.258	5.595	2.282
Olympic Park	14	12	2.568	3.033	3.597	18	15	3.033	3.715	5.763	8	8	5.258	5.572	5.09
Qianmen	18	19	2.571	3.029	3.402	16	14	3.029	3.702	4.863	7	4	5.278	5.522	2
Beihai Park	19	18	2.565	3.036	2.32	12	10	3.036	3.715	4.399	6	7	5.294	5.517	2.152
Dashilan	5	8	2.558	3.004	1.187	7	11	3.004	3.686	2.328	6	6	5.273	5.556	1.823
Yonghe Palace	12	12	2.558	3	3.651	10	3	3	3.661	1.264	5	8	5.283	5.528	1.588
Zoo	16	17	2.566	3.007	4.234	4	7	3.007	3.679	3.645	5	2	5.223	5.489	1.249
Mutianyu Great Wall	5	5	2.547	3.009	2.945	5	13	3.009	3.675	0.077	4	5	5.243	5.533	0.821
798 Art District	16	16	2.563	3.027	5.131	11	12	3.027	3.702	1.116	3	2	5.116	5.473	0.085
Ming Tombs	17	18	2.555	3.036	5.339	12	9	3.036	3.693	6.154	2	2	5.135	5.413	0.16

). The top five nodes are the Palace Museum, Tiananmen Square, the Summer Palace, the Temple of Heaven, and the Badaling Great Wall.

Based on their centrality metrics from 2020 to 2024, the key nodes were classified according to their respective positions in the three centrality indicators. Yuanmingyuan, Tsinghua University, and the National Museum exhibit a comparative advantage in degree centrality, indicating their strong attraction capacity and consistent ability to draw tourist flows. Attractions such as Prince Gong’s Mansion, Universal Studios, and Nanluoguxiang show a comparative advantage in closeness centrality, reflecting their high accessibility and capacity to enhance flow connectivity with other nodes. Sites such as the Olympic Park, Yuanmingyuan, and Shichahai demonstrate a comparative advantage in betweenness centrality, signifying their critical role as intermediaries that channel and redistribute tourist flows and efficiently absorb overflow from other nodes.

4.4. Influencing Factors of Tourist Flow Structure

This study employs the QAP correlation analysis function in UCINET software to examine the relationship between Beijing’s tourism flow structure matrix and various influencing factor matrices [50]. By conducting 5000 random permutations, the study quantifies the influence of six key factors on the structure of tourism flows in Beijing. As shown in

Table 5. Analysis of factors affecting the network structure of tourism flows in Beijing.

Influencing Factors	Correlation Analysis		Regression Analysis
	Correlation Coefficient (R)	Significance Level (P1)	Regression Coefficient (β)	Significance Level (P2)
Government policy orientation	0.537 ***	0.000	0.329 ***	0.000
Tourist satisfaction	0.106 ***	0.000	−0.010	0.433
Tourism reception services	−0.02	0.439	-	-
Tourism resource level	−0.269 ***	0.000	−0.130 **	0.031
Tourism information search	0.679 ***	0.000	0.238 ***	0.000
Tourism resource concentration	0.301 ***	0.003	0.220 ***	0.000

Note: ***, ** represent significance levels of 0.01, 0.05, respectively.

, there are significant correlations between the tourism flow structure and factors such as government policy orientation, tourist satisfaction, tourism information search intensity, tourism resource level, and resource agglomeration (P₁ < 0.05). Specifically, government policy support (R = 0.537, P₁ < 0.01) and tourism information search volume (R = 0.679, P₁ < 0.01) exhibit strong positive correlations, indicating that both policy support and online attention substantially drive the evolution of tourism flow structures. However, we found that planning-oriented policies are significantly positively correlated with the tourism flow structure matrix, while the correlation with maintenance-oriented policies is relatively weak. This supports the assumption of a policy-driven rather than purely result-driven mechanism. The agglomeration of tourism resources (R = 0.301, P₁ < 0.05) is moderately positively correlated, likely because areas with dense tourism resources generate scale and synergy effects, attracting more visitors. In contrast, tourism resource level (R = −0.269) and tourism reception services (R = −0.02) are negatively correlated; however, the latter did not reach statistical significance.

Based on the above correlation analysis, tourist reception services were excluded from further modeling due to a lack of statistical significance. Ultimately, five core explanatory variables—tourism information search intensity, resource level, resource agglomeration, tourist satisfaction, and policy orientation—were identified to construct the matrix regression model (Table 5). The regression results indicate that the coefficients (β) for government policy support, tourism information search, and resource agglomeration were 0.329, 0.238, and 0.220, respectively, all of which significantly contributed to the formation and development of the tourism flow structure in Beijing (P₂ < 0.05). Conversely, the coefficient for tourism resource level was negative (β = −0.130), primarily because several popular destinations—such as Universal Studios, Tsinghua University, Peking University, and the 798 Art District—lack official grading. Additionally, traffic congestion near high-level attractions, especially during peak seasons, often leads tourists to opt for alternative destinations [2]. Tourist satisfaction did not reach statistical significance, suggesting its direct influence on tourism flow structure is limited.

4.5. SWOT Synthesis of the Observed Tourist Flow Patterns

To translate the empirical findings of this study into managerial implications, a SWOT analysis was conducted to summarize the above results [51]. This analysis distinguishes between internal attributes of the tourist destination (strengths and weaknesses) and external environmental factors (opportunities and threats), as shown in

Table 6. SWOT analysis for the tourist flow patterns in Beijing.

	Positive	Negative
Internal	Strengths: Dominance of positive emotions; Stable network structure.	Weaknesses: The gravity center of tourist flow is concentrated.
External	Opportunities: Strong government policy support; High online visibility.	Threats: The distribution of tourism resource grades is uneven.

.

In terms of strengths, tourist sentiment in Beijing is predominantly positive, with the proportion of high-intensity positive emotions increasing from 64.94% in 2012 to 90.29% in 2024. This demonstrates a high level of tourist satisfaction and approval of the tourism experience in Beijing. Therefore, this strength should be sustained, and the positive perception of Beijing’s tourist attractions should be actively reinforced through social media platforms. Additionally, the overall tourist flow network structure remains stable, with the top five nodes consistently being the Forbidden City, Tiananmen Square, the Summer Palace, the Temple of Heaven, and the Badaling Great Wall.

As for weaknesses, the gravity center of tourist flows is overly concentrated within an 18 km radius northeast of the Olympic Park, which may lead to traffic congestion around popular attractions. This is particularly problematic during peak seasons, when such congestion may prompt visitors to choose alternative destinations. To address this issue, increasing the number of attractions with relatively high centrality and enhancing the appeal of peripheral nodes can improve accessibility for tourists and strengthen their connectivity within the network. For instance, the development of the Universal Studios complex has caused a southward shift in the tourism gravity center, offering both theoretical and practical guidance for transforming this weakness into a potential strength.

Regarding opportunities, government policy support and online visibility have significantly influenced the structural evolution of tourist attractions. This suggests that well-designed policies can effectively encourage and direct tourism stakeholders to invest in and develop new attraction nodes, thereby influencing the spatial configuration of tourism flows. Furthermore, online attention—as measured by tourism information search volume (R = 0.679, P₁ < 0.01)—has a pronounced impact on structural evolution. Therefore, marketing strategies such as short videos and social media promotions could be employed to redirect visitor flows toward peripheral nodes.

In terms of threats, the uneven distribution of tourism resource grades undermines the structural balance of the tourist flow network. Specifically, tourist flows tend to concentrate around a limited number of high-grade attractions, while others—such as the 798 Art District—remain underdeveloped due to the lack of official ratings, thereby limiting their visibility and investment potential. This imbalance can negatively influence both the spatial distribution of tourist flows and the overall quality of the visitor experience.

5. Discussion

5.1. Theoretical and Practical Implications

As Beijing is the capital of China and a global metropolis, it is essential to understand the spatiotemporal patterns and influencing factors of its tourist flows to support its transformation into a world-class tourism city and a global tourist destination. Unlike previous studies, this study investigates the fine-grained spatiotemporal patterns and influencing factors of intra-city tourist flows in a megacity through the dual lenses of tourist perception and actual mobility. This dual-perspective approach provides a fresh analytical framework for examining urban tourism flows.

In the spatiotemporal patterns of tourism flows, Beijing’s tourism flow network exhibits a pronounced core–periphery structural characteristic (Figure 5), consistent with Zhao et al. [52], who examined the city’s tourism flow network from January 2018 to December 2019. Furthermore, Beijing’s tourism flow network lacks strong internal connectivity and suffers from significant regional imbalances. These findings align with the spatial differentiation reported by Wang et al. [42], who described Beijing’s urban–suburban ecotourism flows as “dense in the near suburbs, sparse in the distant suburbs; dense in the west, sparse in the east.” Regarding influencing factors, the study employed the number of hotels within a 5 km radius around each tourism node as an indicator of tourism reception services. However, due to limitations in data availability and time constraints, other potentially significant variables—such as subway coverage and restaurant density—were not included in the model. This omission may partly explain why tourism reception services did not reach statistical significance in the final model [53].

From a theoretical perspective, this study moves beyond traditional single-perspective approaches for analyzing the spatiotemporal evolution of tourist behavior [54,55]. Instead, it adopts a relationship-data-based perspective by integrating tourist perception and tourism flows to comprehensively uncover the evolving spatial configuration of tourism flows in Beijing. This approach enriches the theoretical understanding of the spatiotemporal evolution of tourism flows and offers a theoretical foundation for understanding the intrinsic connection between tourist behavior and the spatial pattern of tourism flows. Additionally, the study introduces the Baidu Index as a novel proxy for online attention to assess how online information-seeking behavior influences tourism decision-making in the digital era [25]. This contribution not only provides new insights into tourist decision-making in the age of big data but also expands the analytical lens for investigating the factors that shape tourism flow networks.

From a practical standpoint, the findings offer insights at both micro and macro levels. At the micro level, the analysis of tourist satisfaction and influencing factors of tourist flow provides practical implications for tourism managers to enhance product offerings, improve visitor experiences, and promote the sustainable development of destinations. For instance, optimizing online information dissemination channels can enhance both digital visibility and tourist satisfaction. With appropriate policy support, a three-tier tourism product system—comprising the “urban cultural tourism circle,” “suburban leisure belt,” and “external ecological zone”—can be established. Examples include cross-sector tourism products, such as the “Palace Museum Digital Exhibition + Universal Studios Night Tour,” which could enrich the visitor experience. At the macro level, the research offers an in-depth examination of the tourism flow network structure and tourism centers, providing strategic guidance for the evidence-based planning of tourism resources and the promotion of coordinated regional development. For example, optimizing the spatial structure through the functional reconfiguration of nodes can alleviate the reception pressure on core nodes while enhancing the visibility and attractiveness of secondary nodes. A coordinated “primary core–secondary core” mechanism should be established to facilitate tourist dispersion and experience enhancement. Furthermore, cultivating secondary nodes can foster a polycentric spatial structure conducive to more equitable and sustainable regional development.

5.2. Recommendations

As an international consumption center city, the development of tourism in Beijing has a profound impact on its cultural and economic landscape. Based on the findings on the spatiotemporal patterns and influencing factors of tourist flows, as well as the results of the SWOT analysis, the following policy recommendations are presented.

At the macro level, although the overall structure of Beijing’s tourist flow network remains stable, the concentration of tourism gravity continues to pose challenges (Table 6). It is recommended that the network evolve toward “internal flattening and external hierarchization”. Internally, reducing hierarchical structures and enhancing efficiency and coordination among attractions would promote information exchange between nodes and improve the overall network’s resilience. Externally, the scope of collaboration should be expanded to establish hierarchical linkages with surrounding areas and cities. Strategic cooperation with tourism cities of varying levels can enhance the robustness and resilience of the regional tourist flow network.

At the micro level, efforts should focus on enhancing the innovation capacity of individual attractions and strengthening collaborative efforts between scenic spots. The impact of public health crises has significantly weakened the interconnectivity of attractions within Beijing, underscoring the need to establish a collaborative development mechanism for tourist sites. On one hand, Beijing should leverage its economic and resource advantages to develop high-value-added, innovative tourism products for core attractions, thereby enhancing their international competitiveness. On the other hand, the strength of core nodes should be utilized to drive the development of other attractions, particularly those in peripheral locations.

5.3. Limitations and Future Work

This study systematically examined the evolutionary characteristics of Beijing’s tourism digital footprint, constructed the city’s tourism flow network structure, and conducted an in-depth analysis of the key factors influencing that structure. However, several limitations should be acknowledged and offer avenues for future research.

First, limitations of the data should be acknowledged. Due to limitations in data access (i.e., reliance on publicly available APIs from third-party online travel platforms), we were unable to directly obtain key demographic characteristics. Therefore, there may be concentration in user age, consumption habits, etc., which may introduce selection bias and limit the generalizability of findings to the broader tourist population. Moreover, the data span an extended time period during which policy and market conditions have fluctuated significantly, potentially affecting the temporal validity of some observations. Future studies could incorporate mobile phone signaling data to extract descriptive user attributes, such as age, and to enable real-time monitoring and trend analysis of tourist behavior [25]. Second, the current analysis focuses exclusively on digital footprints within Beijing and does not account for spatial interactions between Beijing and its surrounding regions. Given the growing emphasis on regional tourism integration, future research should expand the geographic scope—such as incorporating the Beijing–Tianjin–Hebei urban agglomeration—to capture intercity tourism dynamics and enhance the comprehensiveness and policy relevance of the findings. Third, limitations of indicator selection should be acknowledged. Although this study considers a wide range of influencing factors, due to time and resource constraints, some potentially important variables may still have been excluded—such as macroeconomic fluctuations and the development of infrastructure such as catering and transportation [53]. Future studies should explore these additional dimensions and develop a composite evaluation framework (such as incorporating multidimensional indicators such as catering, transportation hubs, and commercial facilities in POI data to more accurately quantify the reception capacity of tourist attractions) to enhance the explanatory power and robustness of tourism flow models.

6. Conclusions

This study utilized online travel diary data from Qunar.com to examine the evolution of Beijing’s tourism digital footprint across different stages, focusing on tourists’ perceptions and tourism flow dynamics. Building on this foundation, social network analysis and the QAP regression model were employed to investigate the spatiotemporal evolution of Beijing’s tourism flow network and to identify the key factors influencing its formation and development. Finally, based on the SWOT analysis, this study puts forward strategic recommendations for the development of tourism in Beijing. The main conclusions are as follows:

(1) Tourists’ perceptions and emotional tendencies toward Beijing tourism exhibit multidimensional and temporal evolutionary characteristics. Overall tourist satisfaction remains high, with a marked increase in the proportion of positive emotional expressions. Tourist interest has gradually shifted from traditional iconic attractions to more diverse and culturally rich experiences, reflecting an ongoing upgrading of tourism demand.

(2) From 2012 to 2024, the overall gravity center of Beijing’s tourism flow has remained relatively stable, with only minor shifts influenced by policy adjustments and the emergence of new attractions. Core historical and cultural districts, as well as prestigious universities, consistently retain strong appeal. Spatial linkages among tourist attractions have generally strengthened over time, and the centrality and influence of core nodes have continued to rise. The emergence of new attractions, such as Universal Studios, has significantly reshaped the network structure and established new tourist hotspots and secondary hubs.

(3) QAP correlation and regression analyses quantitatively assessed the driving factors of the tourism flow network. The results demonstrate that government policy orientation, online information search intensity, and tourism resource agglomeration exert significant positive effects on the formation and evolution of the tourism flow structure. In contrast, the hierarchical level of tourism resources does not significantly promote network development, suggesting that unclassified or non-traditional attractions may also exert considerable influence.

(4) Through a systematic SWOT analysis, this study identifies the key strengths, weaknesses, opportunities, and threats associated with the development of tourism in Beijing. The results suggest that the strengths considerably outweigh the weaknesses, and the opportunities exceed the threats. Specifically, the strengths include tourists’ predominantly positive emotional experiences and the overall stability of the tourist flow network structure. The weaknesses are exemplified by the spatial concentration of the gravity center of tourist flows in specific areas.