Route Generation and Built Environment Behavioral Mechanisms of Generation Z Tourists: A Case Study of Macau

Abstract

Personalized travel experiences have become a growing priority for tourists, while the built environment increasingly shapes tourists’ behavior. However, limited research has integrated behavioral drivers with algorithmic travel route optimization, particularly in the context of Generation Z tourists. To address this gap, this study proposes a hybrid framework that combines behavioral modeling with enhanced algorithmic techniques to generate customized travel itineraries for Generation Z. A behavioral influencing factors model is first constructed based on the Theory of Planned Behavior (TPB) and Social Influence Theory (SIT), identifying media influence (MI), subjective norms (SNs), and perceived built environment (PBE) as potential determinants of travel behavioral intention (BI). A Structural Equation Model (SEM) is then applied to empirically validate the hypothesized relationships. Results reveal that all three factors have a significant and positive impact on BI (p < 0.05). Building on this behavioral mechanism, an interest-based Ant Colony Optimization (ACO) algorithm is implemented by incorporating point-of-interest (POI) preferences and distance matrices to improve personalized route generation. Comparative analysis using social media keyword data demonstrates that the proposed method outperforms conventional travel route planning approaches in terms of route relevance and overall path satisfaction. This study offers a novel integration of psychological theory and computational optimization, providing both theoretical insights and practical implications for urban tourism planning and the development of smart tourism services.

1. Introduction

With the acceleration of urbanization and the rising demand for personalized travel, tourists’ expectations have shifted from functional experiences toward more contextualized and individualized journeys [1]. In response, urban tourism research is undergoing a paradigmatic shift—from traditional route design to behavior-oriented and spatially integrated modeling frameworks [2]. This transformation is particularly salient in high-density urban environments, where the quality of the built environment—such as accessibility, green space, and perceived safety—has become a significant determinant of route selection and visitor satisfaction [3]. Amid these changes, Generation Z tourists have emerged as a distinct and influential demographic in the tourism sector. Born into the digital era, this cohort is characterized by high digital literacy, strong social connectivity, and a marked preference for personalized, technology-enhanced experiences [4]. Investigating their travel behaviors and route decision mechanisms not only enhances the precision of personalized travel route recommendations but also introduces novel perspectives for urban spatial governance and tourism service optimization in dense metropolitan contexts [5]. Macau, as one of the world’s most densely populated cities, provides a unique context due to its compact urban structure blending historical, cultural, ecological, and entertainment spaces [6,7,8]. Driven by social media and digital navigation platforms, Generation Z tourists increasingly prefer exploring hidden alleys and artistic neighborhoods, expanding both spatial tourism patterns and their subjective perception of urban spaces [9,10]. However, the current literature exhibits several limitations, such as fragmented analyses of behavioral mechanisms [11,12], weak integration of behavioral insights with path optimization algorithms [13], and insufficient systematic incorporation of built environmental factors into behavioral and route decision models [14].

Although increasing academic attention has been directed toward Generation Z tourists—exploring their travel preferences from generational traits, motivations, and digital media usage [15,16,17,18]—and despite significant advancements using frameworks like the Theory of Planned Behavior (TPB) and Social Influence Theory (SIT) [19,20,21], translating psychological insights into practical route planning tools remains limited. Algorithms such as Ant Colony Optimization (ACO), while improving routing efficiency, typically neglect psychological or user-specific preference data, constraining personalization [22,23,24]. Furthermore, few approaches incorporate psychological or interest-based data, reducing personalization. Integrated models tailored to the distinct behaviors and preferences of Generation Z tourists—an increasingly influential group in smart tourism—remain notably scarce.

To address the aforementioned research gaps, this study selects Macau as the primary study area. As a world-renowned leisure tourism destination favored by young travelers, Macau provides a representative context for examining the behavioral patterns of Generation Z tourists. Focusing on their travel behavior characteristics, this study constructs a structural equation model (SEM) based on the Theory of Planned Behavior (TPB) and Social Influence Theory (SIT) to investigate how social media, peer influence, and perceptions of the built environment affect travel intentions. Building on the behavioral analysis, a personalized point-of-interest (POI) scoring system is developed to reflect individual preferences, which is then embedded into an Ant Colony Optimization (ACO) algorithm to generate customized travel routes that integrate both interest-based and spatial constraints. The specific research questions addressed in this study are as follows:

(1) Are the travel behavioral intentions of Generation Z tourists jointly influenced by social media, peer effects, and perceptions of the built environment?

(2) Does the structural model—developed by integrating the Theory of Planned Behavior (TPB) and Social Influence Theory (SIT)—exhibit strong explanatory and predictive power for the behavioral formation of Generation Z tourists?

(3) How can behavioral intention outcomes be effectively translated into quantifiable interest preference scores and further integrated with spatial distance information for use in travel route generation?

(4) Under multiple constraints—including interest preferences, built environment qualities, and spatial distances—is it feasible to apply an Ant Colony Optimization (ACO) algorithm to generate diverse travel paths that meet the personalized needs of Generation Z tourists?

The structure of this study is organized as follows: In Section 1, the relevant literature on Generation Z tourists’ behavioral characteristics, perceptions of the built environment, and route optimization approaches are reviewed, providing the theoretical foundation for model development. In Section 2, the hypothesized research model that integrates the Theory of Planned Behavior (TPB) and Social Influence Theory (SIT) is introduced, incorporating variables such as social media influence, peer influence, and built environment perception. In Section 3, the research methodology, covering data collection procedures and variable measurements, and a multi-dimensional scoring mechanism is presented. In Section 4, the path optimization model, which is developed using Ant Colony Optimization (ACO) based on behavioral intention scores and classified interest preferences, is presented. In Section 5, the generated travel routes are empirically validated through social media text analysis. Finally, in Section 6, the main findings are summarized, the relationships between behavioral mechanisms and route generations are discussed, practical recommendations are provided, and directions for future research are outlined.

2. Literature Review

2.1. Built Environment and Travel Behavior of Generation Z Tourists

The built environment typically refers to human-constructed physical spaces created through planning and design, encompassing elements such as street configurations, land use patterns, accessibility, green infrastructure, and public facilities [25]. Within the context of urban tourism research, the built environment is recognized as a critical external factor that shapes tourists’ perceptions, mobility patterns, and overall experiences [26]. Existing studies have commonly quantified the built environment using the following four core dimensions: convenience, greening ratio, tranquility, and safety [27]. These attributes are frequently operationalized through questionnaire surveys and factor-based scoring approaches, enabling the empirical assessment of the built environment’s influence on tourist behavior. Given that tourists’ perceptions and responses to the built environment may vary significantly across age groups, it becomes essential to focus on specific generational cohorts whose behavioral patterns are shaped by distinct socio-technological contexts.

There is currently no universally accepted definition of Generation Z, although most scholars define this cohort as individuals born between 1995 and 2012 [28,29,30]. Based on the context of their development within a mobile technology and social media environment, in this study, the birth range of 1997 to 2012 is adopted to define Generation Z [31,32]. As a generation raised in a highly digitalized environment, Generation Z exhibits greater autonomy and exploratory tendencies in their travel behaviors [33]. Compared to older age groups, they place a higher emphasis on immersive and sensory-rich tourism experiences, with particular attention to spatial comfort, walkability, and perceived safety [34,35]. In spatially dense tourism cities such as Macau, Generation Z tourists tend to prefer areas characterized by rich cultural ambiance and visually appealing environments. These behavioral patterns suggest a heightened sensitivity to the built environment among this cohort [36]. Field observations indicate that when faced with spatial constraints—such as steep slopes or mobility barriers—Generation Z tourists are more likely to adjust their travel routes promptly [37]. The behavioral response further emphasizes how built environment features significantly affect their perceived behavioral control and travel intention. However, existing studies often treat the built environment as a static background variable, overlooking how tourists—especially younger cohorts like Generation Z—dynamically respond to spatial constraints in real time. Moreover, the interaction between spatial perception and digital behaviors, such as real-time navigation or social sharing, remains underexplored. This study addresses these gaps by integrating built environment perception into a broader behavioral framework that reflects the digitally mediated travel practices of Generation Z.

2.2. Theory of Planned Behavior and Social Influence Pathways in Travel Behavior Modeling

The behavioral response further emphasizes how built environment features significantly affect their perceived behavioral control and travel intention [38]. According to TPB, behavioral intention is shaped by the following three key psychological constructs: attitude toward the behavior, subjective norms, and perceived behavioral control [39]. In the context of tourism, TPB has been extensively employed to explain various forms of behavioral intention, including revisit intentions, willingness to recommend, and engagement in emerging tourism practices such as ecotourism and immersive cultural experiences [40,41]. These applications highlight the theory’s flexibility in addressing both traditional and novel travel behaviors across diverse tourist segments. However, some studies have noted that TPB alone may not fully capture the influence of external social and technological contexts, particularly in digital native populations such as Generation Z. This limitation has prompted scholars to explore model extensions or integrations to better reflect contemporary behavioral drivers.

To further enhance the explanatory power of the TPB model in the context of Generation Z tourism behavior, this study incorporates Social Influence Theory (SIT) as an external variable framework. Originally proposed by Kelman, SIT posits that individual behavior is shaped through the following three key mechanisms: internalization, identification, and compliance [42]. In digitally mediated tourism contexts, social media content and peer recommendations have emerged as the primary channels through which social influence is exerted [43]. Generation Z not only consumes social content but also actively engages as a content producer [44,45,46,47]. Previous studies have demonstrated that peer behavior and recommendations significantly influence destination choices among Generation Z tourists. Moreover, such influences have been shown to reinforce both behavioral attitudes and subjective norms within the TPB framework [48,49]. Nonetheless, existing research tends to focus primarily on either media influence or peer behavior in isolation, lacking an integrated view of how these social factors jointly shape intention formation. By incorporating both dimensions under SIT, this study addresses that gap and provides a more holistic understanding of social mechanisms in digitally driven travel decisions.

2.3. Applications of Ant Colony Optimization in Built Environment Contexts

Ant Colony Optimization (ACO), benefiting from distributed computation capabilities and strong algorithmic stability, has been broadly applied to complex combinatorial optimization problems. It has demonstrated notable performance in solving classical problems such as the Traveling Salesman Problem (TSP) [50]. Its distributed search mechanism, adaptive learning capability, and multi-objective compatibility make it a powerful tool for simulating travel behavior within built environments [51]. In recent years, ACO has been increasingly employed in urban tourism route planning, pedestrian navigation, and traffic management, with a growing trend toward the incorporation of environmental perception variables [52]. Specifically, features of the built environment—such as walkability, green coverage, and perceived safety—have been embedded into the heuristic evaluation functions of ACO to enhance the behavioral relevance of route recommendations [53]. Compared to other commonly used heuristic algorithms such as Genetic Algorithms (GAs) and Particle Swarm Optimizations (PSOs), ACO offers distinct advantages in the context of route planning. Unlike GA, which relies on global crossover and mutation strategies that may disrupt local optima, ACO utilizes a pheromone-based learning mechanism that allows for more stable convergence and localized optimization—particularly valuable in spatially constrained, user-specific tourism routing scenarios. Similarly, while PSO excels in continuous optimization problems, it often struggles with the discrete, path-dependent nature of travel route design, where ACO’s node-based path construction is inherently more applicable.

These enhancements have enabled the generation of personalized travel routes more closely aligned with individual preferences. Further studies have explored the integration of ACO with behavioral theories, such as the Theory of Planned Behavior, and hybrid computational approaches to simulate psychological variables including travel intention, environmental perception, and personal preference [54,55]. Moreover, ACO’s modular framework allows seamless integration of external data inputs such as survey-derived preference scores or environmental evaluations, enabling a more comprehensive representation of individual tourist behavior. As a flexible computational framework, ACO effectively supports the integration of environmental perception and behavioral decision-making in the context of personalized route optimization. This not only expands its applicability in complex spatial problem-solving but also introduces new methodological pathways for analyzing the decision-making behaviors of specific tourist groups.

3. Materials and Methods

To clearly illustrate the overall methodological framework of this study, a step-by-step research design process is presented in Figure 1. This framework integrates theoretical model development, hypothesis formulation, structural equation model testing, route generation using the Ant Colony Optimization (ACO) algorithm, and validation through social media data. Each step reflects the logical progression from conceptual foundations to empirical analyses and strategic implications. This process ensures both theoretical rigor and practical applicability in examining the travel behavior of Generation Z tourists within high-density urban environments such as Macau.

3.1. Research Hypotheses

To investigate the intrinsic relationships among built environment perception, behavioral theory, and route optimization, this study proposes a structural equation model (SEM) that integrates the Theory of Planned Behavior (TPB), Social Influence Theory (SIT), and perceptions of the built environment. This model is designed to explain the travel interests and route preferences of Generation Z tourists (see Figure 2). Key constructs from the TPB—attitude toward behavior (ATT), subjective norms (SNs), perceived behavioral control (PBC), behavioral intention (BI), and actual travel behavior (AB)—are incorporated into a unified framework. In addition, media influence (MI), peer influence (PI), and perceived built environment (PBE), derived from the SIT and environmental context, are included to enhance the model’s explanatory power. The variables and their corresponding measurement dimensions are summarized in Table 1.

Based on the existing literature and the proposed research framework, the following hypotheses are formulated:

H1. 

Attitude toward behavior (ATT) has a significant positive effect on behavioral intention (BI).

H2. 

Subjective norms (SNs) have a significant positive effect on behavioral intention (BI).

H3. 

Perceived behavioral control (PBC) has a significant positive effect on behavioral intention (BI).

H4. 

Behavioral intention (BI) has a significant positive effect on actual travel behavior (AB).

H5. 

Media influence (MI) has a significant positive effect on attitude toward behavior (ATT).

H6. 

Media influence (MI) has a significant positive effect on subjective norms (SNs).

H7. 

Peer influence (PI) has a significant positive effect on subjective norms (SNs).

H8. 

Peer influence (PI) has a significant positive effect on behavioral intention (BI).

H9. 

Perceived built environment (PBE) has a significant positive effect on perceived behavioral control (PBC).

Figure 1. Research framework and workflow.

Figure 1. Research framework and workflow.

Figure 2. Hypothesized structural mode.

Figure 2. Hypothesized structural mode.

Table 1. Variables and corresponding measurement dimensions.

Variable	Code	Measurement Item	Reference
Attitude Toward Behavior	A1	I think Macau is a city worth visiting for tourism.	[11,12,19,20,38,39,40,41]
	A2	Traveling to Macau is a pleasant and enjoyable experience.
	A3	Macau’s tourism culture attracts me to explore the city.
Subjective Norms	B4	My family or friends support my decision to travel to Macau.	[11,12,19,20,38,39,40,41,49,56]
	B5	Many people in my social circle have traveled to Macau.
	B6	If I travel to Macau, I believe they would approve of my decision.
Perceived Behavioral Control	C7	I have enough time to plan a trip to Macau.	[11,12,19,20,38,39,40,41]
	C8	If conditions allow, I am willing to choose Macau as my travel destination.
	C9	I am highly likely to recommend traveling to Macau to others.
Tourism Behavior	E13	My most recent visit to Macau was for tourism purposes.	[19,20,40,41,48,57]
	E14	I have traveled to Macau within the past year.
	E15	During my trip to Macau, I participated in many cultural or leisure activities.
Media Influence	F16	I often obtain information about Macau tourism through social media (e.g., Xiaohongshu, TikTok).	[16,20,43,45,46,47,48,57]
	F17	Promotional videos about Macau on online platforms have increased me interest in visiting.
	F18	Recommendations from travel bloggers or influencers on social media affect my travel attitude toward Macau.
Peer Influence	G19	Travel destinations recommended by my friends or peers influence my choices.	[21,40,41,49,56]
	G20	I am more likely to travel to Macau if my friends plan to go.
	G21	Traveling with friends in Maca enhances my overall travel experience.
Perceived Built Environment	H22	I find walking around Macau for tourism to be very convenient.	[9,10,25,26,27,36,58]
	H23	I think tourist areas in Macau have rich greenery, which makes me feel relaxed.
	H24	I can find quiet places while visiting tourist attractions in Macau.
	H25	I feel very safe when visiting tourist attractions in Macau.

Table 1. Variables and corresponding measurement dimensions.

Variable	Code	Measurement Item	Reference
Attitude Toward Behavior	A1	I think Macau is a city worth visiting for tourism.	[11,12,19,20,38,39,40,41]
	A2	Traveling to Macau is a pleasant and enjoyable experience.
	A3	Macau’s tourism culture attracts me to explore the city.
Subjective Norms	B4	My family or friends support my decision to travel to Macau.	[11,12,19,20,38,39,40,41,49,56]
	B5	Many people in my social circle have traveled to Macau.
	B6	If I travel to Macau, I believe they would approve of my decision.
Perceived Behavioral Control	C7	I have enough time to plan a trip to Macau.	[11,12,19,20,38,39,40,41]
	C8	If conditions allow, I am willing to choose Macau as my travel destination.
	C9	I am highly likely to recommend traveling to Macau to others.
Tourism Behavior	E13	My most recent visit to Macau was for tourism purposes.	[19,20,40,41,48,57]
	E14	I have traveled to Macau within the past year.
	E15	During my trip to Macau, I participated in many cultural or leisure activities.
Media Influence	F16	I often obtain information about Macau tourism through social media (e.g., Xiaohongshu, TikTok).	[16,20,43,45,46,47,48,57]
	F17	Promotional videos about Macau on online platforms have increased me interest in visiting.
	F18	Recommendations from travel bloggers or influencers on social media affect my travel attitude toward Macau.
Peer Influence	G19	Travel destinations recommended by my friends or peers influence my choices.	[21,40,41,49,56]
	G20	I am more likely to travel to Macau if my friends plan to go.
	G21	Traveling with friends in Maca enhances my overall travel experience.
Perceived Built Environment	H22	I find walking around Macau for tourism to be very convenient.	[9,10,25,26,27,36,58]
	H23	I think tourist areas in Macau have rich greenery, which makes me feel relaxed.
	H24	I can find quiet places while visiting tourist attractions in Macau.
	H25	I feel very safe when visiting tourist attractions in Macau.

3.2. Study Area

The Macau Special Administrative Region was selected as the study area due to its distinctive diversity in tourism resources, spatial configuration, and tourist behavior patterns. As one of the world’s most densely inhabited cities, Macau not only offers a wealth of historical and cultural tourism assets but also features a highly compact urban structure and a well-developed pedestrian network. This enables a highly accessible system of points-of-interest (POIs) for visitors [59]. In recent years, Macau’s historic districts, urban spaces, and art exhibitions have increasingly become popular destinations for Generation Z tourists. Amplified through social media platforms, these areas have gained widespread exposure, showcasing Macau’s notable advantages in visual perception and digital dissemination [60]. Locations such as Love Lane and Patio de Chon Sau have attracted significant attention from Generation Z visitors, owing to their unique artistic ambiance and cultural significance. As a short-distance, high-frequency, and experience-intensive urban tourism destination, Macau’s manageable scale and controllable spatial layout support the self-guided, in-depth, and exploratory travel behaviors commonly preferred by Generation Z tourists. Its compact and diversified spatial pattern provides an ideal testing ground for examining how the built environment influences tourist behavior, route choice, and overall travel experience. Based on these characteristics, Macau not only offers a practical foundation for theoretical model construction but also delivers ample data support for empirical validation and algorithmic optimization.

3.3. Target Group

This study focuses on Generation Z tourists, defined as individuals born between 1997 and 2012. This cohort has grown up with digital technology and is characterized by advanced media literacy, a strong sense of autonomy in decision-making, and a clear preference for personalized travel experiences [57]. Generation Z travelers tend to rely heavily on social media during the decision-making process and exhibit heightened sensitivity toward spatial exploration and environmental perception [58]. The Statistics and Census Service of the Macau Special Administrative Region reported that the number of inbound tourists reached 31.35 million from November 2023 to October 2024, with young travelers increasingly constituting a major segment of Macau’s tourism market [61]. Generation Z tourists typically reconstruct their travel logic through short-distance, in-depth experiences, urban roaming, and check-in-style visits. Their behavioral decisions are highly influenced by social dynamics and perception-based factors. In particular, they are significantly affected by social media recommendations and peer opinions, and they display pronounced sensitivity to spatial features of the built environment, such as walkability, green coverage, and perceived safety [62]. Focusing on Generation Z as the target group not only allows for an in-depth examination of the path selection mechanisms of contemporary young tourists under multi-dimensional influences but also provides valuable empirical support for the development of a model that integrates behavioral theory, environmental perception, and algorithmic optimization.

3.4. Data Collection and Processing

3.4.1. Descriptive Statistical Analysis

In this study, 200 questionnaires were distributed, yielding 158 valid responses that were subsequently analyzed. The demographic profile of the sample is shown in

Table 2. Descriptive statistical analysis of sample.

Variable	Category	Frequency	Percentage (%)
Gender	Female	113	71.52
	Male	45	28.48
Age Group	18–25 years	62	39.24
	26–35 years	96	60.76
Place of Residence	Mainland China	157	99.37
	Other parts of Asia	1	0.63
Number of Visits to Macau	First-time visitor	71	44.94
	Two-to-three visits	77	48.73
	More than four visits	10	6.33
Occupational Sector	Corporate Employee	103	65.19
	Student	42	26.58
	Freelancer	1	0.63
	University Faculty or Staff	5	3.16
	Other	7	4.43
	Short Trip (1–2 days)	41	25.95
Duration of Stay in Macau	Corporate Employee	103	65.19
	Student	42	26.58
Historical and Cultural Sites		104	65.82
Religious Architecture		35	22.15
Modern Entertainment Venues		116	73.42
Natural Landscapes		105	66.46
Culinary Experiences		125	79.11
Art Exhibitions		71	44.94
Total		158	100.0

. The authors are aware that the total population of Generation Z tourists visiting Macau fluctuates seasonally and is not fixed; however, based on recent tourism statistics published by the Macau Government Tourism Office, a conservative estimate suggests that millions of Gen Z tourists visit the region annually. Given this, the sample size of 158 valid responses meets the minimum requirement for SEM analysis and reflects a sufficiently diverse subset. Moreover, the sample structure—covering different travel frequencies, interest types, and length of stay—was carefully balanced to ensure representativeness. Therefore, the responses can be considered reasonably representative of the broader population of Generation Z tourists visiting Macau. The structure of the Questionnaire is shown in Figure 3. The sample includes a wide range of attributes, such as gender, age, place of residence, frequency of visits to Macau, occupation, duration of stay, and travel interests—offering a comprehensive representation of Generation Z tourists. In terms of travel preferences, the sample displays marked diversity, consistent with the behavioral patterns typically observed in the Generation Z tourism market. Respondents exhibit a strong tendency toward social engagement and show a preference for short-duration, experience-oriented travel activities. Such behaviors are often reflected in urban roaming and “check-in” tourism practices and are driven by a combination of motivations, including culinary exploration, entertainment, and cultural engagement. These results provide a solid empirical foundation for the study. They not only facilitate a deeper understanding of the behavioral characteristics and travel motivations of Generation Z tourists in Macau but also establish a robust data basis for subsequent structural equation modeling (SEM) and path analysis.

3.4.2. Reliability and Validity Testing

To ensure the quality of the measurement instruments, reliability and validity tests were conducted for each construct. First, Cronbach’s alpha coefficients were calculated to assess the internal consistency of the measurement dimensions. As shown in

Table 3. Reliability assessment.

Dimensions	Alpha	Number of Scale Items
Media Influence	0.841	3
Peer Influence	0.825	3
Built Environment Influence	0.838	3
Attitude Toward Behavior	0.808	3
Subjective Norms	0.817	3
Perceived Behavioral Control	0.802	3
Behavioral Intention	0.809	3
Tourism Behavior	0.862	4

, all constructs reported Cronbach’s alpha values greater than 0.8, indicating high internal consistency. These results meet the standard threshold for reliability assessment, confirming the scale’s strong internal reliability.

Furthermore, to assess the structural validity of the measurement scale the Kaiser–Meyer–Olkin (KMO) test and Bartlett’s test of sphericity were performed for each latent construct (see

Table 4. Validity assessment.

Dimensions	KMO	Bartlett’s Test of Sphericity	Degrees of Freedom	Statistical Significance
Media Influence	0.723	242.367	3	0.000
Peer Influence	0.716	219.759	3	0.000
Built Environment Influence	0.727	233.609	3	0.000
Attitude Toward Behavior	0.713	194.527	3	0.000
Subjective Norms	0.717	205.684	3	0.000
Perceived Behavioral Control	0.708	188.610	3	0.000
Behavioral Intention	0.707	198.521	3	0.000
Tourism Behavior	0.827	353.165	6	0.000

). The KMO values for all dimensions exceeded 0.7, indicating that the sample data were adequate for factor extraction and thus suitable for structural analysis. In addition, Bartlett’s test results were all highly significant (p < 0.001), confirming that sufficient correlations exist among the observed variables to justify factor analysis. Taken together, these results demonstrate that the designed measurement constructs possess strong factorial validity, thereby providing a sound empirical foundation for the subsequent structural equation model fitting.

The outcomes of the confirmatory factor analysis (CFA) are displayed in

Table 5. Standardized path coefficients from confirmatory factor analysis (CFA).

Dimensions	Indicators	Estimate	S.E.	C.R.
Media Influence	MI3	1.000
Media Influence	MI2	0.994	0.088	11.234
Media Influence	MI1	0.999	0.088	11.329
Attitude Toward Behavior	ATT1	1.000
Attitude Toward Behavior	ATT2	0.971	0.088	10.972
Attitude Toward Behavior	ATT3	0.974	0.088	11.059
Subjective Norms	SN1	1.000
Subjective Norms	SN2	0.960	0.100	9.600
Subjective Norms	SN3	0.982	0.099	9.898
Perceived Behavioral Control	PBC1	1.000
Perceived Behavioral Control	PBC2	1.002	0.098	10.181
Perceived Behavioral Control	PBC3	0.995	0.099	10.053
Behavioral Intention	BI1	1.000
Behavioral Intention	BI2	1.109	0.117	9.504
Behavioral Intention	BI3	1.063	0.115	9.244
Tourism Behavior	AB1	1.000
Tourism Behavior	AB2	1.068	0.109	9.807
Tourism Behavior	AB3	1.054	0.111	9.524
Peer Influence	PI1	1.000
Peer Influence	PI2	0.918	0.085	10.773
Peer Influence	PI3	0.902	0.088	10.274
Built Environment Influence	XX1	1.000
Built Environment Influence	XX2	0.975	0.087	11.154
Built Environment Influence	XX3	0.956	0.089	10.741
Built Environment Influence	XX4	1.028	0.091	11.276

. All indicators demonstrated statistically significant factor loadings on their respective latent constructs, with critical ratio (C.R.) values well above conventional thresholds for significance. These findings indicate that the observed variables effectively represent their intended constructs, providing evidence of strong convergent validity within the model. The distribution of factor loadings falls within an acceptable range, with no abnormally high or low values observed. This suggests a consistent and balanced structure across the measurement items. In addition, standard error values were generally low, further supporting the robustness and reliability of the model estimates. It is worth noting that certain constructs, such as perceived built environment, were measured using more than three observed variables. The path estimates for these extended constructs remained within a consistent range when compared with other dimensions, indicating the statistical soundness of the extended measurement design.

3.4.3. POI Clustering Analysis

To further clarify the classification characteristics of tourist attractions in Macau, this study selected 30 representative sites published by the Macau Government Tourism Office [56] as the research sample. These attractions were chosen due to their high visibility, frequent appearances on major tourism platforms, and substantial visitor traffic, which collectively reflect the core features of Macau’s tourism landscape. In addition, user reviews and preference tags from online platforms were analyzed to supplement and validate the functional attributes of each site, ensuring alignment with the actual interests of Generation Z tourists. Based on the functional attributes of the attractions and visitor preferences, a cluster analysis was conducted, categorizing the 30 attractions into 6 groups: historical and cultural, modern entertainment, natural scenery, culinary experience, art exhibitions, and religious architecture. The classification results for each category are presented in

Table 6. Classification of tourism attractions in Macau.

Category	Tourist Attractions
Historical and Cultural Sites	Lilau Square, Mandarin’s House, Rua da Felicidade and adjacent streets, Ruins of St. Paul’s, Guia Fortress (including Guia Lighthouse and Chapel), A-Ma Temple, St. Lawrence’s Church, St. Joseph’s Seminary and Church, Our Lady of Carmel Church, Coloane Historical and Cultural Village, Section of the Old City Wall, Tam Kung Temple, Na Tcha Temple, St. Francis Xavier Church, Coloane Kun Iam Temple, and Coloane Tin Hau Temple.
Modern Entertainment Venues	The Venetian Macau Resort, Macau Cable Car, Macau Tower Convention and Entertainment Centre, and Rua do Cunha.
Natural Landscapes	Taipa Grande Viewing Platform, Coloane Pier, and Hac Sa Beach.
Culinary Experiences	Rua do Cunha.
Art Exhibitions	Tap Seac Gallery, Iao Hon Market Pavilion and Art Space, Creative Park No. 10, and Macau Science Center.
Religious Architecture	A-Ma Temple, St. Lawrence’s Church, St. Joseph’s Seminary and Church, Our Lady of Carmel Church, Na Tcha Temple, St. Francis Xavier Church, Kun Iam Temple, Tin Hau Temple, and Tam Kung Temple.

.

To determine the optimal number of clusters, the elbow method was applied to perform K-means clustering on the attraction selection data of Generation Z tourists. The within-cluster sum of squares (WCSS) was calculated for different values of k, and an elbow plot was generated (see Figure 4). The results indicate that the rate of WCSS reduction begins to plateau at k = 3 or 4, forming a clear “elbow” shape. Considering both the behavioral complexity of Generation Z tourists and the interpretability of the clustering results, four clusters were ultimately selected. Based on the clustering output and the previously discussed characteristics of Generation Z tourists, the tourist population was classified into the following four groups: gastronomic and nature seekers, cultural integration explorers, artistic and immersive humanities travelers, and heritage-oriented traditionalists. Attraction types were matched to each cluster based on their average scores (mean values), with detailed classification results presented in

Table 7. Classification of tourist attraction types in Macau.

	Historical and Cultural Sites	Modern Entertainment Venues	Natural Landscapes	Culinary Experiences	Art Exhibitions	Religious Architecture
Cluster	mean	mean	mean	mean	mean	mean	silhouette
Gastronomic and Nature-Oriented Type	0	0.39	0.46	0.54	0.25	0.18	0.19
Cultural Integration Type	1	0.73	0.14	0.48	0.11	0.05	0.33
Artistic and Humanistic Type	1	0.76	0.73	0.68	0.95	0.89	0.29
Historical and Cultural Type	1	0	1	0.62	0.58	0.19	0.34

and

Table 8. Generation Z tourist clusters and corresponding types of tourism attractions.

	Type of Tourist Attraction
Gastronomic and Nature-Oriented Type	Culinary Experiences and Natural Landscapes.
Cultural Integration Type	Historical and Cultural Sites and Modern Entertainment.
Artistic and Humanistic Type	Historical and Cultural Sites, Art Exhibitions, and Religious Architecture.
Historical and Cultural Type	Natural Landscapes and Historical and Cultural Sites.

.

3.4.4. Ant Colony Optimization Algorithm

To generate travel routes tailored to different types of Generation Z tourists, an Ant Colony Optimization (ACO) approach was utilized to build a path selection model, building upon the prior clustering analysis. The model aims to balance multiple objectives—such as social media popularity and spatial distance—when identifying optimal travel paths. Such routes are considered the most appealing to Generation Z tourists. In the era of location-based services, social media platforms increasingly support the embedding of spatial data into online social networks via location-capturing technologies. These technologies serve as rich sources of big data, enabling researchers to collect large-scale, user-generated movement trajectories on a daily basis. These behavioral patterns are especially relevant in spatially dense environments like Macau and should be integrated into route planning strategies. In this study, the objectives were represented in the form of matrices, with one encoding the pairwise distances between tourist attractions and the other capturing built environment evaluation scores based on factors such as walkability, green coverage, and safety. These matrices were input into the ACO algorithm.

The ACO model consists of the following two core computational steps:

(1) Constructing solution paths to address the routing problem.

(2) Updating pheromone trails along the paths.

The ACO is as follows:

$${\mathrm{P}}_{\mathrm{i}\mathrm{j}}(\mathrm{t})={\displaystyle \frac{{\left[{\mathsf{\tau }}_{\mathrm{i}\mathrm{j}}\left(\mathrm{t}\right)\right]}^{\mathsf{\alpha }}·{\left[{\mathsf{\eta }}_{\mathrm{i}\mathrm{j}}\right]}^{\mathsf{\beta }}}{\sum _{\mathrm{k}\in {\mathcal{N}}_{\mathrm{i}}} {\left[{\mathsf{\tau }}_{\mathrm{i}\mathrm{k}}\left(\mathrm{t}\right)\right]}^{\mathsf{\alpha }}·{\left[{\mathsf{\eta }}_{\mathrm{i}\mathrm{k}}\right]}^{\mathsf{\beta }}}}$$

(1)

where ${\mathrm{P}}_{\mathrm{i}\mathrm{j}}\left(\mathrm{t}\right)$ is the probability that an ant located at city i will move to city j at iteration, ${ \mathsf{\tau }}_{\mathrm{i}\mathrm{j}}\left(\mathrm{t}\right)$ is the pheromone concentration on path (i, j), and ${\mathsf{\eta }}_{\mathrm{i}\mathrm{j}}$ is the inverse of the path length. For the pheromone update rule, after each iteration pheromone levels are updated based on the performance of the ants. The update is conducted using the following formula:

$${\mathsf{\tau }}_{\mathrm{i}\mathrm{j}}\left(\mathrm{t}+1\right)=\left(1-\mathsf{\rho }\right)·{\mathsf{\tau }}_{\mathrm{i}\mathrm{j}}\left(\mathrm{t}\right)+\mathsf{\Delta }{\mathsf{\tau }}_{\mathrm{i}\mathrm{j}}$$

(2)

After determining the selection and configuration of key parameters in the Ant Colony Optimization (ACO) algorithm, the model was further refined to reflect the contextual background in which Generation Z tourists are influenced by both spatial distance and characteristics of the built environment. Under these circumstances, the relative emphasis on pheromone information α and heuristic information β was meticulously fine-tuned. The core framework of the ACO algorithm, incorporating these considerations, is illustrated in Figure 5.

4. Research Results

4.1. Hypothesis Testing

In this study, the proposed structural equation model’s goodness-of-fit was assessed using a comprehensive set of indices, encompassing absolute fit, incremental fit, and parsimony-adjusted measures.

Table 9. Model fit assessment for structural equation model.

Indicator	Reference Standard	Measured Value
CMIN/DF	A value between 1 and 3 is considered excellent, while a value between 3 and 5 is regarded as good.	1.637
RMSEA	A value of <0.05 indicates excellent model fit, and values <0.08 are considered good.	0.057
NFI	A value of >0.90 reflects excellent reliability, and >0.80 indicates good reliability.	0.832
TLI	A value of 0.90 reflects excellent validity, and >0.8 indicates good validity.	0.916
GFI	A value of >0.90 reflects excellent structural consistency, and >0.80 indicates good consistency.	0.855
CFI	A value of >0.90 reflects excellent internal coherence, and >0.80 indicates good coherence.	0.926

shows that all indices meet or exceed the suggested thresholds, reflecting a strong fit between the model structure and the observed data. In terms of absolute fit, the model exhibits low residuals and structural error, aligning well with the theoretical model assumptions. Incremental fit indices further demonstrate a substantial improvement over the baseline model, particularly in terms of structural validity and explanatory power. Although some fit indices fall slightly below the threshold for “excellent” fit, they remain within the “acceptable”-to-“good” range, suggesting the model is stable overall and free of a significant misfit. The results confirm a high degree of consistency between the measurement model and the hypothesized structural pathway, thus confirming the validity of the subsequent path coefficient analysis and the strength of the model’s explanatory power.

The results of the structural equation model (SEM) analysis, as illustrated in Table 10 and Figure 6, reveal the underlying behavioral mechanisms of Generation Z tourists in the context of Macau’s urban tourism. The model identifies key factors influencing both travel intention and actual behavior by examining the path relationships between latent constructs and their corresponding observed indicators. In the diagram, ellipses represent latent variables—including media influence (MI), peer influence (PI), attitude toward behavior (ATT), subjective norms (SNs), perceived behavioral control (PBC), behavioral intention (BI), actual behavior (AB), and perceived built environment (XX)—while rectangles denote the measurement items for each construct. Standardized path coefficients (β values) reflect the strength of influence between variables. All hypothesized paths were statistically significant, confirming the model’s theoretical validity and structural soundness.

The path analysis indicates that media influence indirectly affects behavioral intention through its impact on attitudes and subjective norms. Peer influence not only strengthens subjective norms but also directly affects behavioral intention. Perceived built environment significantly enhances perceived behavioral control, which in turn positively influences behavioral intention. Behavioral intention emerges as a central mediating variable, demonstrating strong predictive power for actual travel behavior.

Overall, the model establishes a closed-loop behavioral pathway that begins with social media exposure and environmental perception, progresses through cognitive evaluation, and ultimately leads to travel action. This framework offers a systematic, data-driven approach for understanding the decision-making processes of Generation Z tourists in high-density urban environments.

Figure 6. Structural equation model (SEM).

Figure 6. Structural equation model (SEM).

Table 10. Path analysis results of the structural equation model.

Path	Estimate	S.E.	C.R.	p
H1: Behavioral Intention <--- Attitude Toward Behavior	0.293	0.064	3.560
H2: Behavioral Intention <--- Subjective Norms	0.321	0.076	3.601
H3: Behavioral Intention <--- Perceived Behavioral Control	0.233	0.066	2.94	0.003
H4: Tourism Behavior <--- Behavioral Intention	0.524	0.105	5.353
H5: Attitude Toward the Behavior <--- Media Influence	0.509	0.095	5.908
H6: Subjective Norms <--- Media Influence	0.374	0.096	3.927
H7: Subjective Norms <--- Peer Influence	0.195	0.082	2.109	0.035
H8: Behavioral Intention <--- Peer Influence	0.173	0.063	2.054	0.040
H9: Perceived Behavioral Control<--- Built Environment Influence	0.441	0.081	5.129

Table 10. Path analysis results of the structural equation model.

Path	Estimate	S.E.	C.R.	p
H1: Behavioral Intention <--- Attitude Toward Behavior	0.293	0.064	3.560
H2: Behavioral Intention <--- Subjective Norms	0.321	0.076	3.601
H3: Behavioral Intention <--- Perceived Behavioral Control	0.233	0.066	2.94	0.003
H4: Tourism Behavior <--- Behavioral Intention	0.524	0.105	5.353
H5: Attitude Toward the Behavior <--- Media Influence	0.509	0.095	5.908
H6: Subjective Norms <--- Media Influence	0.374	0.096	3.927
H7: Subjective Norms <--- Peer Influence	0.195	0.082	2.109	0.035
H8: Behavioral Intention <--- Peer Influence	0.173	0.063	2.054	0.040
H9: Perceived Behavioral Control<--- Built Environment Influence	0.441	0.081	5.129

4.2. Route Generation

Based on the research hypotheses, K-means clustering results, and Ant Colony Optimization (ACO) algorithm, four distinct types of travel routes were generated (see

Table 11. Categorized travel routes for Generation Z tourists.

Type	Tourism Route
Gastronomic and Nature-Oriented Type	Coloane Pier, Taipa Grande Viewing Platform, Rua do Cunha, Hac Sa Beach, and Coloane Pier.
Cultural Integration Type	Mandarin’s House, St. Lawrence’s Church, St. Joseph’s Seminary and Church, Rua da Felicidade and adjacent streets, Na Tcha Temple, Section of the Old City Wall, Ruins of St. Paul’s, Macau Cable Car, Guia Fortress (including the lighthouse and chapel), St. Francis Xavier Church (Coloane), Coloane Tin Hau Temple, Coloane Kun Iam Temple, Tam Kung Temple, The Venetian Macau Resort, Rua do Cunha, Coloane Historical and Cultural Village, Our Lady of Carmel Church, Macau Tower, A-Ma Temple, Lilau Square, and Mandarin’s House.
Artistic and Humanistic Type	Coloane Tin Hau Temple, Coloane Kun Iam Temple, Tam Kung Temple, St. Francis Xavier Church (Coloane), Guia Fortress (including the lighthouse and chapel), Macau Science Center, Tap Seac Gallery, Ruins of St. Paul’s, Creative Park No. 10, Iao Hon Market Pavilion and Art Space (Mercy Market House), St. Joseph’s Seminary and Church, St. Lawrence’s Church, Mandarin’s House, Lilau Square, A-Ma Temple, Rua da Felicidade and adjacent streets, Na Tcha Temple, Section of the Old City Wall, Our Lady of Carmel Church, Coloane Historical and Cultural Village, and Coloane Tin Hau Temple.
Historical and Cultural Type	Section of the Old City Wall, Ruins of St. Paul’s, Macau Cable Car, Guia Fortress (including the lighthouse and chapel), St. Francis Xavier Church (Coloane), Coloane Tin Hau Temple, Coloane Kun Iam Temple, Tam Kung Temple, The Venetian Macau Resort, Rua do Cunha, Coloane Historical and Cultural Village, Our Lady of Carmel Church, Macau Tower, A-Ma Temple, Mandarin’s House, St. Lawrence’s Church, Lilau Square, St. Joseph’s Seminary and Church, Rua da Felicidade and adjacent streets, Na Tcha Temple, and Section of the Old City Wall.

). Gastronomic and nature-oriented route: Centered on leisure experiences and natural landscapes, this route features a concise, loop-based path design emphasizing environmental comfort and ease of movement. It is particularly suitable for tourists who prefer light, relaxing, and low-intensity travel. Cultural integration route: This route offers broad spatial coverage, connecting key historical landmarks and religious sites across Macau while incorporating modern entertainment venues and iconic architectural structures. It reflects a blended experience of “tradition and modernity”, catering to tourists seeking cultural variety and architectural diversity. Artistic and immersive humanities route: Focused on museums, art galleries, religious architecture, and culturally rich neighborhoods, this route prioritizes esthetic value and cultural immersion. It is ideal for visitors who place greater emphasis on spiritual enrichment and deep urban exploration. Heritage-oriented historical route: Designed to trace the historical evolution of Macau, this route links major heritage nodes across the city. It appeals to tourists with a strong interest in historical continuity and deep cultural meaning.

4.3. Analysis of Trending Keywords on Social Media

To validate the rationality of the categorized travel routes, keyword data were collected from social media platforms frequently used by Generation Z tourists—specifically Xiaohongshu and Weibo—using Python 3.10.13-based web scraping techniques. Prior to implementing the topic modeling process, a structured keyword preprocessing procedure was carried out to enhance the accuracy and relevance of the dataset. This involved the removal of stop words, emojis, punctuation marks, and duplicate entries, alongside standard text normalization steps such as lowercasing and lemmatization. To minimize semantic noise, a minimum frequency threshold was applied, whereby only keywords appearing in at least five distinct posts were retained for analysis. In addition, generic terms with low semantic specificity—such as “Macau”, “travel”, and “play”—were manually excluded to improve the interpretability of the resulting topics. Latent Dirichlet Allocation (LDA) topic modeling was then employed to perform semantic analysis of the extracted keywords. The analysis yielded both thematic distributions and semantic features of the keywords. The left side of Figure 7 displays the main topics extracted by the LDA model, while the right side illustrates the high-frequency keywords associated with each topic. In the frequency distribution visualization, blue bars represent the frequency of keywords across the entire corpus, whereas red bars indicate the frequency of keywords within each specific topic. The results demonstrate a strong semantic alignment between the identified topics and the corresponding types of attractions, confirming the conceptual validity of the travel route classifications.

(1) Gastronomic and Nature-Oriented Theme

This topic was characterized by frequent occurrences of keywords such as “food”, “travel”, and “check-in”, which closely align with popular gastronomic and leisure destinations such as Rua do Cunha and Hac Sá Beach. These results reflect Generation Z tourists’ strong preferences for culinary experiences and natural landscapes.

(2) Cultural Integration Theme

Keywords such as “World Heritage”, “culture”, and “city” were dominant in this topic, corresponding well with multi-cultural landmarks like the Ruins of St. Paul’s and A-Ma Temple. This highlights tourists’ interest in cultural experiences and urban exploration.

(3) Artistic and Immersive Humanities Theme

Terms like “exhibition”, “interactive”, and “immersive” were frequently observed in this topic, which strongly correspond with artistic spaces such as Tap Seac Gallery and the Ox Warehouse. These findings indicate Generation Z’s heightened interest in artistic engagement and deep cultural immersion.

(4) Heritage-Oriented Historical Theme

Keywords such as “ruins”, “temple”, and “tradition” prominently appeared in this topic and closely matched heritage sites like the Mandarin’s House and the Old City Wall. This confirms a pronounced preference for historical and culturally authentic environments among heritage-focused tourists.

Through the above analysis, a high degree of semantic consistency was observed between the four identified topics and the corresponding travel route categories. This finding confirms the conceptual validity and scientific soundness of the route classification framework. It further demonstrates that a classification approach based on Generation Z tourists’ behavioral characteristics and semantic patterns can effectively capture the needs and preferences of different tourist segments. This provides a solid foundation for subsequent research in route optimization and built environment analysis.

Figure 7. Semantic clustering and keyword distributions across four experience-oriented travel types [63,64].

Figure 7. Semantic clustering and keyword distributions across four experience-oriented travel types [63,64].

5. Discussion

This study reveals the hierarchical behavioral mechanisms underlying Generation Z tourists’ travel intentions and their route selection in high-density urban environments. Structural equation modeling results show that subjective norms exert a significantly stronger influence on behavioral intentions than attitudes toward behavior or perceived behavioral control. This suggests that Generation Z tourists prioritize peer evaluations and group identity over individual judgment [65], largely shaped by their digital upbringing in a social-media-saturated environment filled with likes, shares, and comments. Compared with earlier studies applying the Theory of Planned Behavior (TPB) to general tourist groups in Western or low-density urban settings, the findings emphasize a stronger role of subjective norms and media-induced social pressure among Generation Z tourists in high-density contexts. In digitally active cities like Macau, social conformity mechanisms appear especially prominent, further validating the critical impact of social context on this cohort’s behavioral intentions.

Further analysis indicates that social media influence indirectly shapes behavioral intentions by reinforcing attitudes and subjective norms, while its direct impact on perceived behavioral control is limited. In contrast, the perceived built environment notably enhances perceived behavioral control by increasing tourists’ sense of route feasibility, illustrating a synergistic relationship between digital social information and physical spatial perception. These interacting cognitive mechanisms collectively influence travel motivations and route preferences [66,67]. Although Generation Z individuals exhibit independent values, their actual behavior often reflects a tendency toward social conformity, underscoring the need to incorporate dynamic social variables into predictive models. Future research could further explore this conformity through qualitative methods. Compared with earlier studies applying the Theory of Planned Behavior (TPB) to general tourist groups in Western or low-density urban settings, the current findings emphasize the stronger role of subjective norms and media-induced social pressure among Generation Z in high-density environments. Specifically, in digitally active cities like Macau, social conformity mechanisms appear more prominent.

Based on these behavioral insights, a hybrid route recommendation strategy is proposed. First, an influencing factor model is constructed, identifying media influence (MI), subjective norms (SNs), and perceived built environment (PBE) as key determinants of behavioral intention (BI), grounded in TPB and Social Influence Theory (SIT). Then, a structural equation model (SEM) is used to empirically validate the relationships among these variables. Following this, behavioral intention scores are transformed into weighted preference matrices and integrated into an Ant Colony Optimization (ACO) algorithm to generate multi-objective travel routes.

This hybrid approach demonstrates methodological advantages over traditional statistical or purely heuristic models by aligning route generation with both psychological and spatial factors. From a practical perspective, understanding the dominant influence of subjective norms and media enables tourism practitioners to design targeted marketing content, foster peer-aligned tourism products, and build interactive online communities. For urban planners, the results highlight the importance of integrating physical design with digital experience infrastructure to create urban environments that are accessible, navigable, and optimized for social sharing. Future work should incorporate real-time behavioral data from mobile apps, dynamic urban metrics, and intergenerational comparisons to further improve the adaptability and personalization of smart tourism services in complex urban contexts.

6. Conclusions

Taking Macau—a representative high-density tourism city—as the study area, this research focused on the travel behavior mechanisms of Generation Z tourists within urban space. A tourism route optimization model was proposed, integrating behavioral theory, environmental perception, and the Ant Colony Optimization (ACO) algorithm. The framework synthesizes the Theory of Planned Behavior (TPB), Social Influence Theory (SIT), and built environment perception factors. Through this study, not only is the theoretical framework for explaining generational travel behavior enriched but a closed-loop behavioral modeling logic is also established that links psychological intention with spatial path generation.

In the structural equation modeling (SEM) stage, nine hypotheses (H1–H9) were tested. The model demonstrated good fit (CFI = 0.855, TLI = 0.916, and RMSEA = 0.057). Path analysis showed that subjective norms (SNs) had the strongest impact on behavioral intention (BI), followed by attitude toward behavior (ATT) and perceived behavioral control (PBC). Media influence (MI) and peer influence (PI) significantly affected both ATT and SNs, while perceived built environment (PBE) significantly influenced PBC, supporting the inclusion of spatial perception in behavioral modeling. Based on the SEM results, a multi-objective path scoring function was developed using BI scores, interest preferences, and built environment attributes (e.g., walkability, greenery, safety). These scores served as heuristic inputs for the Ant Colony Optimization (ACO) algorithm, enabling the generation of tailored route types for different tourist profiles (e.g., culture-oriented, culinary-nature, art-immersive). To assess the practical relevance of the generated routes, social media text analysis was conducted using user-generated content (UGC) from platforms such as Xiaohongshu and TikTok. Keyword clustering and topic relevance analysis showed strong alignment between ACO-recommended POIs and popular user themes, demonstrating the model’s consistency across data sources and its real-world applicability. Overall, these findings confirm the feasibility of integrating behavioral theory with intelligent routing algorithms. This study contributes both a validated empirical framework and a transferable method for smart tourism systems that adapt to user preferences and urban spatial conditions.

Despite these contributions, the study has several limitations:

(1) Limited regional applicability:

The empirical data used in this study are primarily drawn from Macau. Given the city’s distinctive urban form and cultural context, the generalizability of the findings may be constrained. Caution is warranted when applying the conclusions to other cities, particularly those with markedly different socio-spatial structures.

(2) Lack of real-time dynamic factors:

The current path model does not sufficiently incorporate real-time dynamic elements, such as disruptions from unexpected events or sentiment fluctuations on social media. As a result, the model may exhibit limitations in adapting to rapidly changing urban environments.

To address these limitations, future research could explore cross-regional comparative studies, longitudinal analyses, and the integration of real-time data sources. Incorporating data from mobile applications, GPS trajectories, location-based services (LBSs), and emotional signals from social media could enhance the model’s robustness, generalizability, and capacity to support real-time decision-making in complex urban systems.