A Landscape Narrative Model for Visitor Satisfaction Prediction in the Living Preservation of Urban Historic Parks: A Machine-Learning Approach

Abstract

Urban historic parks face the dual challenge of achieving the living preservation of historic buildings while enhancing contemporary visitor satisfaction. In the context of accelerating urbanization and growing demand for immersive cultural experiences, it is increasingly important to conserve historical and cultural values while maintaining relevance and emotional engagement. This study adopts a mixed-methods approach to develop a predictive model for visitor satisfaction within the framework of living preservation, using Yingzhou West Lake in Fuyang City, Anhui Province, as a representative case. Qualitative methods were employed to identify key landscape narrative dimensions, while quantitative data from structured questionnaires highlighted critical experiential elements such as environmental restoration perception, flow experience, and cultural identity. Three machine-learning algorithms—random forest, Support Vector Machine (SVM), and XGBoost—were applied, with the most accurate model used to analyze the relative contribution of each component to visitor satisfaction. The findings revealed that immersive experiential elements play a central role in shaping satisfaction, while physical and cultural elements, particularly historic buildings and their contextual integration, provide essential structural and emotional support. This study offers data-driven insights for the adaptive reuse and interpretive activation of historic architecture, proposing practical strategies to harmonize cultural continuity with visitor engagement in the sustainable management of urban historic parks.

1. Introduction

In the context of globalization and informatization, cultural heritage tourism has not only become a significant driver of economic growth but also a vital platform for promoting cultural exchange. However, its rapid development has also brought about numerous sustainable development challenges, especially the increasingly prominent issue of protecting the authenticity of heritage [1]. As a cultural heritage country, China is increasingly focusing on the sustainable tourism development of urban historical and cultural parks, with particular attention to the living conservation of historical buildings and landscapes. The 14th Five-Year Plan for Culture and Tourism Development, issued by the General Office of the State Council in 2022, explicitly proposes to “promote the in-depth integration of culture and tourism” and emphasizes the need to “enhance the level of cultural heritage protection and inheritance, and innovate the way of utilization”, highlighting the need to “enhance the level of cultural heritage protection and inheritance, and innovate the way of utilization” [2]. A longitudinal study of World Heritage sites in China points out that there is a fundamental contradiction between heritage conservation and tourism development: on the one hand, tourism brings visibility and economic benefits; on the other hand, over-exploitation leads to resource decline, cultural dilution, and loss of authenticity [3,4]. Many urban historical heritage sites are already experiencing problems of over-commercialization, overloaded hospitality, and blurring of local cultural identities. The continuity and vitality of historical and cultural landscapes are seriously threatened.

In this context, how to strike a balance between tourism development and heritage conservation requires not only regulation at the policy and infrastructure levels, but also an in-depth understanding of visitors’ experience and emotional connection at heritage sites. Traditional satisfaction models focus on service quality and functional evaluation, neglecting the deeper experience of cultural and emotional dimensions [5]. In recent years, landscape narrative experience, as a kind of experience that integrates spatial construction and cultural expression, has gradually been regarded as an effective path to enhance visitor satisfaction, improve cultural participation, and continue the vitality of heritage. In order to meet this challenge, it is necessary to reconstruct the sustainable development strategy of historical heritage sites from visitors’ experience. As an emerging theoretical framework, landscape narrative experience combines spatial design with cultural narrative, emphasizes emotional arousal and cultural resonance, and has become an important means to enhance satisfaction and achieve long-term participation.

Specifically, narrative landscapes enhance visitors’ understanding of heritage significance and emotional resonance through the integration of cultural symbols and spatial fields; environmental restoration perception focuses on the positive impact of historic gardens on psychological restoration, which can effectively alleviate urban pressures and enhance psychological satisfaction [6]; flow experience emphasizes immersion and in-depth participation, which is particularly prominent in cultural heritage sites [7]; and cultural identity is the core driver of repeat visits and loyalty. Cultural identity is the sense of belonging and identity formed by visitors during their visit, which is the core driver of repeat visits and loyalty [8]. This study integrates three theoretical frameworks: landscape narrative theory, visitor experience theory, and machine-learning theory. Landscape narrative theory, proposed by Potteiger and Purinton, emphasizes the role of landscape as a medium of cultural expression [9]; visitor experience theory includes three key dimensions—environmental restoration perception, flow experience, and cultural identity—and explains the formation mechanism of visitor satisfaction from the aspects of psychological restoration, immersive participation, and identity, respectively.

Methodologically, the study uses machine-learning methods to construct a prediction model to explore the relationship between the dimensions of landscape narrative experience on visitor satisfaction. The random forest algorithm was specifically chosen due to its significant advantages in nonlinear modeling, feature importance analysis, and handling of missing data [10]. Compared with algorithms such as Support Vector Machine (SVM) and Gradient Boosting Tree (XGBoost), random forest is more suitable for modeling tourist satisfaction in terms of handling multi-dimensional and heterogeneous data and maintaining a high level of accuracy [11,12,13]. It has been demonstrated that machine-learning methods significantly outperform traditional regression methods in predicting satisfaction with tourism and cultural heritage [14].

Therefore, this study aims to explore the influence mechanism of landscape narrative experience on visitor satisfaction, establish a satisfaction prediction model based on machine learning, and provide a scientific basis for landscape design and operation management of urban historical and cultural parks. The core research questions include the following: (1) What are the main dimensions of landscape narrative experience in historical and cultural parks? (2) How does the random forest algorithm assess the relative importance of various experience dimensions and visitor experience characteristics in satisfaction prediction? (3) How can the random forest model in machine learning be used to extract key factors affecting visitor satisfaction and translate them into design and management strategies that support the living conservation of cultural heritage?

2. Literature Review

2.1. Landscape Narrative

Landscape narrative stems from the concept of narrative, a common form of expression that, from a philosophical and dialectical perspective, signifies the symbolic representation of a sequence of events with temporal logic or causal relationships [15]. It involves the application of narrative theory to landscape design, utilizing narratological theories and methods to analyze, comprehend, and evaluate the elemental attributes, spatial structures, and cultural semantics of landscapes, as well as their construction strategies. Potteiger and Purinton emphasize that “places constitute narratives, and landscapes are not merely settings for stories but evolving stories themselves,” highlighting the two-way interaction inherent in the landscape narrative process [16]. This underscores that the process is reciprocal rather than a one-way transmission. Landscape narrative not only addresses the functionality and practicality of the landscape but also prioritizes the creation of spaces imbued with emotional resonance and artistic ambiance through storyline and imagery. Employing landscape design from a narrative perspective, the spirit of place and the cultural connotations of landscape spaces are harnessed, focusing on enhancing communication and interaction between the subject and the landscape. This approach imbues the space with interest and poetry, encouraging individual participation, appreciation, reflection, and resonance [17].

In urban heritage parks, landscape narrative manifests across two dimensions, physical and cultural elements, collectively fashioning a multi-dimensional cultural experience space. Physical elements encompass natural and man-made elements. Cultural elements are divided into the narrative theme, structure and rhythm, artistic conception, folklore activities, and educational significance [18].

Physical elements construct the foundational narrative model through both natural and man-made components. Natural elements encompass terrain, water systems, and flora. The terrain design, featuring the “one pond and three mountains” layout, mirrors the pursuit of a mythical realm in traditional Chinese culture, inspired by the legend of the three immortal mountains—Penglai, Yingzhou, and Fangzhang. Examples include the layered islands in the Kunming Lake of the Summer Palace, which immerse visitors in a mythological ambiance [19]. The water system manifests as “living water,” where static bodies like lakes symbolize the sedimentation of history while flowing waters like streams represent the passage of time. For instance, the “Ten Scenic Spots” of West Lake narrate cultural tales through water system transformations [20]. Plants like bamboo symbolize a high cultural standard and quality of life, and pine trees represent longevity, conveying cultural meanings [21].

Man-made elements primarily include historic structures, cultural landscapes, and public art installations. Pavilions serve functional and cultural purposes, with pavilions embodying gathering places for scholars and refined individuals, while bridges symbolize communication and connection [22]. Cultural landscapes and public art recount historical narratives through mediums like rocks, sculptures (e.g., the sculpture depicting the Legend of the White Lady at Hangzhou’s West Lake), murals, and wall plaques, complemented by traditional park amenities, collectively forming a cohesive cultural milieu [23].

In summary, landscape narrative enriches the theoretical connotation and practical methods in landscape design by applying narrative theory. In urban heritage parks, landscape narrative unfolds through two dimensions: physical elements and cultural elements. The physical elements are divided into natural elements (terrain, water system, plants) and man-made elements (architecture, cultural landscape, and public art), and physical elements construct the basic landscape narrative model. In contrast, the cultural elements are divided into the narrative theme (theme of ancient style, the theme of historical figures), narrative structure and rhythm (attraction arrangement, tour route arrangement, narrative rhythm), artistic conception (humanistic conception, emotional), folk activities (festivals, folklore experience), and educational significance (transmitting historical and cultural heritage, children, and youth education). These elements are intertwined and work together to not only show the material form of the urban heritage park but, more importantly, to convey the deep historical and cultural connotations through rich narrative techniques, create a cultural experience space that can trigger visitor participation, reflection, and emotional resonance, and realize the sublimation of the landscape narrative from material space to cultural heritage.

Landscape narratives have increasingly garnered scholarly interest as a key medium linking spatial environments with social memory and cultural values, particularly in the context of urban heritage sites. As Cresswell argues, places are not merely physical settings but are imbued with meaning; landscape narratives represent a cultural process through which such meanings are actively constructed [24]. By engaging language, imagery, symbolism, and other representational modes, landscape narratives reinforce emotional identification and memory formation among visitors [25,26].

A substantial body of research has examined the material narrative carriers embedded in heritage landscapes, such as inscriptions, architectural forms, and decorative elements, primarily analyzing their symbolic significance and pathways of cultural transmission [27]. However, these studies tend to concentrate on textual and semiotic interpretation, often overlooking how visitors subjectively comprehend and respond to narrative content. This has left a notable gap in understanding the mechanisms by which narratives influence visitor perception and experience.

In response to this limitation, a number of scholars have begun exploring landscape narratives from an experiential perspective. For instance, Larsen and Svabo introduced the notion of “performative narratives,” which underscores the embodied and sensory dimensions of visitor participation in the narrative process [28]. Similarly, Boboc et al. highlight the communicative and affective potentials of multimodal narratives, particularly those delivered through digital tours and augmented reality, within the context of urban heritage interpretation [29].

Nevertheless, research remains underdeveloped in systematically addressing how narrative content translates into psychological and behavioral responses. Few studies have articulated clear frameworks that link narrative perception with affective engagement or satisfaction outcomes. To address this gap, the present study conceptualizes landscape narratives as antecedent variables and seeks to construct theoretical pathways that incorporate visitors’ psychological and emotional responses. In doing so, it offers a novel analytical perspective on how narrative-rich environments shape visitor satisfaction in urban heritage parks.

2.2. Visitors’ Experience

Visitors’ experience has always been a core issue in tourism research. In recent years, the integration of psychological dimensions has provided a more nuanced perspective for understanding the mechanisms underlying visitor experience. Notably, the dimensions of environmental restorative perception, flow experience, and cultural identity have been frequently adopted in heritage and natural landscape studies. Environmental restorative perception originates from Kaplan’s Attention Restoration Theory, which posits that natural or cultural environments can help individuals to regain mental energy through four key features: being away, extent, compatibility, and fascination [7,30]. Flow experience, proposed by Csikszentmihalyi, refers to the state of deep immersion, focus, and control that individuals experience during highly engaging activities. Cultural identity, on the other hand, emphasizes the individual’s sense of belonging and self-recognition within a specific cultural context [10]. Collectively, these dimensions reveal the underlying psychological mechanisms of visitor experience in terms of mental restoration, cognitive immersion, and emotional attachment.

2.2.1. Environmental Restoration Perceptions

Researchers have paid more attention to the psychological factors influencing visitor satisfaction in recent years. The theory of restorative perception states that locations with suitable ecological and landscape environments can provide psychological recovery and relaxation for visitor satisfaction overall. Environmental restorative perception refers to an individual’s perception of a specific environment that has qualities that promote physical and mental recovery [31]. This includes natural environments such as green spaces, parks, and waters, which have been shown to positively reduce stress, enhance emotional states, and promote cognitive recovery [32]. Much of the previous research on the restorative perception of environments is based on attention restoration or restorative environmental perception research in environmental psychology, representing a new research direction in integrating psychology and tourism.

The physical elements (natural and man-made elements) within urban historic parks help to promote relaxation among tourists by creating a tranquil and harmonious atmosphere, positively affecting tourists’ perceptions of environmental restoration. This state of relaxation effectively relieves stress and anxiety and promotes psychological healing, which is closely related to environmental restoration [33]. In addition, visitors’ experience of pleasure in the park is closely linked to psychological recovery. When visitors are immersed in the natural and cultural environments provided by parks, they are more likely to recover from negative emotions and stress. This opportunity for cognitive recovery can be realized through the rich historical and cultural contexts and unique physical elements within parks that help visitors to recover from attention fatigue [34]. In these spaces, visitors experience emotional fulfillment through emotional resonance with the cultural context, reducing stress and fatigue [35].

Environmental restorative perception significantly impacts visitor satisfaction, especially in urban heritage parks. The natural landscape and cultural ambiance of parks provide visitors with opportunities for psychological rehabilitation, which can effectively relieve accumulated psychological stress and reduce fatigue [36]. By focusing on four critical factors in restorative theory—being away, extent, compatibility, and fascination—visitor satisfaction and willingness to revisit can be further enhanced [37]. As an integral component of historical and cultural parks, physical elements within parks offer visitors visual enjoyment and serve as an essential source of psychological recovery and emotional connection through the fusion of nature and culture. In the study of optimizing and applying machine-learning models for predicting visitor satisfaction in urban historic parks in China, the perception of environmental restoration can serve as an effective predictor to aid in explaining and forecasting visitor satisfaction, thereby providing data-driven decision support for managers of these parks [38].

2.2.2. Flow Experience

The psychologist Csikszentmihalyi proposed the term “flow experience” as a state of mind with a sense of total concentration and engagement accompanied by pleasure. In tourism, the flow experience usually refers to the high concentration and focus visitors experience when deeply engaged and immersed [39]. This state of mind is pleasurable when visitors achieve their goals by interacting with an urban historic park’s natural landscape and cultural environment [40]. The prerequisites for a flow experience include clear goals, detailed feedback, and a balance of skills [41]. For example, visitors to urban historic parks reduce stress by visiting natural landscapes and old buildings, fulfilling clear goals. In addition, visitors need to learn about the park’s history and culture while visiting the park, which also fulfills the condition of skill balance. Detailed feedback is provided by the historical and cultural information that the park conveys to visitors’ experience, and when this information enhances visitors’ perception, the flow experience is realized.

Flow experience significantly impacts visitor satisfaction, especially in urban heritage parks, where it can deepen visitors’ emotional connection and cultural identity. Visitors’ experience in the park can enter a state of immersion through interaction with historical and cultural elements, and this experience not only satisfies visitors’ psychological needs but also makes them feel deeply connected to history and culture [42]. In the process, through the three characteristics of a flow experience–action–consciousness integration, focus on the task, and a sense of control, visitors’ experience becomes more focused on their experience. This heightened state of concentration allows visitors to forget about time and self and become fully immersed in the historical and cultural atmosphere, which in turn enhances their satisfaction with urban historic parks [43].

Research has shown that visitor satisfaction and emotional connection are significantly enhanced when they realize a flow experience through interaction with cultural elements in urban historic parks [44]. The flow experience has three key dimensions: prerequisites, characteristics, and outcomes. In the context of urban historic parks, cultural narrative themes, such as ancient styles and scenes commemorating historical figures, provide visitors with a clear goal, and this goal guides visitors to detach themselves from the minutiae of their lives and focus on experiencing the unique ambiance of culture and history to realize the prerequisites of the flow experience. At the same time, the narrative structure and rhythm maintain visitors’ engagement through the arrangement of attractions and the planning of visit routes, enabling them to maintain a sense of control over the landscape, i.e., the features of the flow experience. Finally, creating a humanistic and emotional atmosphere further enhances visitors’ experience of flow, enabling them to gradually enter a state of selflessness during their visit, forgetting about external distractions and thus achieving a state of peace of mind. This cultural resonance and emotional connection deepen visitors’ satisfaction and sense of belonging.

The flow experience is enhanced through participation in folk cultural activities such as festivals and folklore experiences, where visitors can actively engage with traditional culture. Educational activities, including cultural heritage transmission and youth education programs, create flow experiences by providing clear goals and immediate feedback [45]. This psychological state of flow, when integrated with various cultural elements, serves as an effective predictor of visitor satisfaction. By incorporating these cultural elements into machine-learning models, park managers can better understand how different cultural experiences contribute to visitor satisfaction and develop more targeted strategies for enhancing the overall visitor experience.

2.2.3. Cultural Identity

The cultural identity of visitors to urban heritage parks pertains to the degree to which they acknowledge the cultural identity or values embodied by the historic and cultural sites [46]. This identity mirrors visitors’ comprehension and empathy toward the values of history, culture, traditional practices, and cultural symbols in the park. Factors such as tourists’ cultural backgrounds, education levels, and travel purposes influence the intensity of their cultural identity [47]. In the context of tourism to historical and cultural parks, tourists’ interest in history, art, and traditional culture augments their cultural identity, while cultural displays and heritage activities within the parks further bolster this sense of identity [48].

Research has demonstrated a significant impact of visitors’ cultural identity on their satisfaction and loyalty [49]. When visitors’ experiences in urban heritage parks are imbued with cultural depth, they will likely develop a profound emotional connection with the park’s history and culture, leading to a greater appreciation of its cultural expression. Visitors with a strong cultural identity are more disposed to actively engage in cultural activities and attain a more profound cultural experience. This emotional connection contributes to their overall satisfaction with the tourism experience. In cultural tourism, tourists with a high cultural identity tend to perceive the environment of urban heritage parks more profoundly, enhancing their positive evaluations of the park’s environment [41].

In urban heritage parks, cultural identity is critical to enhancing visitors’ experience and satisfaction. Cultural identity includes three main components: emotional identity, local and historical identity, and social identity. Firstly, emotional identity involves tourists’ emotional attachment to the historical and cultural values carried by heritage sites, and this identity can inspire a sense of pride and belonging among tourists. Research has shown that by showcasing China’s rich historical heritage, especially by highlighting major historical events and unique cultural values, heritage parks can strengthen tourists’ emotional connection to their culture and make them feel a more profound sense of cultural pride [50].

Secondly, local and historical identity is reflected in tourists’ knowledge of specific regional cultures and historical figures, such as the influence of historical figures, so that tourists feel a common historical background and national emotions when visiting heritage parks. Such a cultural atmosphere enhances tourists’ sense of belonging and promotes their emotional connection to the heritage site [51]. In addition, social identity is closely related to tourists’ active participation in cultural activities; by participating in cultural heritage activities such as those in China, tourists can deepen their understanding of their own cultural identities and enhance their pride in the country and collective identity in the heritage park environment. This process promotes visitors’ cultural identity and increases their satisfaction and willingness to revisit, thus supporting the sustainable management of urban heritage parks.

Cultural identity exerts a direct positive influence on visitor satisfaction. When visitors encounter a pronounced cultural atmosphere in urban historic parks, their cultural identity intensifies considerably, elevating their favorable assessment of the park’s environment [41]. A deep understanding and empathy for the culture heightens their satisfaction with the tourism experience. Research has revealed that tourists with a stronger cultural identity can better perceive the unique allure of urban historic parks, thereby enhancing their satisfaction, particularly in cultural tourism, where cultural identity plays a pivotal role in shaping visitors’ experiences and satisfaction [52]. In optimizing and applying machine-learning models for predicting visitor satisfaction in Chinese urban historic parks, cultural identity can be an effective predictor, improving the precision of satisfaction forecasts and offering robust data support for managing these parks.

However, the existing literature has mostly studied the above dimensions in isolation and lacks an integrated perspective. These three dimensions have an obvious chain relationship in tourism behavior: restorative perception provides the psychological foundation, flow experience drives behavioral engagement, and cultural identity is the result of emotion and meaning construction [26]. This study will use this logic to construct a mediated pathway model of visitors’ experience, which will provide support for explaining the psychological mechanisms between narrative and satisfaction. In addition, in cultural identity research, few studies have focused on how tourists’ emotional identification and value resonance with heritage landscapes is achieved through a sense of place [53]. In this study, we propose to quantitatively examine the sense of place as a sub-dimension of cultural identity to fill the gap in existing research.

2.3. Visitor Satisfaction

Visitor satisfaction is a crucial indicator of the overall experience of visitors, directly influencing their likelihood to revisit and recommend the destination [54]. Relevant research has demonstrated that visitor satisfaction is pivotal for the sustainable development of tourism destinations and offers a theoretical foundation for tourism management [55]. It is the overall assessment of tourists’ satisfaction based on comparing the services received and their expectations [56]. This satisfaction not only forecasts future visitor behavior but also constitutes an essential means of evaluating the effectiveness and sustainability of a tourist spot [57]. Numerous studies have affirmed that visitor satisfaction significantly and positively impacts customer loyalty within destinations and across service and product categories [58]. Consequently, overall satisfaction is a vital benchmark for assessing destination quality in the tourism industry.

In urban historic parks, visitor satisfaction is critical to sustainable development. Firstly, higher satisfaction fosters a positive emotional bond and identification between visitors and heritage sites, encouraging compliance with conservation rules, minimizing environmental harm, and amplifying the societal reach of heritage conservation through verbal recommendations [59]. Secondly, elevated satisfaction levels can alleviate the management burden on heritage sites and bolster cultural and environmental sustainability. Satisfied tourists are more likely to offer constructive feedback, aiding managers in refining management strategies and service standards. They are also more inclined to revisit and recommend the site, providing financial backing for maintaining and restoring urban historic parks [60].

Furthermore, delighted visitors often exhibit a stronger sense of environmental protection and engagement in cultural heritage preservation. They can mitigate the adverse environmental effects of tourism and support local environmental conservation projects and cultural preservation endeavors [61]. Hence, enhancing visitor satisfaction is not merely the core of tourism development but also a strategic approach to promoting the sustainable development of urban historic parks.

Visitor satisfaction is not only an important indicator of the quality of the visitors’ experience, but also a core basis for the optimization of destination management and marketing. Traditional satisfaction studies have used structural equation modeling (SEM) or multiple linear regression for causal analysis [62]. In recent years, studies have begun to focus on the interaction mechanism between visitors’ satisfaction and visitors’ behaviors (revisit intention, word-of-mouth communication, etc.) and environmental perceptions (e.g., sense of safety, cleanliness) [63]. Particularly in the context of cultural heritage sites, visitors’ satisfaction levels are closely influenced by their perceptions of environmental ambience, cultural narratives, and facility services.

Although there have been many explorations of the factors influencing satisfaction, there are still three problems: firstly, the choice of variables is too single and lacks composite dimensional integration; secondly, the methodology mostly relies on linear models, which makes it difficult to deal with the nonlinear and multi-level relationships between variables; and thirdly, there is a lack of a dynamic prediction mechanism, which makes it difficult to achieve advance intervention and optimal management of satisfaction. Therefore, this paper intends to introduce machine-learning models to carry out satisfaction prediction analysis on the basis of the research on the path of satisfaction construction, explore the nonlinear causal relationship between “narrative–experience–satisfaction”, and carry out interpretable modeling of the factors influencing high satisfaction, so as to provide data support for management practice.

2.4. Machine Learning

Machine learning, a vital branch of artificial intelligence, enables systems to improve performance through algorithms without explicit programming [64]. In tourism management, machine learning excels in identifying key factors from large visitor datasets using data mining and pattern recognition, effectively predicting and analyzing visitor behavior and satisfaction through advanced models [65,66]. Unlike traditional statistical methods, machine learning handles complex nonlinear relationships and multi-dimensional data, enhancing prediction accuracy and providing deeper insights.

This study adopts three representative machine-learning algorithms—random forest, Support Vector Machine (SVM), and XGBoost—to predict visitor satisfaction based on landscape narrative experiences. These algorithms were selected not only for their strong performance in previous tourism and environmental studies [12,64], but also for their ability to model complex nonlinear relationships and interpret variable contributions. Random forest is well-suited for structured yet medium-sized datasets, offering robustness and clear interpretability of feature importance [64]. SVM excels in handling small, high-dimensional datasets and is frequently used in classification and satisfaction prediction tasks [67]. XGBoost, widely adopted in recent urban mobility and perception studies [12], balances accuracy and computational efficiency through iterative gradient optimization. These three algorithms provide complementary advantages for evaluating visitor satisfaction in a heritage landscape context.

Unlike traditional statistical methods, machine-learning models can capture complex interactions between variables—such as physical features, cultural perception, and experiential dimensions—and accommodate nonlinear causality [65,66]. To enhance interpretability, this study further integrates SHAP (Shapley Additive Explanations) analysis, which visualizes the contribution of each predictor to model output. The results reveal that demographic factors such as education level and place of residence have minimal predictive value, whereas experience-based features (e.g., cultural identity, environmental restoration perception) contribute significantly to satisfaction prediction.

While recent research has begun to explore machine-learning applications in tourism prediction [68], studies focused on visitor satisfaction in cultural heritage sites remain scarce. Most have emphasized model accuracy without addressing the interpretability of key variables or practical decision support. This study bridges that gap by combining machine learning with behavioral psychology theory, thereby advancing methodological innovation in satisfaction modeling and providing actionable insights for heritage site management [7,10,49,63].

In this study, we explore the influence mechanism and prediction model of tourists’ satisfaction with the help of machine-learning methods from the integration path of “narrative–experience–satisfaction”, trying to fill the above research gaps. On the one hand, restorative perception, heart flow experience, and cultural identity are integrated into the visitor experience structure; on the other hand, environmental perception, behavioral intention, and cultural identity are introduced into the satisfaction modeling process to construct predictable causal paths; finally, three representative algorithms, namely, random forest, SVM, and XGBoost, are introduced for comparative analysis, exploring their applicability and effectiveness in modeling satisfaction in urban heritage parks, aiming at theoretical integration and methodology development, and exploring the influence mechanism and prediction model with the help of machine learning. Finally, three representative algorithms, SVM, and XGBoost, are introduced for comparative analysis to explore their adaptability and effectiveness in modeling satisfaction in urban heritage parks, aiming to make a breakthrough in both theoretical integration and methodological innovation.

In summary, this study constructed a landscape narrative experience model for Chinese urban heritage parks, as shown in the figure below. The model includes physical elements (natural and man-made elements), cultural elements (narrative theme, narrative structure and rhythm, artistic conception, folk activities, and educational significance), and experiential elements (environmental restoration perception, flow experience, and cultural identity). This study innovatively introduces machine-learning techniques, providing a new methodological perspective for predicting and analyzing the impact of landscape narrative elements (as independent variables) on visitor satisfaction (as the dependent variable) through its ability to handle high-dimensional data and its advantage in assessing the importance of variables, a diagram of the model for this study can be seen in Figure 1.

3. Materials and Methods

3.1. Study Area

In this study, Yingzhou West Lake is selected as a research case. Yingzhou West Lake is in Fuyang City, Anhui Province, with geographic coordinates of 115°37′21″–115°39′18″ East, 32°54′05″–32°56′25″ North, and a total area of 24.32 square kilometers. The road around the lake bounds the east and north sides of the park, the west side relates to the nature reserve, and the south side is bordered by the road around the lake, detailing the road’s scope and the Caohe River’s wetland. As one of China’s four central West Lakes, Yingzhou West Lake has been listed as a Provincial Scenic Spot and a National Wetland Park, an urban heritage park integrating cultural heritage protection, landscape protection, and urban development functions.

Yingzhou West Lake has a deep historical and cultural heritage and a rich landscape narrative experience environment. During historical development, the Yingzhou West Lake has accumulated a rich cultural heritage as a place where the literati and elegant people gather. According to statistics, since the Tang and Song dynasties, the literati have left nearly 400 poems and songs, in which the works of Ouyang Xiu, Su Shi, and other historical figures not only vividly depicted the natural beauty of the West Lake but also contained a profound humanistic heritage. Regarding landscape creation, the Yingzhou West Lake integrates the essence of traditional Chinese garden art. It cleverly integrates poetry, calligraphy, painting, and other art forms into the landscape narrative, forming a unique space for cultural experience. In addition, the garden retains a wealth of folk cultural activities, further enhancing visitors’ experience of culture. These unique landscape narrative features and deep cultural deposits demonstrate the potential of the Yingzhou West Lake to become a world cultural landscape and provide a typical case study for researching the landscape narrative experience of urban historic parks in China.

In summary, the Yingzhou West Lake, one of China’s four famous West Lakes, is an ideal case for studying the narrative experience of urban historic park landscapes. Yingzhou West Lake possesses rich physical elements, including natural elements (terrain, water system, vegetation), man-made elements (historical buildings, humanistic landscape), and multiple cultural elements expressed through narrative themes, spatial rhythms, artistic conceptions, and folklore. These comprehensive landscape narrative features provide an excellent empirical basis for validating the theoretical framework of landscape narrative experience in urban historic parks. By field-investigating visitors’ experiences at Yingzhou West Lake, this study aims to explore how these physical and cultural elements enhance visitor satisfaction and support the sustainable development of urban historic parks in China.

3.2. Research Method

This study employs a mixed-methods approach, integrating qualitative and quantitative methods to analyze landscape narrative experiences in urban heritage parks. The qualitative phase includes an inventory survey and on-site observation, offering in-depth insights that guide questionnaire design and feature selection for machine-learning analysis. The quantitative phase validates and extends the qualitative findings, providing a holistic view of how landscape narratives impact visitor satisfaction. This combination allows the exploration of detailed qualitative experiences alongside broader quantitative trends.

The research process consists of three phases. First, a literature review constructs a theoretical framework identifying three dimensions of landscape narratives: physical elements (natural and man-made), cultural elements (narrative themes, structure and rhythms, artistic conception, folk activities, and educational significance), and experiential elements (environmental restoration perception, flow experience, and cultural identity).

Second, field research at Yingzhou West Lake refines these elements using resource inventory and on-site observation. Based on the findings, a questionnaire was designed to collect data on visitors’ perceptions and satisfaction. This data is analyzed using machine-learning techniques, employing random forest, SVM, and XGBoost algorithms for cross-validation. The most accurate model is used for the final analysis, revealing precise relationships between narrative elements and satisfaction.

3.3. Data Collection

3.3.1. Field Investigation

Field research will be conducted at Yingzhou West Lake to systematically identify, document, and interpret the physical and cultural elements that constitute its landscape narrative. This investigation aims not only to record existing features but also to reveal how tangible and intangible elements interact to shape visitors’ narrative experiences. For the physical elements, a resource inventory method will be applied to catalog the natural components (terrain, water system, plants, and trees) and man-made features (including architecture, cultural landscapes, and public art facilities). Special attention will be paid to the spatial distribution and functional characteristics of these elements. For the cultural elements, a participatory observation approach will be adopted. The research team will engage directly with tourist experiences by joining sightseeing routes, observing folk activities, and attending cultural events. Through this immersive engagement, the study will explore how visitors perceive and interpret cultural narratives, focusing on dimensions such as narrative themes, narrative structure and rhythms, artistic conception, folk cultural activities, and educational significance.

To enhance methodological transparency, a procedural diagram illustrating the full field investigation process—including the preparation stage of the investigation, data collection, data organization and analysis, and output—is proposed (see Figure 2). This will help to clarify the multi-phase nature of the fieldwork. A total of five trained investigators will participate in the on-site survey. Each team member will be assigned specific thematic tasks, ensuring both breadth and depth in data collection. This integrated method, combining structured observation with in situ participation, will provide a comprehensive understanding of the landscape narrative system at Yingzhou West Lake and its role in shaping visitor perception and satisfaction.

A range of software tools was employed during the on-site investigation and documentation to ensure clarity and professionalism in data presentation. Adobe InDesign 2024 was used to design the layout and structure of procedural diagrams, while Adobe Photoshop 2024 was utilized for image processing tasks. In addition, both handwritten notes and digital tools were used to record field observations and inventory details, ensuring the accuracy and completeness of data collection.

3.3.2. Survey of Questionnaires

This study employed a structured paper-based questionnaire, which was distributed on-site to collect quantitative data. The questionnaire consisted of four main sections. Yingzhou West Lake, the case site of this study, was selected due to its strong landscape narrative characteristics, which align well with the research focus on visitors’ experiential perception and cultural interpretation in heritage settings. The site’s rich historical and cultural context provides an ideal environment for exploring the relationships among narrative landscapes, visitor experience, and satisfaction.

• Physical Elements: This section includes natural elements (e.g., terrain, water systems, plants, and trees) and man-made elements (e.g., architecture, cultural landscapes, and public art facilities).
• Cultural Elements: This section covered narrative themes, narrative structure and rhythms, artistic conception, folk cultural activities, and educational significance.
• Experiential Elements: This section focused on environmental restoration perception, flow experience, and cultural identity.
• Visitor Satisfaction and Demographic Variables: Visitor satisfaction was measured in terms of overall satisfaction, while demographic variables included the number of visits, gender, age, education level, income, and place of residence.

All items, except for demographic variables, were measured on a five-point Likert scale (1 = strongly disagree, 2 = disagree, 3 = neutral, 4 = agree, 5 = strongly agree). The questionnaire was distributed to tourists visiting “Yingzhou West Lake.” Random convenience sampling was employed, with participants required to be 18 years or older. Each questionnaire took approximately 10–15 min to complete. The target sample size was 10% of the estimated 3000 daily visitors, resulting in a target of 300 questionnaires for the main study and 30 for the pilot test. Data collection took place at Yingzhou West Lake on 12, 13, 19, 20, 26, and 27 August 2023. A total of 380 questionnaires were returned, of which 340 were valid after excluding incomplete responses, yielding a valid response rate of 93.75%.

3.4. Data Analysis

3.4.1. Measurement Scale Validation

Before conducting the primary data analysis, the measurement scales will be tested for reliability and validity to ensure the accuracy and reasonableness of the questionnaire’s measurements. Reliability analyses will be conducted using Cronbach’s alpha (Cronbach’s coefficient), which will assess the internal consistency of the scale items and ensure that the scale possesses sufficient reliability. Additionally, the measurement model will be validated through a confirmatory factor analysis (CFA) to test the structural validity of the scale and confirm that the factors and indicators accurately reflect the dimensions of the landscape narrative experience.

3.4.2. Machine-Learning Model Analysis

This study uses three machine-learning algorithms—random forest (RF), Support Vector Machine (SVM), and XGBoost—to analyze the relationship between landscape narrative features and visitor satisfaction. A 10-fold cross-validation process was applied to ensure reliability and minimize overfitting risks. Metrics such as accuracy, precision, recall, and F1-score evaluated model performance, with the model showing the highest average accuracy across folds selected for further analysis. Python (version 3.10) and libraries like Pandas, Scikit-learn, SHAP, and XGBoost facilitated data preparation, cleaning, and encoding. Data were split into training and test sets (80:20), and SVM data were standardized using Scikit-learn’s Standard Scaler [55,56].

Random forest was trained to calculate feature importance, leveraging its robustness in handling nonlinear relationships and large datasets. SVM proved effective in high-dimensional spaces, while XGBoost demonstrated efficiency in handling missing data and regularization. After cross-validation, the model with the best balance of accuracy and performance metrics underwent feature importance and interpretability analysis. SHAP values quantified feature contributions to predictions, providing insights into their influence on visitor satisfaction. Partial dependence plots further illustrated the relationships between critical variables and satisfaction [57].

Data visualization tools like Matplotlib 3.8.2 and Seaborn 0.13.2 presented results. Correlation heatmaps identified relationships between features and satisfaction, while feature importance and SHAP summary plots visualized feature contributions and interactions. This approach ensures robust findings, bridging predictive performance and interpretability, and offers actionable insights into how landscape narrative features influence visitor satisfaction [56,57]. All data processing, analysis, and visualization in this study were carried out using Python (version 3.10). The use of programming ensured rigorous data cleaning, effective visual representation, and interpretable modeling, contributing to the overall scientific validity and reproducibility of the analytical process.

4. Results

4.1. Results of Inventory

In this section, a qualitative research inventory survey of the physical features of “Yingzhou West Lake” identifies the specific composition of its natural and man-made elements to reveal how these features construct the landscape narrative features of the urban historic park.

4.1.1. Natural Elements

The natural elements of the West Lake of Yingzhou present a unique narrative conception through the traditional layout of “one pond and three mountains,” which demonstrates the method of the Chinese classical gardens’ algal wells. The “one pool” is the lake, which provides the park with a sense of open space and the contrast between reality and reality of the reflection in the water, creating a peaceful and deep atmosphere; the “three mountains” are symbolically arranged on the water, symbolizing the imagery of the Penglai Immortal Realm or Immortal Mountain on the sea, which conveys the spatial symbolic meaning of the traditional culture. The “three mountains” are symbolically placed on the water, symbolizing the image of Penglai Immortal Land or Immortal Mountain on the sea, conveying the spatial symbolism of traditional culture. This layout enriches the visual level and provides visitors with a narrative experience by combining natural landscape and artistic conception, as shown in Figure 3.

The natural elements of West Lake in Yingzhou primarily encompass the terrain, water system, and abundant flora, which collectively shape the “one pond and three mountains” configuration, a classic landscape design technique in traditional Chinese gardens. This traditional layout imparts a sense of visual hierarchy and spatial arrangement to the park and conveys narrative concepts through the symbolism embedded in the natural scenery. Specific plant groupings enhance the landscape’s depth and color palette, with 11 key areas featuring landscape plants, as shown in Figure 4.

• Ginkgo, Luan, Peach Tree Plant Viewing Area;
• Almond and Peach Blossom Plant Landscape Area;
• Flower, Peach, Plum, Bamboo, and Pine Plant Landscape Area;
• Willow Plant Landscape Area;
• Lotus Plant Landscape Area;
• Peach, Cherry, and Willow Planted Landscape Areas;
• Pine, Poplar Plant Landscape Area;
• Acacia, Luan, Plant Landscape Area;
• Pine, Luan Plant Landscape Area;
• Acorus Calamus Plant Landscape;
• Fir Plants Landscape Area.

These plant elements embody the diversity and storytelling of traditional Chinese gardens through the rich blend of species and distinctive layout, allowing visitors to immerse themselves in the historical and cultural concepts conveyed by the visual and atmospheric qualities of the natural landscape.

4.1.2. Man-Made Elements

Man-made elements, mainly historic architecture, cultural landscapes, and public artworks, are also essential in the Yingzhou West Lake landscape narrative, as shown in Figure 5. Historical buildings include the following:

• Feigai Bridge;
• Xiefang Garden;
• Hu Pavilion;
• Lan Garden;
• Yi Garden;
• Hanpu Garden.

Cultural landscapes and public artworks of Yingzhou West Lake demonstrate the architectural aesthetics of classical gardens while providing an essential historical context for the landscape narrative experience, as shown in Figure 6. Cultural landscapes and public artworks, include the following:

• Statues;
• Stone tables and walls;
• Facilities (signage and guiding facilities, resting facilities, lanterns, decorative fixtures, gates, and pagodas).

These enhance the landscape’s beauty and enrich the park’s narrative theme through historical or cultural symbols. They also complement the overall landscape with their traditional design characteristics.

Through the harmonious integration of natural and man-made elements, “Yingzhou West Lake” embodies the rich features of classical Chinese gardens in its landscape narrative. It expresses historical and cultural connotations, enabling visitors to experience a strong cultural atmosphere and storytelling. Combining these physical and cultural elements shapes the visual aesthetics and conveys a unique cultural conception that forms the park’s narrative space.

4.2. Results On-Site Observation

4.2.1. Narrative Theme

Historical figures and poetic culture deeply influence the narrative theme of the West Lake of Yingzhou, and this theme deeply integrates the lives of historical celebrities and literary heritage into the landscape of the lake area, giving it rich cultural significance.

• Ancient Styles: The natural beauty of the West Lake in Yingzhou has inspired a rich tradition of poetry and art. Poets and scholars throughout the ages have been captivated by the beauty of its lake and mountains and have composed more than 400 poems praising the West Lake, including as many as 187 during the Tang and Song dynasties. The poetic culture formed around the West Lake has highlighted its importance as a symbol of traditional Chinese aesthetics and cultural inspiration. It has become a unique blend of natural beauty, history, and culture.
• Historical Figures: Due to its stunning natural landscape, Yingzhou’s West Lake has drawn many renowned literati and calligraphers, including iconic figures like Ouyang Xiu and Su Shi. Their visits transformed West Lake into a prominent cultural landmark in ancient China, cementing its place as an essential destination for scholars and artists.

The West Lake of Yingzhou is a natural landscape and a “narrated” landscape with deep historical and cultural connotations. The interweaving of historical figures and poetic culture gives the lake area a multi-layered narrative dimension, enabling visitors to experience historical continuity and cultural resonance and deepening their identification with and emotional connection to the urban historic park.

4.2.2. Narrative Structure and Rhythm

Figure 7 and Figure 8 show the narrative structure and rhythm of the Yingzhou West Lake through the arrangement of attractions, the master plan, the narrative rhythm control points, and route planning.

• Arrangement of attractions (Main Entrance, Secondary Entrance, Song Dynasty Memories, Yinzhou Nostalgia Area, Ancient Yingzhou Flower Culture Area, Scenic Byway around the Lake Area, Main Entrance Service Area);
• Master plan (South Entrance, Main Visitors’ Service Centre, Emerald Waves and Glazed Reflections, Ying Chrysanthemum Garden, Xiefang Pavilion, Around the Lake Road, Peony Garden, Su Causeway, Fei Gai Bridge, Hanchun Garden, Yi Yuan Bridge, Yi Garden, Lan Garden, Island in the Middle of the Lake, Hu Pavilion, Jetty, Secondary Visitors’ Service Centre, West Lake Management Committee, North Parking, South Parking);
• Narrative rhythm (Main Landscape Control Points, Secondary Landscape Control Points, View Corridor, Landscape Interface);
• Route planning (Walking Route, Cycling Route, Boat Cruise Route, Jetty).

4.2.3. Artistic Conception

• Humanistic Conception: With its rich humanistic tradition, the West Lake of Yingzhou blends natural beauty with cultural narratives. Poems penned by renowned writers such as Ouyang Xiu and Su Shi impart a multi-layered historical and emotional depth to the lake, elevating it beyond a mere scenic spot. The lakeside scenery offers visual pleasure and invites visitors into an emotional dialogue with the past’s culture, transforming the West Lake into a narrative space where history, culture, and nature intertwine.
• Emotional Conception: The West Lake of Yingzhou evokes solid emotional resonance and fosters a profound sense of local cultural identity among visitors through the harmonious fusion of natural landscapes with the history and culture of Fuyang City. As visitors embark on their journey, they sense the deep affection that generations of literati have held for this land, making the West Lake a bridge to the past. Through this emotional journey, visitors enjoy visual delights and deepen their understanding and connection to the history and culture of Fuyang City.

4.2.4. Folk Cultural Activities

• Festivals (Chinese New Year, Mid-Autumn Festival, etc., as shown in Figure 9).
• Folklore experience (Archery, Tea Culture, Ancient Chinese Traditional Costume Experience, Calligraphy, as shown in Figure 10).

4.2.5. Education Significance

• Transmit History and Cultural Heritage: The West Lake of Yingzhou is not merely a natural scenic spot but also an invaluable heritage site with profound history and culture. Landmarks and structures within the lake area, such as Feigai Bridge, Yi Garden, Xie Fang Pavilion, Su Causeway, Orchid Garden, and others, are steeped in historical tales and cultural significance. By preserving and showcasing these landscapes, tourists can perceive the cultural footprints left by ancient literati and ink masters, thereby enhancing their comprehension of major historical events in Fuyang City. This cultural heritage transcends mere visual experience and triggers emotional resonance among visitors through stories, poems, architecture, and other multi-dimensional means, serving as a crucial conduit for transmitting history and culture.
• Children and Youth Education: The West Lake of Yingzhou offers an ideal venue for children and youth educational activities. Through diverse study programs, students can deepen their understanding of the history, culture, and natural resources of Yingzhou West Lake, fostering a sense of identity and affection for traditional culture, as shown in Figure 11. Simultaneously, the educational curriculum emphasizes cultivating students’ practical abilities and teamwork spirit, enabling them to acquire historical knowledge, ecological conservation concepts, and traditional art skills through interaction and hands-on experiences.

In the context of landscape narrative, both the material and cultural elements of Yingzhou West Lake significantly influence tourists’ satisfaction. The spatial distribution and functional characteristics of natural features—such as terrain, water systems, and vegetation—as well as man-made landscapes, directly shape visitors’ environmental perception and touring comfort. These aspects form the foundation for restorative experiences and the development of spatial identity. A well-designed landscape layout and thoughtful facility planning can enhance aesthetic appreciation and convenience, thereby improving overall visitor satisfaction. Furthermore, cultural elements—including narrative themes, structural rhythm, mood creation, folk activities, and educational value—deepen the emotional and cognitive engagement of tourists. Participatory observation reveals that immersive involvement in folk events, cultural storytelling, and site-specific experiences fosters emotional resonance and cultural identification. This sense of connection and understanding significantly elevates visitors’ evaluations of their overall experience. Thus, while the physical landscape sets the stage for the narrative, cultural components provide its content. Together, they interact to form the core mechanism underlying tourist satisfaction.

4.3. Results of Survey

4.3.1. Reliability Analysis

Reliability is an indicator that assesses a measurement instrument’s consistency, stability, and reliability (e.g., questionnaire, scale, etc.) during the measurement process. In other words, reliability means the trustworthiness of the measurement results, which is usually judged by the value of the Cronbach alpha coefficient. The Cronbach alpha coefficient for the physical elements is 0.91, the Cronbach alpha coefficient of the cultural elements is 0.9245, the Cronbach alpha coefficient of the environmental restoration perception is 0.942, the Cronbach alpha coefficient of the flow experience is 0.913, the Cronbach alpha coefficient of cultural identity is 0.948, and the Cronbach alpha coefficient of visitor satisfaction is 0.890. In summary, the reliability value of each dimension is more significant than 0.7, indicating that the reliability of the questionnaire is good, as shown in

Table 1. Reliability analysis.

Elements	Item Count	Cronbach α
Physical Elements	PN1 1–PN3 1, PM4 2, PM5 2,	0.910
Cultural Elements	CNT6 3, CNT7 3, CAA8 4, CRP9 5, CAC10–CAC13 6, CFC14 7, CFC15 7, CES16-18 8	0.9245
Environmental Restoration Perception	ERP19-28 9	0.942
Flow Experience	FE29-34 10	0.913
Cultural Identity	CI35-41 11	0.948
Visitor Satisfaction	VS42 12	0.890

¹ Physical element—natural elements (PN). ² Physical element—man-made elements (PM). ³ Cultural element—narrative theme (CNT). ^4,5 Cultural element—narrative structure and rhythm (CAA, CRP). ⁶ Cultural element—artistic conception (CAC). ⁷ Cultural element—folk cultural activities (CFC). ⁸ Cultural element—educational significance (CES). ⁹ Experience element—environmental restoration perception (ERP). ¹⁰ Experience element—flow experience (FE). ¹¹ Experience element—cultural identity (CI). ¹² Visitor satisfaction (VS).

below.

4.3.2. Validation Factor Analysis

In this paper, confirmatory factor analysis (CFA) was used to test the overall fitness of the research model. These metrics help the researcher to determine if the model can reasonably explain the data and if the model is too complex or too simple. The overall model combined with the CFA results showed that CMIN/DF = 1.319, RMR = 0.023, GFI = 0.917, AGFI = 0.905, NFI = 0.941, IFI = 0.985, TLI = 0.984, CFI = 0.995, and RMSEA = 0.026. All model indicators meet the requirements, indicating a good model fit, as shown in

Table 2. Validation factor analysis.

Model Fit Indicators	Statistical Value	Standard Value
CMIN/DF	1.319	1–3
RMR	0.023	0.05
GFI	0.917	≥0.9
AGFI	0.905	≥0.9
NFI	0.941	≥0.9
IFI	0.985	≥0.9
TLI	0.984	≥0.9
CFI	0.985	≥0.9
RMSEA	0.026	≤0.08

.

Reliability and validity tests, including Cronbach’s alpha for internal consistency and CFA for construct validity, ensured that the measurement instrument accurately captured the intended dimensions of the landscape narrative experience. These tests confirmed the robustness of the data and provided a solid foundation for subsequent research. Random forest analyses were conducted by verifying that each variable reliably represented its respective construct.

4.3.3. Algorithm Validation and Optimal Selection

This study evaluates the performance of three machine-learning models—random forest, Support Vector Machine (SVM), and XGBoost—in predicting the impact of landscape narrative features on visitor satisfaction. Hyperparameter tuning was conducted for each model through grid search and cross-validation, and their performance was evaluated based on the optimal hyperparameter combination.

Figure 12 illustrates the cross-validation accuracy of the three models. Support Vector Machine (SVM) achieves the best performance with a regularization parameter C= 1, kernel parameter γ = 0.1, and the Radial Basis Function (RBF) kernel, achieving a cross-validation accuracy of 62.12%, the highest among the three models. The random forest model performs best with a maximum tree depth of 20 and 200 trees, yielding a cross-validation accuracy of 59.77%. The XGBoost model achieves a cross-validation accuracy of 58.50% with a tree depth of 7, a learning rate of 0.1, and a feature sampling ratio of 0.8. A comprehensive analysis shows that Support Vector Machine (SVM) exhibits the most superior predictive performance.

The classification performance of Support Vector Machine was further analyzed in detail, as shown in Figure 13, which presents the confusion matrix for the SVM model. From the confusion matrix, it can be observed that the model achieves the highest classification accuracy for category 3, while its performance is weaker for categories 1, 4, and 5. This indicates that although the SVM model performs best overall, there is still room for improvement in classifying specific categories.

The Support Vector Machine (SVM) was ultimately selected as the optimal model due to its superior accuracy and classification performance. Random forest and XGBoost, as alternative models, still offer the potential for further optimization in future research. This study’s findings provide an important reference for data-driven decision-making in landscape design and management, highlighting the critical role of nonlinear features in predicting visitor satisfaction.

4.3.4. SHAP Analysis and Feature Interpretation

This study uses the SHAP (Shapley Additive Explanations) method to analyze the Support Vector Machine (SVM) model’s prediction results and understand feature contributions. SHAP quantifies each feature’s influence, importance, and direction, providing insights for optimizing landscape design and visitor satisfaction management.

Through the replacement feature importance analysis (shown in Figure 14), “P-NE2”, “C-AC13”, and “E-ERP26” were identified as the most important model features with the highest average importance values (0.0126, 0.0109, and 0.0095), respectively. These features significantly influence the predictive performance of the model and are key drivers of visitor satisfaction. In addition, features such as “C-SR9”, “E-FE31”, and “P-ME4” also contributed significantly to the model prediction and played an important role in capturing the potential association between visitors’ behavior and satisfaction. On the contrary, features such as “education”, “residence”, and “income” are of lower importance, indicating that their influence on the model output is more limited, which could be considered for simplification in subsequent model optimization.

The SHAP summary graph (shown in Figure 15) provides a more intuitive picture of the contribution and influence of each feature. The analysis shows that the feature “P-NE2” has a significant positive contribution to the model prediction, and its SHAP value increases significantly when the feature value is higher, indicating that it can effectively improve the accuracy of the prediction results. Meanwhile, the contribution of the feature “C-AC13” to the prediction in different eigenvalue ranges shows obvious nonlinearity, which reflects its complex mechanism. The SHAP values of the tail features, such as “education”, have a narrower distribution and are concentrated around the zero value, which further confirms the limited contribution of these features to the model. Overall, the SHAP values of the core features have a wide range, indicating that these features play an important role in predicting visitor satisfaction.

To explore the mechanism of model prediction for individual samples, the SHAP Force Plot (shown in Figure 16) provides the study with a precise and personalized analysis tool. By analyzing the prediction path for a particular sample, the positive and negative effects of features on the prediction results can be observed. For example, in one of the test samples, the feature “C-AC10” made a significant negative contribution to the predicted value, while the features “E-ERP25” and “C-NT6” drove an increase in the predicted value. The length and color of the arrows in the Force Plot are a direct reflection of the balance of positive and negative features, suggesting that the model output is the result of multiple features working together. This individual-level interpretation provides a basis for more refined model optimization and paves the way for a deeper understanding of the mechanisms of key features.

According to the SHAP dependency analysis, the five subplots (Figure 17a–e) systematically illustrate the influence patterns of key features. Figure 17a (0.02, 0.00, −0.04, −0.06, −0.08) indicates a weak or negative contribution of the feature “P-NE2” across non-class 1 categories. In contrast, Figure 17b (1.0–3.0) and Figure 17c (3.5–5.0) demonstrate a strong positive contribution of this feature within class 1, with contribution values increasing progressively alongside feature values, from 1.0 to 5.0. This stepwise growth underscores a robust positive correlation at higher feature levels. Although the current visualization does not include features such as “C-AC13” and “E-FE29,” similar patterns—if observed in Figure 17d or Figure 17e—may reveal alternating SHAP values, suggesting a bidirectional contribution mechanism. This analysis not only confirms the inter-class differences in the behavior of “P-NE2” (i.e., strong positive impact in class 1 versus minimal or negative impact elsewhere) but also offers practical insights for feature optimization: high “P-NE2” values should be prioritized for class 1 predictions, while down-weighting may be advisable in other categories. Additionally, further SHAP analyses of features exhibiting dual contributions are recommended to enhance explanatory depth.

Finally, the SHAP decision diagram (shown in Figure 18) provides a comprehensive picture of the cumulative contribution of the model to the multisampling predictions. The results show that “C-ES18”, “C-NT6”, and “E-FE29” are the features that contribute the most to the model output, and the cumulative contribution paths of these features are consistent. The cumulative contribution paths of these features show consistency, indicating that the model can consistently utilize these features for prediction. Most of the features contribute positively, driving the predicted values to the positive region, while a few features show negative contributions, reducing the model output values. The consistency of the sample paths in the SHAP decision diagram further validates the robustness of the model across samples, providing a basis for subsequent model optimization. Meanwhile, the optimization or simplification of low-contribution features can also improve the efficiency and interpretability of the model to some extent.

In conclusion, the SHAP analysis not only enhances the transparency of the model by revealing the key features of the Support Vector Machine model and its mechanism of action but also provides scientific guidance for optimizing the prediction model of visitor satisfaction. The analysis shows that features such as “C-ES18”, “C-AC13”, and “E-FE29” have significant contributions in multiple categories and are the core model prediction performance drivers. On the other hand, low-contribution features such as “gender” and “residence” have a limited impact on the prediction results, and it is recommended that they be filtered and optimized in future studies. By combining the results of SHAP analysis, this study proposes a new idea of data-driven optimization of landscape design and visitor experience management and promotes the scientific and practicality of the prediction model in practical application.

5. Discussion

This study investigates the influence of landscape narrative experience on visitor satisfaction by integrating machine-learning algorithms (Support Vector Machine, random forests, XGBoost) and SHAP (Shapley Additive Explanations) analysis, with Yingzhou West Lake serving as a representative case of an urban historic park. The results emphasize the significance of physical, cultural, and experiential elements in shaping the visitor experience and supporting the living preservation of historic buildings.

This study integrates multiple machine-learning algorithms (SVM, random forest, and XGBoost) and SHAP analysis to investigate how landscape narrative experiences influence visitor satisfaction in urban historic parks, using Yingzhou West Lake as a case study. The findings highlight the differential impacts of physical, cultural, and experiential dimensions on satisfaction, offering both theoretical insights and practical guidance for the living preservation of historic buildings.

In terms of physical components, natural features such as terrain, water systems, and vegetation, along with man-made elements like historic buildings and cultural structures, form the spatial foundation of narrative experience. High-impact features such as “P-NE2” (natural environment composition) and “C-AC13” (cultural architecture) demonstrate that these elements not only serve as historical carriers but also enhance visitor immersion and identification. This aligns with the argument by Dai et al. that authenticity in historic landscapes arises from the interplay between natural and cultural elements [3]. Similarly, Zhou et al. (2023) have emphasized that symbolic spatial elements and narrative structures are key to emotional engagement in urban heritage environments [26].

Cultural and experiential dimensions played an even more significant role in predicting satisfaction, confirming the growing importance of user-centered design in landscape narratives. For instance, “C-ES18” (educational significance) and “E-ERP26” (environmental restoration perception) reflect the dual needs of visitors for emotional resonance and cognitive stimulation. These findings are consistent with Kaplan and Kaplan’s theory of restorative environments and echo the conclusion of Mo et al. that immersive cultural storytelling significantly enhances satisfaction in heritage contexts [7,69,70].

Moreover, experiential factors such as cultural identity and flow experience were shown to significantly influence satisfaction, underscoring the need to optimize landscape strategies across aesthetic, cultural, and psychological dimensions. Previous research emphasized the importance of cultural context in shaping immersion [71]. The present study builds on that conclusion through quantitative modeling and provides a replicable methodological framework.

The survey instrument exhibited high reliability and validity, with Cronbach’s α coefficients exceeding 0.7 and a KMO value of 0.871, alongside a significant Bartlett’s test (p < 0.001). These indicators confirm the internal consistency and structural validity of the measurement model. In particular, cultural and experiential dimensions exhibited strong convergent validity, supporting Hypotheses H2 and H3—that narrative cultural elements and subjective experience factors have a significant positive impact on visitor satisfaction. For instance, “C-AC13” had a mean score of 4.27 with a standard deviation of 0.61, indicating broad agreement among respondents on the importance of cultural architecture. Additionally, “E-FE29” (flow experience), though a non-physical variable, displayed high marginal contributions in SHAP analysis, confirming its influence on satisfaction through psychological immersion.

Among the three models, Support Vector Machine achieved the highest predictive accuracy, reflecting the robustness of the variable structure and survey design. SHAP analysis provided an intuitive interpretation of variable importance and revealed that demographic factors such as education and place of residence contributed minimally to prediction, thereby suggesting a shift in design priorities from static profiling to dynamic experience optimization. These insights support the development of landscape strategies that are driven by culture, guided by narrative, and enhanced through experience. As Zhang and Chen (2020) argue, the emotional stickiness of heritage sites is shaped by contextual design and narrative activation—an idea reaffirmed by the empirical findings of this study [41,49,72].

6. Conclusions

This study explores the influence of landscape narrative experience on visitor satisfaction through the integration of machine-learning algorithms—namely, Support Vector Machine, random forest, and XGBoost—complemented by SHAP (Shapley Additive Explanations) analysis. Using Yingzhou West Lake as a representative case of an urban historic park, the findings highlight the multifaceted role of physical, cultural, and experiential dimensions in shaping visitor engagement and supporting the living preservation of historic buildings.

From a physical standpoint, natural elements such as topography, water bodies, and vegetation, alongside man-made structures like heritage architecture and culturally symbolic installations, form the structural basis for landscape narratives. Key features such as “P-NE2” and “C-AC13” illustrate the significance of conserving and meaningfully integrating these components to enhance the experiential value. Historic architecture, in particular, gains renewed vitality when narratively embedded within the landscape, serving both as a testament to authenticity and as a conduit for cultural continuity.

Culturally and experientially, attributes such as narrative theme, artistic conception, and educational value (e.g., “C-ES18”, “C-AC13”) are essential to deepening visitors’ cultural resonance. Meanwhile, experiential factors—namely, environmental restoration perception, flow experience, and cultural identity (“E-ERP26”, “E-FE29”)—play a decisive role in fostering emotional satisfaction and psychological well-being. These results reinforce the need for immersive and context-aware design strategies that actively support both heritage interpretation and visitor fulfillment.

In terms of methodological contribution, model comparisons indicate that Support Vector Machine achieved the highest prediction accuracy, while SHAP analysis provided interpretive clarity regarding the relative impact of each variable. The limited predictive value of demographic variables such as education level and place of residence suggests a redirection of design emphasis toward experience-centric elements. SHAP’s transparent interpretability further enhances the applicability of machine learning in the evidence-based management of urban historic parks.

Altogether, this study offers both theoretical insight and practical implications for achieving synergy between heritage conservation and contemporary user experiences. It underscores the potential of narrative-driven and technologically informed strategies in sustaining the vitality of historic environments. Looking ahead, future research could explore how this modeling framework performs across different cultural and geographic settings, particularly in non-Chinese contexts where heritage interpretation may follow different narrative traditions. Additionally, integrating longitudinal or real-time behavioral data—such as movement tracking, emotional analytics, or social media interaction—may offer further precision in evaluating the dynamic relationship between visitors and landscape narratives. Expanding the dataset to include multisensory and participatory variables could also enhance the richness of experiential modeling and inform more holistic approaches to the adaptive reuse and public communication of historic landscapes.

7. Recommendations

Based on the findings of this study, the following recommendations are proposed for the landscape design and management of urban historic parks, with a particular focus on achieving the living preservation of historic buildings:

(1)

Optimize environmental restoration perception and immersive experiences.

Enhancing the psychological restorative function of the natural environment and the fluidity of visitor engagement is key to increasing satisfaction. Design strategies should include optimizing greening layouts, enriching water systems, and integrating natural landscapes with cultural meanings linked to historical buildings [73]. Creating immersive experiences through structured narrative paths, adaptive pacing of attractions, and interactive technologies (such as AR/VR) can help to situate historic architecture within dynamic and engaging visitor journeys, thereby reinforcing both experiential quality and heritage appreciation.

(2)

Strengthen the integration and expression of cultural elements rooted in historic architecture.

Cultural elements—especially those tied to historical buildings—play a crucial role in fostering emotional resonance and a sense of place. It is recommended to integrate local history, traditional symbols, and vernacular art forms into the interpretive framework of historic buildings. Strategies such as curated storytelling, interactive cultural displays, and seasonal folk activities can deepen narrative structures and reinforce the identity and continuity of historic spaces, aligning with the goals of living heritage preservation.

(3)

Prioritize core material elements and refine resource allocation.

Material components—particularly historic buildings and landmark natural features—form the foundation of landscape narratives and contribute directly to satisfaction. Management should prioritize the conservation, adaptive reuse, and interpretive revitalization of these key elements to ensure they remain functionally integrated within contemporary park life. Meanwhile, resources associated with less impactful demographic factors (e.g., education level, place of residence) as identified in SHAP analysis can be moderately reallocated to focus on high-contribution elements, enhancing both operational efficiency and visitor experience.

(4)

Implement data-driven and adaptive heritage park management mechanisms.

Establish a feedback system to regularly collect and analyze visitor experiences through machine-learning tools. Techniques like random forest and SHAP can help to reveal the evolving impact of various design and management strategies, allowing for dynamic adjustments that align with both heritage conservation goals and visitor expectations. Such an adaptive management approach echoes recent urban climate research, which emphasizes the role of data-driven tools in identifying behavioral responses and optimizing public engagement strategies under complex environmental conditions [74]. This data-informed framework supports sustainable planning, ensures effective resource use, and facilitates the coexistence of preservation and innovation in the long-term development of urban historic parks.