A traffic optimization method and system for park vitality improvement

Abstract

This application discloses a traffic optimization method and system for enhancing park vitality: It acquires multi-source data on urban parks in a target area; calculates park visitor density index F, park visitor sentiment index E, and park scene diversity index H respectively; and constructs a comprehensive park vitality index (CVI) based on each index; based on the CVI and traffic network data, it constructs a multi-dimensional index system for urban parks; using a pre-trained XGBoost-SHAP machine learning model, it obtains the influence weights of each index on the CVI, as well as the threshold ranges for the number of bus stops, subway stations, and road network density in the traffic accessibility index; based on the CVI and its corresponding influence weights, and the number of bus stops, subway stations, and road network density and their corresponding threshold ranges, it generates traffic optimization strategies for different park vitality levels. This application improves the matching degree of resource demand.

Description

Technical Field

[0001] This application relates to the field of planning and design, and more specifically, to a traffic optimization method and system for enhancing park vitality. Background Technology

[0002] Urban parks are an important component of the urban open space system, and their service capacity is closely related to the surrounding traffic organization. The accessibility level of parks directly affects the spatiotemporal distribution of people entering the park, further influencing park usage intensity and service efficiency. With the improvement of methods for acquiring mobile communication data, online text data, landscape image data, and traffic network data, it has become possible to quantitatively analyze the operational status of urban parks using multi-source data. Therefore, how to integrate multi-source heterogeneous data with traffic network data to construct a park vitality analysis method that supports traffic optimization decisions has become an important research direction in related fields.

[0003] Existing urban park traffic optimization technologies typically employ empirical rules, single visitor flow indicators, or static spatial attributes to analyze park service levels. However, these methods, due to their limited data sources and analytical dimensions, struggle to accurately establish a quantitative correlation between park vitality characteristics and traffic accessibility factors. Furthermore, existing technologies often use linear analysis methods when analyzing the impact of traffic accessibility factors on park vitality, assuming a linear or monotonic relationship between indicators such as the number of bus stops, subway stations, and road network density and park vitality. However, different parks vary in size, facility configuration, surrounding land use structure, and road conditions. The impact of these traffic accessibility indicators on park vitality often exhibits non-linear characteristics and threshold effects. Using a uniform linear analysis logic or empirical threshold can easily lead to a mismatch between traffic facility configuration and actual demand.

[0004] Existing technologies, even when incorporating data such as mobile phone signaling, social media text, or landscape images, often rely on single-dimensional extraction or simple overlay, lacking a unified and comprehensive representation mechanism. This makes it difficult to generate stable evaluation results for park vitality. Furthermore, it struggles to further identify the weight and threshold boundaries of traffic accessibility indicators on park vitality, thus failing to provide reliable quantitative data for traffic optimization at different vitality levels. Based on these issues, there is an urgent need to propose a traffic optimization method and system for enhancing park vitality, generating traffic optimization strategies tailored to different park vitality levels. Summary of the Invention

[0005] To address the problem that existing technologies for urban park traffic optimization mainly rely on empirical rules or single-indicator analysis, making it difficult to accurately identify the nonlinear impact of traffic accessibility indicators on park vitality and their threshold boundaries, resulting in a lack of quantitative basis for optimization strategies, this application provides a traffic optimization method and system for enhancing park vitality. It can identify the impact weights and threshold ranges of traffic accessibility indicators such as the number of bus stops, the number of subway stations, and road network density, and generate traffic optimization strategies for different park vitality levels accordingly.

[0006] One aspect of this application provides a traffic optimization method for enhancing park vitality, comprising: Acquire multi-source data of urban parks in the target area. The multi-source data includes mobile phone signaling data, social text data, landscape image data, and traffic network data within the preset range of urban parks. Preprocess the acquired multi-source data. Based on the user spatiotemporal distribution data in the preprocessed mobile signaling data, calculate the park visitor density index F; Based on the preprocessed social text data, emotional features are extracted, and the park visitor sentiment index E is calculated. Based on the preprocessed landscape image data, scene features are extracted, and the park scene diversity index H is calculated. Based on the park visitor density index F, the park visitor emotion index E, and the park scene diversity index H, a comprehensive park vitality index CVI is constructed. Based on the Comprehensive Park Vitality Index (CVI) and traffic network data, a multi-dimensional indicator system for urban parks is constructed. The multi-dimensional indicator system includes: park physical characteristics indicators, surrounding environment characteristics indicators, and traffic accessibility indicators. Using each indicator in the multi-dimensional index coefficient as input, and employing the pre-trained XGBoost-SHAP machine learning model, we obtain the influence weight of each indicator on the park vitality comprehensive index CVI, as well as the threshold range of the number of bus stops, the number of subway stations, and the road network density in the traffic accessibility index. Based on the Park Vitality Index (CVI) and its corresponding influence weights, as well as the number of bus stops, subway stations, and road network density and their corresponding threshold ranges, traffic optimization strategies are generated for different park vitality levels. These strategies include adjusting bus routes, adding or removing stops, and upgrading the road network.

[0007] Another aspect of this application provides a traffic optimization system for enhancing park vitality, used to implement a traffic optimization method for enhancing park vitality according to this application.

[0008] Compared to existing technologies, the advantages of this application are: This application does not directly optimize based on a single traffic indicator, but rather constructs a comprehensive park vitality index (CVI) based on multi-source data such as mobile phone signaling, social text, and landscape images, transforming park vitality from a local judgment into a comprehensive representation. Since the CVI is formed by coupling multiple heterogeneous indicators, it exhibits a non-linear relationship with the traffic accessibility indicator, characterized by multiple factors working together, mutual constraints, and dynamically changing contributions. That is, the impact of the same traffic indicator on park vitality depends not only on its own value but also on factors such as park size, facility configuration, and surrounding environment. To address this complexity, this application uses the XGBoost model to learn the non-linear mapping relationship between multi-dimensional features and the CVI, and then uses the SHAP framework to decompose it into the marginal contributions of each indicator. This not only identifies the influence weights but also extracts the trends and threshold boundaries of the effects of bus stops, subway stations, and road network density across different value ranges, making the implicit traffic impact patterns explicit. Therefore, this application can identify the actual impact of different traffic elements on park vitality under different conditions and value ranges from a mechanistic perspective, avoiding the optimization inaccuracies caused by existing technologies based on empirical rules or linear assumptions, generating quantitatively based traffic optimization strategies for parks with different vitality levels, and improving the pertinence and reliability of optimization. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating a traffic optimization method for enhancing park vitality according to this application; Figure 2 This is a flowchart illustrating the multi-source data input and processing process of this application; Figure 3 This is a flowchart illustrating the comprehensive measurement of urban park vitality in this application; Figure 4 This is a framework diagram for analyzing the factors influencing the vitality of urban parks in this application; Figure 5 This is a graph showing the results of the analysis of the relative importance of factors influencing urban park vitality based on SHAP values ​​in this application; Figure 6 This is a schematic diagram illustrating the impact analysis of the park's land area in this application; Figure 7 This is a schematic diagram illustrating the impact analysis of surface relief in this application; Figure 8 This is a schematic diagram illustrating the impact of the water area ratio on the analysis of this application; Figure 9 This is a schematic diagram illustrating the impact analysis of vegetation cover rate in this application. Figure 10 This is a schematic diagram illustrating the impact analysis of green space type diversity in this application; Figure 11 This is a schematic diagram illustrating the impact analysis of pedestrian path density in this application; Figure 12This is a schematic diagram illustrating the impact analysis of the number of park parking lots in this application; Figure 13 This is a schematic diagram illustrating the impact analysis of the number of park restrooms in this application; Figure 14 This is a schematic diagram illustrating the impact analysis of the POI functional hybridity in this application; Figure 15 This is a schematic diagram illustrating the impact analysis of the distance to the city center matrix in this application; Figure 16 This is a schematic diagram illustrating the impact of the surrounding land use diversity analysis on this application; Figure 17 This is a schematic diagram illustrating the impact of surrounding population density on this application. Figure 18 This is a schematic diagram illustrating the impact of traffic accessibility on park vitality in this application. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings and tables. It should be understood that the following embodiments are only used to explain this application and do not constitute a limitation on the scope of protection of this application.

[0011] Example 1 like Figure 1 As shown in the figure, this application takes 160 urban parks in Nanjing as the research object and provides a method for assessing and optimizing urban park vitality based on multi-source data fusion. By acquiring multi-source data such as mobile phone signaling data, social media comment text data, landscape image data, park basic geographic data, POI data, land use data, population data, and transportation network data, the system calculates the footprint vitality index, emotional intensity index, and visual diversity index from three dimensions: visitor spatiotemporal activity, visitor emotional experience, and park visual environment perception, and constructs a comprehensive urban park vitality index. Then, it constructs a multi-dimensional index system by combining park ontological characteristics, surrounding environmental characteristics, and transportation accessibility indicators, and uses the XGBoost-SHAP machine learning model to analyze the nonlinear impact relationship of each index on urban park vitality and identify the threshold range of key parameters. Finally, based on the analysis results, it generates differentiated optimization strategies for parks with different vitality levels.

[0012] Acquire multi-source data on urban parks in the target area and preprocess the acquired multi-source data. For example... Figure 2As shown, multi-source data can include one or more of the following: mobile phone signaling data, social media comment text data, landscape image data, basic geographic data of urban parks, surrounding environment data, and transportation network data. Specifically, mobile phone signaling data is used to characterize users' spatiotemporal activities within the park; social media comment text data is used to characterize visitors' subjective evaluations and emotional experiences; landscape image data is used to characterize the park's visual environment and scene composition; basic geographic data of urban parks is used to characterize attributes such as park boundaries, topography, and internal facilities; surrounding environment data is used to characterize factors such as land use and population density around the park; and transportation network data is used to characterize transportation conditions such as bus stops, subway stations, road networks, and accessibility around the park.

[0013] Optionally, mobile signaling data can be sourced from the operator's big data platform, using the spatial location and time information of the user's base station to determine whether the user is located within the urban park area. Social media comment text data and landscape image data can be obtained from online comment platforms, using the park name as a search keyword to retrieve comment text and user-uploaded photos. Basic geographic data for urban parks may include park boundaries, entrance points, internal roads, and service facility POIs. Surrounding environment data may include land use data, population density data, and commercial service facility data. Transportation network data may include road networks, bus stops, subway stations, and isochronous zone data.

[0014] After acquiring multi-source data, it is preprocessed to eliminate problems such as noise, missing data, duplication, coordinate inconsistencies, and spatial scale inconsistencies. For example, comment data can be manually screened to remove irrelevant, duplicate, or commercial comments, obtaining valid comments and comment images.

[0015] Optionally, to calculate the accessibility of the area around the park, isochronous circles can be generated based on the park entrance. For example, first, the park entrance image is georeferenced and the entrance point coordinates are vectorized; then, the isochronous circle API interface is used to generate isochronous circles with the park entrance as the starting point and a 15-minute walking time threshold, which are used to count the number of bus stops, subway stations, and road network lengths within the reachable area of ​​the park, thereby more accurately reflecting the user's actual walking reach.

[0016] Based on the user spatiotemporal distribution data in the preprocessed mobile signaling data, the park visitor density index F is calculated. The park visitor density index F is used to characterize the actual usage intensity of the target city park within a preset time period.

[0017] Specifically, the location information and timestamp of the user's base station can be extracted from the preprocessed mobile signaling data. The location information of the user's base station can be spatially matched with the preset park geofence data to filter out the user's stay records located within the park geofence.

[0018] Optionally, after filtering out user dwell records, they can be grouped into three time categories—holidays, weekends, and weekdays—based on timestamps. For each time category, the number of visitors in each time period can be counted by hour or other preset time granularity to generate visitor count time series data.

[0019] Optionally, to further quantify park activity, the average visitor activity value can be calculated during the daytime period from 10:00 to 19:00. Specifically, the number of visitors in each time period within a preset time frame can be averaged to obtain the park visitor density index F. This park visitor density index F can be expressed as: ;in, This represents the number of visitors located within the park's geofence during the t-th time period, where t represents the number of time periods within the preset time period; and T represents the total number of time periods within the preset time period. Using this method, the actual usage intensity of urban parks can be quantified from an objective behavioral perspective based on mobile signaling data.

[0020] Based on the preprocessed social media comment text data, sentiment features are extracted, and the sentiment index E of park visitors is calculated. This index is used to characterize the statistical distribution of evaluative words, degree modifiers, negative words, punctuation marks, and semantic polarity features in the relevant comment texts of the target city park.

[0021] Specifically, the text semantic analysis tool VADER can be used to identify semantic polarity features in comment texts. VADER can map feature values ​​to evaluative words in the text using a pre-built lexical feature library, and then, combined with grammatical and syntactic rules, quantify the semantic polarity features and their strengths. For each comment text, VADER can identify evaluative words and assign semantic polarity feature values, adjusting these values ​​based on negation words, degree adverbs, and punctuation marks. The final output is a positive, negative, neutral semantic feature score, and a comprehensive semantic feature score. The comprehensive semantic feature score is a standardized value between -1 and 1.

[0022] Optionally, the average of the comprehensive semantic feature scores of all review texts corresponding to the target park can be calculated as the park visitor sentiment index E, denoted as: ;in, Let represent the comprehensive semantic feature score of the j-th comment text, and M represent the number of comment texts corresponding to the target park.

[0023] The above methods can convert unstructured text data in social media comments into structured sentiment feature indicators that can be used for model calculations, supplementing the footprint vitality indicator which only reflects spatial residence, thereby improving the data dimensionality completeness of multi-source data fusion evaluation results.

[0024] Based on the preprocessed landscape image data, scene features are extracted, and the park scene diversity index H is calculated. In some embodiments, the park scene diversity index H can also be called the visual diversity index, which is used to characterize the richness of the visual environment and scene composition of urban parks.

[0025] Specifically, a ResNet-18 deep learning model can be pre-trained for scene recognition on landscape image data. ResNet-18 can include convolutional layers, residual connections, and fully connected layers. Residual connections alleviate the vanishing gradient problem during deep neural network training, enabling the network to extract hierarchical features from images more effectively. Optionally, the ResNet-18 model can be pre-trained on scene recognition datasets such as Places365, giving it basic recognition capabilities for natural landscapes, urban landscapes, and public space scenes.

[0026] When transferring the model to a park landscape image recognition task, the output layer of the ResNet-18 model can be fine-tuned by setting the number of nodes in the output layer to [value missing]. One, corresponding to Park scene categories. For example, It can be set to 5, corresponding to 5 categories of park scenes. It should be noted that the number of park scene categories can be adjusted according to the actual research task and training samples; this application does not limit this.

[0027] Optionally, the output values ​​of the ResNet-18 model can be converted into a probability distribution using the Softmax activation function. For each input landscape image, the model can output the probability value of the image belonging to each park scene category, and select the category with the highest probability as the scene classification result for that image.

[0028] Optionally, during training, the cross-entropy loss function can be used to measure the difference between the model's predictions and the true labels, and the model parameters can be updated using the stochastic gradient descent (SGD) optimizer. To improve the model's generalization ability, data augmentation techniques such as image rotation, scaling, flipping, cropping, and color jittering can be used to increase the diversity of training samples and reduce the risk of model overfitting.

[0029] After completing image scene recognition, the scene classification results of all landscape images of the target park can be statistically analyzed, the frequency of occurrence of each scene category can be calculated, and the probability distribution of scene categories can be obtained: ;in, This represents the frequency of occurrence of the i-th scene category in the target park landscape image.

[0030] Then, based on the probability distribution of scene categories, the scene diversity value can be calculated using the Shannon entropy formula, and the calculated Shannon entropy value can be used as the park scene diversity index H. The Shannon entropy formula can be expressed as: Among them, when At that time, it can make .

[0031] The above methods can convert visual scene information from park landscape images uploaded by visitors into quantifiable diversity indicators. Compared to traditional manual interpretation methods, automated image recognition based on ResNet-18 can improve the efficiency of large-scale image processing, reduce human subjective bias, and make park visual environment perception an important component of park vitality evaluation.

[0032] A comprehensive park vitality index (CVI) is constructed based on the park visitor density index (F), park visitor emotional index (E), and park scene diversity index (H). For example... Figure 3 As shown, this application comprehensively measures the vitality of urban parks from three dimensions: footstep vitality, emotional experience, and visual perception. Since the park visitor density index F, park visitor emotional index E, and park scene diversity index H have different sources and dimensions, directly adding them may lead to one index having a large dimension or numerical range, which could unreasonably affect the overall evaluation result. Therefore, before constructing the comprehensive park vitality index CVI, F, E, and H can be normalized separately to obtain a normalized park visitor density index. Normalized park visitor sentiment indicators and normalized park scene diversity index .

[0033] Then, the Park Vitality Index (CVI) can be calculated based on the three normalized indicators: ;in, These are the weighting coefficients, and they satisfy... In one specific embodiment, it can be set to... Specifically, the footprint vitality index has a weight of 0.6, the emotional intensity index has a weight of 0.2, and the visual diversity index has a weight of 0.2. It should be noted that the above weights are merely examples; in other embodiments, weight coefficients can be determined based on expert scoring, entropy weighting, analytic hierarchy process (AHP), principal component analysis, or other weight determination methods.

[0034] In this way, this application integrates the objective usage intensity reflected by mobile phone signaling data, the subjective emotional experience reflected by social media comment text, and the visual scene diversity reflected by landscape image data into a unified park vitality comprehensive index (CVI), which expands the park vitality evaluation from a single dimension to a multi-dimensional comprehensive representation, thereby improving the comprehensiveness and reliability of the vitality evaluation results.

[0035] Based on the Comprehensive Park Vitality Index (CVI) and multi-source spatial data, a multi-dimensional indicator system for urban parks is constructed. After obtaining the CVI, a further multi-dimensional indicator system can be built to explain the mechanism of park vitality formation. For example... Figure 4 As shown, this multi-dimensional indicator system can include indicators of park characteristics, indicators of the surrounding environment, and indicators of public transportation accessibility. Specifically, as shown in Table 1, the indicator system for influencing factors of urban park vitality constructed in this application includes 3 primary indicators, 9 secondary indicators, and 29 specific indicators.

[0036] Table 1. Indicator System, Sources, and Descriptions of Factors Influencing Urban Park Vitality

[0037] Park characteristics indicators are used to describe the park's spatial form, landscape pattern, internal walking conditions, and service facility configuration. They can include secondary indicators such as basic attributes, landscape pattern, walkability, and infrastructure construction.

[0038] The basic attributes may include the operating years P1 and the park's land area P2. The operating years P1 is calculated based on the time span from the park's construction and opening to the current statistical period, in years; the park's land area P2 is calculated based on the area of ​​the non-water area within the park's boundaries, in hectares.

[0039] Landscape pattern can include surface relief (P3), landscape shape index (P4), water area ratio (P5), vegetation coverage (P6), and green space diversity (P7). Surface relief (P3) is calculated by extracting the elevation difference between the highest and lowest points within the park from DEM data; landscape shape index (P4) is calculated based on the deviation of the park's outer contour shape from a regular shape of the same area; water area ratio (P5) and vegetation coverage (P6) are calculated based on the ratio of the corresponding area to the total park area, respectively; and green space diversity (P7) is calculated using the Shannon entropy index based on land use classification results.

[0040] Walkability can include pedestrian path density P8, which is calculated as the ratio of the total length of walkable trails within the park to the park's land area.

[0041] Infrastructure construction may include the number of entrances / exits (P9), parking lots (P10), restrooms (P11), catering facilities (P12), shopping facilities (P13), entertainment facilities (P14), sports facilities (P15), cultural facilities (P16), and the POI functional mix (P17). The POI functional mix (P17) is calculated using the Shannon entropy index based on the distribution of different types of service facilities within the park.

[0042] The environmental characteristics indicators surrounding a park are used to describe the park's location, land use structure, and population base within its service area. These indicators may include secondary indicators such as park location, land use, and population density.

[0043] Park location can include distance from the city center E1, such as the average straight-line distance between all park entrances and the city center. Land use can include the proportion of commercial and service land E2, residential land E3, science, education, culture and health land E4, higher education land E5, transportation land E6, park green space land E7, and land use diversity E8, calculated within a 15-minute walking radius of the park. Population density can include the population density around the park E9.

[0044] Public transport accessibility indicators are used to describe the level of public transport service and road network connectivity around the park. These indicators may include the number of bus stops (A1), the number of subway stations (A2), and the road network density around the park (A3). The number of bus stops (A1) and subway stations (A2) are the sum of the number of stops within a 15-minute walking radius based on the road network, taking each park entrance as a starting point. The road network density around the park (A3) is the ratio of the total length of the road network within the isochronous circle to the area of ​​the isochronous circle.

[0045] Furthermore, the aforementioned park-specific characteristic indicators, surrounding environment characteristic indicators, and traffic accessibility indicators can be quantified to construct a multi-dimensional indicator matrix X. Each row in the multi-dimensional indicator matrix X corresponds to a sample of urban parks, and each column corresponds to an indicator variable. For example, for a dataset containing M urban parks and n indicator variables, the multi-dimensional indicator matrix can be represented as: ;in, This represents the quantified value of the m-th park sample on the n-th indicator.

[0046] After constructing the multi-dimensional indicator matrix X, the Park Vitality Comprehensive Index (CVI) can be aligned with the multi-dimensional indicator matrix X according to park identifiers to obtain a multi-dimensional indicator system for urban parks. Through data alignment, it can be ensured that the CVI of each park sample corresponds one-to-one with its park characteristics, surrounding environment characteristics, and transportation accessibility characteristics, avoiding sample mismatch caused by different data sources.

[0047] In this way, the application can unify the park vitality evaluation results with the park's own conditions, surrounding environment and traffic accessibility factors into the same analytical framework, providing structured input for subsequent nonlinear modeling and influence mechanism identification.

[0048] Table 2 presents the descriptive statistical results of the factors influencing the vitality of urban parks. As shown in Table 2, there are significant differences among different parks for various indicators, indicating that urban parks in Nanjing have strong heterogeneity in terms of scale, facilities, landscape pattern, location conditions, and transportation accessibility.

[0049] Table 2. Descriptive Statistics of Factors Influencing the Vitality of Urban Parks

[0050] Regarding the characteristics of the parks themselves, the minimum value of the park's operating years P1 is 0 years, the maximum value is 97 years, and the average value is 11.59 years; the minimum value of the park's land area P2 is 0.79 hectares, the maximum value is 2350.28 hectares, the average value is 91.82 hectares, and the standard deviation is 315.20, indicating that there are significant differences in size between different parks.

[0051] In terms of landscape pattern, the surface relief P3 ranges from 4 meters to 301 meters, with an average of 32.34 meters; the water area ratio P5 has a maximum value of 0.72 and an average value of 0.12; the vegetation coverage P6 has an average value of 0.69; and the green space type diversity P7 has a maximum value of 1.48 and an average value of 0.45.

[0052] In terms of infrastructure construction, the maximum number of park restrooms (P11) is 33, with an average of 1.88; the maximum number of catering services (P12) is 47, with an average of 1.49; the maximum number of entertainment services (P14) is 17, with an average of 1; and the maximum POI functional mix (P17) is 1.89, with an average of 0.76, indicating that most parks have relatively limited internal service facilities.

[0053] Regarding the environmental characteristics surrounding the park, the minimum distance from the city center (E1) is 1136.77 meters, the maximum is 81020.87 meters, the average is 18961.63 meters, and the standard deviation is 18778.74; the average proportion of surrounding residential land (E3) is 0.24; and the average land use diversity (E8) is 1.24, with a standard deviation of 0.30.

[0054] Regarding public transportation accessibility, the minimum number of surrounding bus stops (A1) is 0, the maximum is 50, the average is 8.89, and the standard deviation is 10.35; the minimum number of surrounding subway stations (A2) is 0, the maximum is 6, and the average is 0.68; the minimum surrounding road network density (A3) is 2.60 meters / hectare, the maximum is 590.49 meters / hectare, and the average is 309.65 meters / hectare. There are significant differences in public transportation service levels and road network connectivity among different parks.

[0055] Descriptive statistics revealed the spatial heterogeneity among the sample parks, providing a data foundation for subsequent machine learning models to identify nonlinear influence relationships and threshold effects.

[0056] Based on the XGBoost-SHAP machine learning model, the influence weights of each indicator on the Comprehensive Vitality Index (CVI) of the park are obtained, and the threshold ranges of key parameters are identified.

[0057] Specifically, each row of the multi-dimensional indicator matrix X can be used as a feature vector of a sample, and the corresponding Park Vitality Index (CVI) can be used as a label value to form a training dataset. Based on the training dataset, the XGBoost gradient boosting algorithm is used to build a park vitality prediction model. The XGBoost model learns the complex nonlinear mapping relationship between the multi-dimensional indicators and the Park Vitality Index (CVI) by integrating multiple weak decision trees.

[0058] Optionally, when training the XGBoost model, the model's hyperparameters can be optimized through grid search and K-fold cross-validation to improve the model's generalization ability.

[0059] Alternatively, a coefficient of determination can be used. The mean squared error (MSE) and root mean square error (RMSE) are used to evaluate the performance of the XGBoost prediction model. The model is used to characterize the ability of the model to explain changes in CVI. MSE is used to characterize the mean squared error between the predicted and actual values, and RMSE is used to characterize the model's prediction error on the same scale as the target variable. In a specific embodiment, the XGBoost model on the training set... The value can be 0.80, MSE can be 19.58, and RMSE can be 4.42; on the validation set... The coefficient of performance (COP) can be 0.73, the mean squared error (MSE) 37.77, and the root mean squared error (RMSE) 6.15. The above model evaluation results demonstrate that the model can fit the relationship between the multidimensional indicators and the overall vitality of urban parks relatively well. It should be noted that these evaluation results are merely illustrative and do not limit the scope of protection of this application.

[0060] Because the XGBoost model has strong nonlinear expressive power, but its internal decision-making process is relatively complex, this application further employs the SHAP interpretation framework to interpret and analyze the trained XGBoost prediction model. The SHAP framework, based on Shapley value theory in cooperative game theory, decomposes the model's prediction result for each sample into the sum of the marginal contributions of each feature, thus obtaining the contribution value of each feature to the prediction result of the corresponding sample.

[0061] Specifically, for the trained XGBoost prediction model, the SHAP interpretation framework can be used to calculate the SHAP value for each sample on each feature dimension. Features can include all secondary indicators from park ontology features, surrounding environment features, and accessibility indicators. Then, the SHAP values ​​of all samples can be arranged according to the feature dimensions to form an m×n SHAP matrix. ;in, This represents the SHAP contribution of the nth feature in the mth park sample to the model's prediction of CVI.

[0062] Furthermore, the average absolute value of each column of the SHAP matrix can be calculated to obtain the average absolute SHAP value of each feature, which can then be used as the weight of the corresponding indicator's influence on the Comprehensive Park Vitality Index (CVI). The average absolute SHAP value can be expressed as: ;in, The value represents the weight of the j-th indicator on the CVI; m represents the total number of samples, i.e., the number of parks participating in the analysis; i and j represent the serial numbers, respectively. This represents the contribution of the j-th indicator to the predicted value of the Park Vitality Index (CVI) for the i-th park sample. By influencing the weights, we can obtain the global importance ranking of each indicator to the Comprehensive Park Vitality Index (CVI). For example... Figure 5 As shown, Figure 5 The results of the relative importance analysis of factors influencing urban park vitality based on SHAP values ​​are presented. The vertical axis represents the influencing factors, and the horizontal axis represents the SHAP values. Color indicates the magnitude of the eigenvalue, with red indicating a higher eigenvalue and blue indicating a lower eigenvalue. The influencing factors are arranged from highest to lowest importance, and each data point corresponds to the SHAP value of a single park sample.

[0063] Depend on Figure 5 It can be seen that factors such as POI functional mix, park land area, number of park restrooms, surface relief, green space type diversity, number of surrounding bus stops, pedestrian path density, number of park parking lots, and distance from the city center are among the most important, indicating that these indicators have a strong explanatory power for the vitality of urban parks.

[0064] Among these factors, the high level of POI (Point of Interest) functional mix indicates that a more diverse range of facilities and functions within a park are more conducive to attracting different types of visitors, thereby enhancing the park's overall vitality. The high ranking of park land area suggests that park size plays a crucial role in the types of activities it can accommodate, the number of visitors it can hold, and the creation of diverse scenarios. Furthermore, high-value samples regarding distance from the city center are mostly concentrated in areas with negative SHAP (Shadows and Points of Interest) values, indicating that parks farther from the city center generally have lower overall vitality; while high-value samples regarding the number of surrounding bus stops and pedestrian path density mostly show positive contributions.

[0065] Overall, the characteristics of the park itself contribute significantly to the explanation of the vitality of urban parks, followed by the characteristics of the surrounding environment and public transportation accessibility.

[0066] In some embodiments, SHAP curves can be fitted to key parameters to identify threshold ranges for those parameters. Key parameters may include the number of bus stops, subway stations, and road network density in transport accessibility indicators, as well as indicators that have a significant impact on CVI, such as park size, POI functional mix, population density, and distance from the city center.

[0067] Specifically, for a target feature, the values ​​of that target feature across all samples can be used as the horizontal axis, and the corresponding SHAP value can be used as the vertical axis to fit a SHAP dependency curve. Then, the first and second derivatives of the SHAP curve are calculated. The first derivative reflects the marginal direction and rate of change of the contribution of the target feature value to CVI, while the second derivative reflects the acceleration or deceleration trend of this marginal change. The threshold range corresponding to the target feature can be determined based on the zero point, sign change point, and local extrema of the first derivative, as well as the inflection point of the second derivative.

[0068] For example, regarding the number of bus stops A1, when the SHAP curve shows an upward trend in the low-value range, it indicates that increasing bus stops may have a positive effect on CVI; when the curve enters a flat range, it indicates that the marginal positive effect of increasing bus stops weakens; when the curve declines or the contribution turns negative, it indicates that excessively high stop density may have a negative impact. In this case, the effective increase range, marginal decrease range, or over-provision range of the number of bus stops can be determined based on the first and second derivatives of the SHAP curve. Similarly, the threshold ranges for the number of subway stations A2 and the road network density around parks A3 can be identified.

[0069] This application further analyzes the nonlinear impact of key park ontological features on the vitality of urban parks based on SHAP dependency graph analysis. For example... Figures 6 to 14 As shown in the figure, the horizontal axis of each graph represents the value of the corresponding indicator, the vertical axis represents the SHAP value, the green scatter points represent the sample parks, the orange curve represents the non-linear effect trend after fitting, and the red dashed line represents the key threshold identified.

[0070] like Figure 6As shown, the land area of ​​a park has a significant non-linear positive impact on the vitality of urban parks. When the park's land area is less than approximately 50 hectares, the SHAP value increases rapidly with increasing area; after exceeding approximately 50 hectares, the curve gradually flattens out. Table 2 shows that the average land area of ​​the sample parks is 91.82 hectares, with a maximum of 2350.28 hectares. Therefore, for small and medium-sized parks with an area of ​​less than 50 hectares, vitality can be enhanced by expanding usable space; for large parks exceeding 50 hectares, the focus should be on improving space utilization efficiency through functional zoning and facility layout.

[0071] like Figure 7 As shown, the relationship between surface relief and urban park vitality shows a positive increasing effect that gradually diminishes. When the surface relief is less than approximately 50 meters, the SHAP value increases significantly with increasing terrain undulation; however, the curve flattens out after approximately 50 meters. Table 2 shows that the average surface relief of the sample parks is 32.34 meters, with a maximum value of 301 meters.

[0072] like Figure 8 As shown, the proportion of water area exhibits a significant threshold effect on park vitality. When the proportion of water area is below approximately 0.3, the overall SHAP value is low or negative; when it approaches and exceeds 0.3, the SHAP value rises rapidly and turns positive. Combined with Table 2, it can be seen that the average proportion of water area in the sample parks is only 0.12, indicating that most parks have insufficient water resources.

[0073] like Figure 9 As shown, a non-linear negative threshold effect exists between vegetation cover and park vitality. When the vegetation cover is below approximately 0.9, the SHAP value remains largely in the positive range; however, above approximately 0.9, the SHAP value drops rapidly and turns negative. Table 2 shows that the average vegetation cover of the sample parks is 0.69.

[0074] like Figure 10 As shown, green space type diversity has a significant positive effect on park vitality. When green space type diversity is below approximately 0.3, the SHAP value is mostly negative; it turns positive and gradually increases after exceeding approximately 0.3. Combined with Table 2, the average green space type diversity of the sample parks is 0.45, and the maximum value is 1.48.

[0075] like Figure 11 As shown, walking path density has a significant threshold effect on park vitality. When the walking path density is below approximately 250 meters per hectare, the SHAP value is mostly in the negative region; after reaching approximately 250 meters per hectare, the SHAP value rises rapidly and tends to stabilize. Combined with Table 2, the average walking path density of the sample parks is 270.53 meters per hectare.

[0076] like Figure 12As shown, the number of parking lots in a park has a rapid increase in activity followed by a plateauing effect. When the number of parking lots is less than approximately 3 to 5, the SHAP value increases significantly with the increase in the number of parking lots; after reaching approximately 3 to 5, the curve tends to stabilize. Table 2 shows that the average number of parking lots in the sample parks is 1.96.

[0077] like Figure 13 As shown, the number of park restrooms has a significant positive threshold effect on activity. When the number of restrooms is less than about 3, the SHAP value is mostly negative or close to zero; after more than about 3, the SHAP value rises rapidly and gradually plateaus. Table 2 shows that the average number of restrooms in the sample parks is 1.88, which is below this threshold.

[0078] like Figure 14 As shown, the functional mix of POIs has a significant positive effect on the vitality of urban parks. When the functional mix of POIs is below approximately 1.1, the SHAP value is mostly negative or at a low level; above approximately 1.1, the SHAP value quickly turns positive and tends to stabilize. Table 2 shows that the average functional mix of POIs in the sample parks is 0.76, which is below this threshold.

[0079] like Figure 15 As shown, distance from the city center has a significant negative impact on the vitality of urban parks. When the distance from the city center is less than approximately 20,000 meters, the SHAP value is mostly positive or close to positive; however, after exceeding approximately 20,000 meters, the SHAP value decreases significantly and turns negative. Combined with Table 2, the average distance of the sample parks from the city center is 18,961.63 meters.

[0080] like Figure 16 As shown, the surrounding land use diversity has a significant positive threshold effect on park vitality. When the surrounding land use diversity is below approximately 1.28, the SHAP value is mostly negative or at a low level; above approximately 1.28, the SHAP value rises rapidly and then tends to stabilize. Table 2 shows that the average surrounding land use diversity is 1.24.

[0081] like Figure 17 As shown in Table 2, the surrounding population density has a weak positive impact on the overall vitality of the park. The average surrounding population density is 21.65 people / hectare, with a maximum of 81.94 people / hectare.

[0082] like Figure 18As shown, the number of surrounding bus stops has a significant positive impact on the vitality of urban parks. When the number of surrounding bus stops is less than about 7, the SHAP value is mostly negative; after reaching about 7, the SHAP value turns positive and continues to rise; after further increases, the curve gradually flattens out. Combined with Table 2, the average number of bus stops around the sample parks is 8.89, with a standard deviation of 10.35, indicating significant differences in public transport accessibility among different parks.

[0083] Through the above methods, this application can not only predict the level of park vitality, but also explain the strength and direction of the impact of different indicators on CVI, and further identify the effective range of key traffic parameters and other built environment parameters.

[0084] Based on the analysis results, differentiated optimization strategies are generated for parks of different types and vitality levels. After obtaining the Park Vitality Index (CVI), the weights of each indicator on the CVI, and the threshold ranges of key parameters, differentiated optimization strategies for urban parks of different types and vitality levels can be generated based on the analysis results.

[0085] Optionally, parks can be categorized into three levels—low vitality, medium vitality, and high vitality—based on the Comprehensive Park Vitality Index (CVI). Optimization priorities can be determined according to the influence weights of each indicator, and corresponding optimization strategies can be developed for parks at different vitality levels.

[0086] based on Figures 5 to 18 The SHAP importance ranking and key threshold identification results shown can divide the park comprehensive vitality index (CVI) into high vitality, medium vitality and low vitality levels. By comparing the actual values ​​of each park in the index system in Table 1 with the identified thresholds, the optimization direction can be determined.

[0087] For low-activity parks, if their POI (Point of Interest) functional mix is ​​less than 1.1, priority should be given to adding mixed service facilities; if there are fewer than 3 restrooms, priority should be given to supplementing basic service facilities; if the pedestrian path density is less than 250 meters / hectare, priority should be given to improving the slow-traffic network; if there are fewer than 7 bus stops nearby, priority should be given to optimizing bus connections and stop layout; if the distance from the city center is more than 20,000 meters, the park's destination function should be strengthened.

[0088] For traffic optimization, the actual values ​​of the number of bus stops, subway stations, and road network density for each park can be obtained, and these values ​​can be compared with corresponding threshold ranges. For example, when the number of bus stops in a low-activity park is below the lower limit of the effective improvement range and has a high impact weight, strategies to increase bus stops or optimize bus route coverage can be generated; when the road network density is below the threshold range, strategies to increase pedestrian and cycling road connections or improve the connectivity of roads around the park entrance can be generated.

[0089] For medium-sized parks with low activity levels, targeted improvements can be made based on their weakest indicators. For example, if the park area exceeds 50 hectares but the functional mix of POIs is insufficient, the focus should be on improving the functional mix; if the vegetation coverage exceeds 0.9, open activity spaces should be appropriately increased; if the water area ratio is less than 0.3, the landscape attractiveness can be enhanced through ecological water features or waterfront spaces.

[0090] For high-activity parks, the focus should be on carrying capacity pressure and diminishing marginal returns. When the number of bus stops or road network density around a high-activity park exceeds the upper limit of the threshold range, and the SHAP curve shows that the contribution of continuing to increase this indicator tends to level off or turn negative, strategies can be generated to optimize the station layout, disperse passenger flow entrances, or improve pedestrian diversion paths, rather than simply increasing the number of stations.

[0091] In summary, as shown in Table 1, this application achieves a systematic expression of the factors influencing the vitality of urban parks. As shown in Table 2, the descriptive statistical results demonstrate that this application can identify significant differences between different parks. Figure 5 The importance ranking of SHAP shown is as follows: Figures 6 to 18 As shown in the SHAP dependency diagram, this application can identify the nonlinear effects and threshold ranges of key indicators. For example, the key threshold for park land area is approximately 50 hectares; the key threshold for water area ratio is approximately 0.3; the negative threshold for vegetation coverage is approximately 0.9; the key threshold for pedestrian path density is approximately 250 meters / hectare; the key threshold for POI functional mixing is approximately 1.1; and the key threshold for the number of surrounding bus stops is approximately 7. Therefore, this application can provide quantitative evidence for urban park renovation and upgrading, facility configuration, traffic optimization, and synergy with surrounding functions, improving the scientific nature of urban park planning, management, and optimization.

[0092] The foregoing illustrative description of the present application and its embodiments is not restrictive and can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. The accompanying drawings are only one embodiment of the present application, and the actual structure is not limited thereto. Therefore, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the present application, such designs should fall within the scope of protection of this application. Furthermore, the word "comprising" does not exclude other elements or steps, and the word "a" preceding an element does not exclude the inclusion of "a plurality" of that element. Terms such as "first," "second," etc., are used to indicate names and do not indicate any specific order.

Claims

1. A traffic optimization method for enhancing park vitality, characterized in that, include: Acquire multi-source data of urban parks in the target area. The multi-source data includes mobile phone signaling data, social text data, landscape image data, and traffic network data within the preset range of urban parks. Preprocess the acquired multi-source data. Based on the user spatiotemporal distribution data in the preprocessed mobile signaling data, calculate the park visitor density index F; Based on the preprocessed social text data, emotional features are extracted, and the park visitor sentiment index E is calculated. Based on the preprocessed landscape image data, scene features are extracted, and the park scene diversity index H is calculated. Based on the park visitor density index F, the park visitor emotion index E, and the park scene diversity index H, a comprehensive park vitality index CVI is constructed. Based on the Comprehensive Park Vitality Index (CVI) and traffic network data, a multi-dimensional indicator system for urban parks is constructed. The multi-dimensional indicator system includes: park physical characteristics indicators, surrounding environment characteristics indicators, and traffic accessibility indicators. Using each indicator in the multi-dimensional index coefficient as input, the pre-trained XGBoost-SHAP machine learning model is used to obtain the influence weight of each indicator on the park vitality comprehensive index CVI, as well as the threshold range of the number of bus stops, the number of subway stations and the road network density in the traffic accessibility index. Based on the Park Vitality Index (CVI) and its corresponding influence weights, as well as the number of bus stops, subway stations, and road network density and their corresponding threshold ranges, traffic optimization strategies are generated for different park vitality levels. These strategies include adjusting bus routes, adding or removing stops, and upgrading the road network.

2. The traffic optimization method for enhancing park vitality according to claim 1, characterized in that: The park's intrinsic characteristics include: land area, years of operation, surface relief, landscape shape index, water area ratio, vegetation coverage, green space type diversity, pedestrian path density, number of entrances and exits, number of parking facilities, number of restrooms, number of catering service facilities, number of shopping service facilities, number of entertainment service facilities, number of sports service facilities, number of cultural service facilities, and POI functional mixing degree. The surrounding environmental characteristics indicators include: distance of the park from the city center, proportion of commercial and service land, proportion of residential land, proportion of science, education, culture and health land, proportion of higher education land, proportion of transportation land, proportion of park green space, land use diversity and population density; Traffic accessibility indicators include: the number of bus stops, the number of subway stations, and road network density.

3. The traffic optimization method for enhancing park vitality according to claim 2, characterized in that: Based on the user spatiotemporal distribution data in the preprocessed mobile signaling data, the park visitor density index F is calculated, including: Extract the location information and timestamp of the user's camped base station from the preprocessed mobile signaling data; Spatially match the location information of the user's base station with the preset park geofence data to filter out the user's stay records located within the park geofence. Based on the timestamp, the filtered user stay records are grouped into three time categories: holidays, weekends, and weekdays; For each time category, the number of visitors in each time period is counted, and time-series visitor data is generated; Based on the time-series data of visitor numbers, the average number of visitors within a preset time period is calculated and used as the park visitor density index F.

4. The traffic optimization method for enhancing park vitality according to claim 2, characterized in that: The calculation of park visitor sentiment index E includes: Using the sentiment analysis tool VADER, the sentiment polarity of preprocessed social text data was identified, including positive, negative, and neutral sentiment polarities. The sentiment analysis tool VADER maps the sentiment values ​​of words in a text to a pre-built sentiment vocabulary and adjusts the sentiment values ​​based on preset negative words, degree adverbs and punctuation marks in the text. Based on the weighted sentiment values, output the positive sentiment score, negative sentiment score, and neutral sentiment score for each social text data. Based on the positive sentiment score, negative sentiment score, and neutral sentiment score, a comprehensive sentiment score is obtained through a normalization algorithm. Calculate the average of the combined sentiment scores of all social text data, and use it as the sentiment index E for park visitors.

5. The traffic optimization method for enhancing park vitality according to claim 2, characterized in that: The calculation of the park scene diversity index H includes: A pre-trained ResNet-18 deep learning model is used for scene recognition on landscape image data. The ResNet-18 deep learning model includes convolutional layers and fully connected layers. The number of output layer nodes of the ResNet-18 deep learning model is set to N2, corresponding to N2 park scene categories, and the output value is converted into a probability distribution using the Softmax activation function; The preprocessed landscape image data is used as input, and the pre-trained ResNet-18 deep learning model is used to obtain the probability value of each image belonging to each park scene category. The category with the highest probability is selected as the scene classification result of the corresponding image. The scene classification results of all landscape images in the park were statistically analyzed, the frequency of each scene category was calculated, and the probability distribution of the scene categories was obtained. ; Based on the probability distribution P of scene categories, the scene diversity value is calculated using the Shannon entropy formula, and the calculated Shannon entropy value is used as the park scene diversity index H.

6. The traffic optimization method for enhancing park vitality according to any one of claims 2 to 5, characterized in that: The Comprehensive Vitality Index (CVI) for parks is constructed, including: The park visitor density index F, park visitor emotional index E, and park scene diversity index H were normalized respectively to obtain the normalized park visitor density index. Park visitor sentiment indicators and park scene diversity indicators ; Based on the normalized park visitor density index Park visitor sentiment indicators and park scene diversity indicators Calculate the Park Vitality Index (CVI): ,in, , , are the weighting coefficients, respectively.

7. The traffic optimization method for enhancing park vitality according to claim 6, characterized in that: Construct a multi-dimensional indicator system for urban parks, including: Data on bus stops, subway stations, and road networks within the pre-defined area of ​​the park were extracted from the pre-processed traffic network data. Bus stop data, subway station data, and road network data will be used as indicators of transportation accessibility. Based on the park's intrinsic characteristics, surrounding environment characteristics, and accessibility indicators, a multi-dimensional indicator matrix X is constructed. By aligning the Park Vitality Comprehensive Index (CVI) and the multi-dimensional indicator matrix X, a multi-dimensional indicator system for urban parks is obtained.

8. The traffic optimization method for enhancing park vitality according to claim 7, characterized in that: Obtain the weights of each indicator on the Comprehensive Park Vitality Index (CVI), and the threshold ranges for the number of bus stops, subway stations, and road network density in the accessibility indicators, including: Each row of the multi-dimensional index matrix X is used as a feature vector of a sample, and the corresponding park vitality index CVI is used as the label value of the corresponding sample. The two are combined to form a training dataset. Based on the training dataset, a park vitality prediction model was established using the XGBoost gradient boosting algorithm. The hyperparameters of the prediction model were optimized through grid search and 5-fold cross-validation. The hyperparameters included the maximum tree depth, learning rate, and regularization parameter. The performance of the prediction model was evaluated using the coefficient of determination R², mean squared error MSE, and root mean squared error RMSE. For the trained XGBoost prediction model, the SHAP interpretation framework is used to calculate the SHAP value of each sample on each feature dimension. The features include all secondary indicators of park ontology features, surrounding environment features, and traffic accessibility. Arrange the SHAP values ​​of all samples according to the feature dimension to form The SHAP matrix, where m is the number of samples and n is the number of features; Calculate the average absolute value of each column of the SHAP matrix to obtain the average absolute SHAP value of each feature, which is used as the weight of each indicator on the Comprehensive Park Vitality Index (CVI). For the number of bus stops, the number of subway stations, and the road network density in the traffic accessibility indicators, corresponding SHAP curves are fitted according to the corresponding SHAP values; the SHAP curves represent the nonlinear relationship between the eigenvalues ​​and the contribution of the corresponding eigenvalues ​​to the Comprehensive Park Vitality Index (CVI). Calculate the first and second derivatives of the SHAP curve, and determine the threshold ranges corresponding to the number of bus stops, the number of subway stations, and the road network density in the traffic accessibility index based on the first and second derivatives.

9. The traffic optimization method for enhancing park vitality according to claim 8, characterized in that: Generate traffic optimization strategies for different park activity levels, including: Based on the Comprehensive Park Vitality Index (CVI), parks are divided into three levels: low vitality, medium vitality, and high vitality. Based on the influence weights corresponding to the Park Vitality Index (CVI), the priority of traffic optimization is determined; Obtain the actual values ​​of the number of bus stops, subway stations, and road network density for each park; Compare each actual value with its corresponding threshold range; Based on the comparison results and the determined priorities, traffic optimization strategies with different vitality levels are generated.

10. A traffic optimization system for enhancing park vitality, used to perform the method described in any one of claims 1 to 9.

Images