Nonlinear impacts of the configuration elements of life services on spatial vitality in the online-merge-offline context: A case study of Shanghai,China

Abstract

Despite the growing prevalence of online-merge-offline life services and increased spatiotemporal flexibility in human activities, few studies examine how social media platforms and spatial layout elements jointly influence spatial vitality, hindering sustainable development in the digital era. To address this gap, this study employs Light Gradient Boosting Machine (LightGBM) interpreted by SHapley Additive exPlanations (SHAP) to analyze the nonlinear impacts of Internet word-of-mouth (IWOM) and spatial layout elements of life services on spatial vitality in Shanghai. Key findings reveal that: (1) IWOM drives a shift in life service access from location determinism to a digital visibility-driven mode through spatial activation and interaction, thereby redistributing spatial vitality; (2) The impact of IWOM varies significantly by service type: retail commerce and convenience services exhibit the strongest enhancement effects, with both showing a minimum threshold in activating vitality; (3) Significant interactive effects exist, particularly between plot ratio and IWOM of multiple life services, reinforcing the trend of online service penetration into vertical spaces. These findings advance the theoretical understanding of how information and communication technologies (ICT) enhance spatial vitality and provide a practical basis for guiding the digital transformation of life service configuration.

Keywords

life service spatial vitality Internet word-of-mouth online-merge-offline nonlinear impact interaction effect

Introduction

With rising living standards, social interactions outside home and workplace have become the primary source of spatial vitality (Oldenburg and Brissett, 1982). As a core component of sustainable development, spatial vitality manifests in multiple dimensions, including the accumulation of social capital at the economic level and the enhancement of life quality and community cohesion at the human-oriented level (Park and Kim, 2025; Williams and Pocock, 2010). The emergence of Online to Offline (O2O) behavior, in which social media platforms influence travel decisions (Wan et al., 2023; Wu et al., 2024a; Xiao, 2022), has further transformed social interaction, imbuing it with characteristics of spatial fragmentation and dynamic continuity (Li et al., 2023; Xiao, 2022). This shift leads to the rapid growth of global O2O market, which reached 178 billion USD in 2023 with a compound annual growth rate of 12.3% (DataIntelo, 2023). Specifically, in the world’s largest digital consumer market, 59% of users in China conduct O2O consumption at least three times weekly (Iimedia Research, 2023). Thus, quantifying spatial vitality in the digital era and identifying its drivers have become research priorities for fostering sustainable urban development within fields such as urban-rural planning, transportation, and consumption studies.

In travel behavior theory, social interaction outside home and workplace refers to the process of accessing life services, encompassing maintenance activities (e.g., shopping, dining, healthcare) and leisure activities (e.g., cultural experiences, entertainment) (He et al., 2024; Wei et al., 2023; Gong et al., 2020). Among them, the configuration of life services is shaped by physical spatial layout and social media platforms (Liao et al., 2024), with the former determining five dimensions of service opportunities: density, diversity, design, transit proximity, and accessibility (Li et al., 2022; Wang et al., 2024; Xia et al., 2020). Meanwhile, social media platforms provide information that influences when, where, and how individuals access these opportunities (Park and Kim, 2024, 2025). Specifically, the information can be categorized into two types: (1) Visual-textual review, which conveys individual experiences and subjective feedback (Chen and Yeh, 2021; Shao et al., 2023); (2) Rating score, often termed digital word-of-mouth (IWOM), derived from public perceptions and algorithmically weighted to represent service quality (Zhang and Long, 2024). Owing to its relative objectivity and trend representativeness, IWOM enables efficient preliminary screening of service providers, making it a key factor in shaping individual decisions and reshaping urban spatial patterns.

Traditionally, spatial vitality has been measured by the intensity of maintenance and leisure activities, which are closely associated with the spatial layout of life services (Jacobs, 1961). This association arises because the location of service opportunities is shaped by urban functions, such as land use and facility density, while their quantity is influenced by building forms, including floor area ratios and architectural design (Pyrialakou et al., 2016; Wang et al., 2025). In addition, individuals’ willingness to access services is affected by their perception of travel convenience, which is largely determined by network connectivity. Accordingly, urban planning practices generally concentrate services in central areas to enhance recognizability and reduce travel costs, linking spatial vitality to the distribution of objective opportunities. However, the emergence of social media platforms has altered this logic by enhancing users’ awareness of available choice sets through location-based recommendations (Wang et al., 2025) and perceived service attractiveness through virtual intentions (Zhang and Long, 2024). This shift promotes decentralized urban activity patterns (Dadashpoor and Yousefi, 2018), boosting activity intensity in peripheral or low-accessibility areas through improved digital accessibility, and in turn regenerating spatial vitality (Zhang et al., 2025).

However, existing research on the relationship between life service configuration and spatial vitality in virtual-physical contexts reveals two notable gaps: (1) while O2O behaviors increasingly exhibit spatially fragmented and dynamically connected forms, most spatial vitality studies still emphasize static agglomeration (Ge et al., 2024), overlooking the role of dynamic interaction reflected in human flows; (2) although the influence of IWOM on activity distribution has been examined (Lin et al., 2023; Zhang and Long, 2024), few studies investigate how IWOM and spatial layout elements jointly shape activity intensity, limiting the design of optimization strategies for digital-era spatial vitality. To address these gaps, this study advances in two ways. First, it measures composite spatial vitality by integrating both static aggregation and dynamic interaction, using Shanghai’s township‑level units for visualization and 1km × 1km grids for analysis. Second, it applies interpretable machine learning to quantify the relative importance and nonlinear interactive effects of IWOM and spatial layout elements across different life services. Theoretically, this study clarifies the social interaction shift from geography-determined to digital visibility-driven mode. Practically, it offers a scientific basis for optimizing the configuration of life services, thereby informing strategies aimed at pursuing sustainable urban spatial vitality (Figure 1).

Figure 1.

Conceptual framework.

Literature review

Measurement of spatial vitality in the online-merge-offline context

Spatial vitality reflects the quality and livability of urban development through multiple dimensions, notably human activity, spatial function, and economic levels. Among these, human activity, particularly maintenance and leisure activities, is a direct indicator of social aggregation and spatial vitality (Gehl, 1992; Jacobs, 1961). Accordingly, spatial vitality is directly reflected in the intensity of human activity within a space (Cao et al., 2022; Yang et al., 2023). The concentration of physical elements (facilities, transport, environment) generates and attracts these activities, while the spatial clustering of industry and investment provides a macro-level socioeconomic indicator (Liu et al., 2022).

The online-merge-offline configuration of life services has rendered residents’ offline activities more spatially fragmented and interactively dynamic. Consequently, measuring vitality must expand beyond static aggregation of people to include dynamic connectivity between activities. However, existing research remains limited: Luo and Zhen (2019) modeled activity links as a weighted network to assess spatial centrality, and Liu et al. (2023b) used mobile data to compare activity and road networks. Few studies integrate both aggregation and connectivity to reflect vitality holistically, though Sun et al. (2022) stressed the need to combine dynamic networks with static elements like density.

In short, current measures rely heavily on population density and activity extent, overlooking activity connectivity and composite perspectives. This limits their ability to capture vitality in an era of spatiotemporally flexible behavior and to assess the attraction of life services in the online-merge-offline context.

Impacts of the configuration elements of life services on spatial vitality

The relationship between IWOM of life services and spatial vitality

Research on IWOM of life services has primarily examined its influence on the distribution of human activities, which indirectly reflects spatial vitality. Studies indicate that IWOM expands the spatial range of service access, reducing travel time and frequency for suburban residents while maintaining established patterns in urban centers (Niu et al., 2023; Yan, 2017). Zhang and Long (2024) observed that, while leisure activities remain concentrated in superior locations at a macro scale, IWOM itself tends to cluster in micro-scale disadvantaged sites. Similarly, implicit consumption spaces, namely the shops attracting customers through IWOM, still follow traditional location logic but are often linked to public spaces via indirect pathways (Jia et al., 2024).

The influence of IWOM also varies by type of activity. For leisure services, it can generate short‑term, highly clustered, and interactive visitation patterns (Zou, 2025). For maintenance activities, IWOM plays a supporting role in providing information and atmosphere, but high frequency and time‑sensitivity keep related behavior spatially bounded (Wu et al., 2024a). Additionally, different forms of IWOM correlate with distinct residential preferences: traditional e‑commerce is linked to suburban living, whereas on‑demand delivery and digital public services are associated with central‑city agglomeration (Wu et al., 2024b).

In summary, existing work details how IWOM reshapes activity patterns and distributions but seldom quantifies its direct effect on spatial vitality as measured through human activity. This gap limits our understanding of how IWOM contributes to vitality formation and to what extent online channels can enhance the utilization efficiency of life services in the digital era.

Nonlinear impacts of the spatial layout elements of life services on spatial vitality

Research on how the spatial layout elements, encompassing the type, quantity, location, and connectivity of resources, shapes spatial vitality is extensive and can be organized by scale. At the regional scale, factors like mixed land use, transport accessibility, and compact urban form are found to enhance socio-economic vitality (Chen et al., 2022; Xie et al., 2024). At the neighborhood scale, the density of public and commercial facilities, walkable street design, and accessible open spaces promote outdoor activity and daily service use, thereby supporting vitality (Gao et al., 2024; Li et al., 2022). At the street scale, elements like intersection density, building façade quality, and parking management influence pedestrian flow and staying.

The influence of spatial layout elements on spatial vitality is often nonlinear, manifesting in three ways: (1) Marginal effects: the impact of a single factor (e.g., building or intersection density) often peaks within an optimal range, exhibiting inverted U-shaped relationships (Li et al., 2022; Xie et al., 2024); (2) Spatial heterogeneity: the effect of the same element can vary in direction or strength across different urban contexts, requiring tailored strategies (Gao et al., 2024); (3) Interaction effects: factors can interact synergistically (e.g., street accessibility and building density) or antagonistically (e.g., road density and sidewalk width), altering their combined impact on vitality (Li et al., 2022; Yang et al., 2023).

While thresholds and interactions among spatial layout elements are well studied, few works specifically examine those elements characterizing life services’ spatial layout. More importantly, the interaction between these elements and the IWOM of services remains unexplored, limiting the ability to integrate virtual and physical factors in digitally-informed planning.

Methodology

Study area

Shanghai is in a leading position in China in the digital field of life services. On the one hand, the government has set goals such as building 100 benchmark scenarios for the digital transformation of life, 50 hospitals for digital transformation, and 15,000 intelligent parcel lockers (Shanghai Municipal People’s Government, 2021). On the other hand, Shanghai ranks first in China’s overall digital development index, demonstrating leadership in digital economy, infrastructure, society, and government (Xinhua News Agency, 2023). Therefore, this study takes Shanghai as the research area (Figure 2) and divides it into two spatial scales: firstly, to characterize the patterns of maintenance and leisure activities, township-level administrative divisions (streets, towns, townships), main urban areas, and urban circles are used as visualization boundaries. It should be noted that compared to administrative divisions, main urban areas and urban circles are basic units for spatial organization and resource allocation in urban centers and suburbs (Shanghai Municipal Planning and Natural Resources Bureau, 2021), better reflecting the characteristics of accessing life services at the regional level. Secondly, considering the sample size requirements of machine learning methods, 1 km × 1 km square grids are used as the basic units for variable statistics, totaling 2,095 grids.

Figure 2.

Research area.

Data

Table 1 presents the description of research data. (1) Geospatial data. Six first-level and 20 second-level life services (Supplementary Note 1) significantly influenced by Information and Communication Technology (ICT) are selected. It should be noted that the IWOM for the above services only serves as an information medium, and individuals still need to visit physical spaces to obtain them; (2) Data from Dianping.com , where the number of reviews for various life service stores is used to evaluate IWOM. Existing reports have found that 63.4% of respondents attach great importance to information on Dianping.com, and 82.7% and 69.2% of respondents use it to determine locations for catering and leisure activities, respectively (Liu and Gao, 2023), indicating the data’s representativeness and authenticity; (3) Mobile signaling data, including Origin-Destination (OD) data for two weekends based on 500 m × 500 m grids. This study focuses on weekends for analyzing maintenance and leisure activities, as residents primarily engage in necessary activities on weekdays.

Table.1.

Description of research data.

Type	Description	Source	Time
POIs	422,037 shapefile point data classified and coded according to POI categories on Gaode Map. Its fields include function type, name, address, longitude and latitude.	Gaode Map	2022.12
Building plan outline	412,099 shapefile polygon data of building model boundaries. Its fields include the number of building floors.		2023.9
Road network	15,388 shapefile line data of built roads. Its fields include grade.		2022.6
Transportation stops	2,104 shapefile point data of subway stations and 12,753 shapefile point data of bus stops. Its fields include name, longitude and latitude.		2022.6
IWOM	212,006 shapefile point data classified according to merchant categories. Its fields include function, store name, number of evaluations, address, longitude and latitude.	Dianping.com	2022.12
Mobile signaling data	23,192,385 travel data of Shanghai’s permanent residents.¹ Its fields include the latitude and longitude of the origin and destination grids, travel purpose,² number of travelers without sampling expansion, and recording time.	Telecom operator	2023.6.3-4、6.10-11

Measurement of spatial vitality based on travel data

Screening of travel data for accessing life services

Table 2 presents a sample of the travel dataset derived from mobile signaling. First, to validate its representativeness of Shanghai residents’ travel patterns, this study estimated the size of the permanent and employed populations within each administrative district based on time segments and travel purposes. Specifically, the permanent population was approximated by the number of users whose destination was “home” during the nighttime period (21:00-08:00), while the employed population was estimated from users with a “home-to-work” trip purpose during daytime hours (09:00-17:00). A Pearson correlation analysis was then performed between these estimates and the corresponding data from the 7th national population census (Table 3). The high correlation coefficients confirm that the dataset reliably captures travel patterns in reality.

Table 2.

Sample of the travel dataset.

Record date	Time period	Origin		Destination		Trip purpose	User count
Record date	Time period	longitude	latitude	longitude	latitude	Trip purpose	User count
2023 0603	20:00-20:30	120.8875	31.0975	120.9125	31.0975	other-to-other	1
	09:00-09:30	120.8875	31.0975	120.9175	31.1175	other-to-home	1
	12:00-12:30	120.8875	31.1175	120.9125	31.0975	home-to-other	1
	19:30-20:00	120.8875	31.1175	120.9175	31.1175	other-to-work	1
	21:00-21:30	120.8875	31.1175	120.9175	31.1175	home-to-other	1
	10:00-10:30	121.1075	31.1575	121.1125	31.1625	home-to-home	4
	06:00-06:30	120.8875	31.1175	120.9175	31.1175	home-to-work	2

Table 3.

Pearson correlation between mobile signaling and census population.

District	Permanent population		Employed population
District	Population census	Mobile signaling	Population census	Mobile signaling
Huangpu	662030	24092	24177	13481
Xuhui	1113078	39310	47562	22763
Changning	693051	20232	30357	11967
Jing’an	975707	39269	39815	22029
Putuo	1239800	39459	55251	22353
Hongkou	757498	22317	31512	13015
Yangpu	1242548	45930	49190	24739
Minhang	2653489	100539	143127	56124
Baoshan	2235218	49848	112071	31109
Jiading	1834258	61913	105761	37967
Pudong	5681512	192915	298063	100029
Jinshan	822776	33344	45147	17191
Songjiang	1909713	75625	111849	44778
Qingpu	1271424	50609	80533	28981
Fengxian	1140872	55898	64627	31120
Chongming	637921	22805	25880	10854
Correlation coefficients	0.975(p<0.001)		0.984(p<0.001)

Second, to extract trips associated with maintenance and leisure activities, records with the travel purpose “others” during the daytime were selected. To ensure representativeness, only trips that appeared consistently throughout the sampling period were retained. This resulted in 1,202,556 valid trip records with both origin and destination located within the study area.

Measurement of spatial vitality

Static aggregation (population density) and dynamic connection (spatial interaction intensity) were used to measure spatial vitality (Table 2).

(1) Population density. This indicator measures the aggregation degree of human activities in urban spaces. In this study, the weighted in-degree centrality $k_{i n, w e i g h t e d}$ per unit area of grid is calculated using the network visualization software Gephi. The calculation formula is:

k_{i n, w e i g h t e d} = \frac{\sum_{i = 1}^{n} w_{i n, i}}{S}

(1)

$w_{i n, i}$ is the total number of travelers from the starting point i to the destination point during the sampling period; n is the total number of starting nodes connected to the destination point; $S$ is the area of a 1 km × 1 km grid.

(2) Spatial interaction intensity. The activity flow and the corresponding number of travelers are abstracted into the edges and their weights of the activity network, reflecting the ability of each spatial unit, as a network node, to promote or attract the flow. Specifically, this study selects four second-level indicators, namely PageRank centrality, closeness centrality, betweenness centrality, and average clustering coefficient, for representation:

PageRank centrality. This indicator measures the importance of nodes in attracting dynamic connections. The calculation formula is:

C_{P R} (v) = (1 - d) + d \sum_{i \in N_{i n} (v)} \frac{P R (i)}{K (i)}

(2)

$C_{P R} (v)$ is the PageRank centrality of the destination point v; $d$ is the damping factor (usually set to 0.85); $N_{i n} (v)$ is the set of starting points connected to the destination point v; $P R (i), K (i)$ are the PageRank value and weighted out-degree of any starting point i connected to v, respectively.

Closeness centrality. This indicator measures the proximity between a node and other nodes in the activity network, reflecting the locational advantage of the spatial node v in promoting activity connections. The calculation formula is:

d_{v} = \frac{1}{n - 1} \sum_{i \neq v} d_{i v}

(3)

C_{C} (v) = \frac{1}{d_{v}} = \frac{n - 1}{\sum_{i \neq v} d_{i v}}

(4)

$C_{C} (v)$ is the normalized closeness centrality of the destination point v; $d_{i v}$ is the graph distance of the shortest path between the destination point v and the starting point i of the activity flow in the network, and the graph distance between interconnected nodes is 1.

Betweenness centrality. This indicator measures the frequency of a node appearing on the shortest path, reflecting the locational characteristics of the spatial node v in attracting people to stay and continue to flow. The calculation formula is:

C_{B} (v) = \sum_{i \neq v \neq j} \frac{σ_{ij} (v)}{σ_{ij}}

(5)

$C_{B} (v)$ is the betweenness centrality of the destination point v; $σ_{i j}$ is the number of shortest paths from node i to node j; $σ_{i j} (v)$ is the number of shortest paths from node i to node j that pass through node.

Local clustering coefficient. This indicator measures the clustering or grouping tendency of a node with other nodes in the network, so as to the activity connection patterns. The calculation ormula is:

C_{L} (v) = \frac{2 M_{v}}{k_{v} (k_{v} - 1)}

(6)

$C_{L} (v)$ is the local clustering coefficient of the destination point v; $M_{v}$ is the actual number of edges generated between node v and its neighbors; $k_{v}$ is the degree of node v.

Next, the weights of the above indicators were derived using the entropy weight method. After normalizing these indicators, the spatial vitality of each unit was calculated by weighted summation. The entropy weight method evaluates the variation of an indicator across samples: a smaller entropy value indicates greater dispersion, and thus a higher assigned weight. This weighting approach is objective and accounts for inter‑relationships among elements, making it suitable for this study. The method was applied twice: first to weight the second-level indicators of dynamic interaction intensity, and then to weight the primary indicators of spatial vitality (Table 4).

z_{i j} = \frac{X_{i j} - X_{m i n}}{X_{m a x} - X_{m i n}}

(7)

e_{j} = - k \sum_{i = 1}^{n} p_{i j} \ln (p_{i j}), j = 1, \dots, m

(8)

d_{j} = 1 - e_{j}

(9)

w_{j} = \frac{d_{j}}{\sum_{j = 1}^{m} d_{j}}

(10)

I = \sum_{j = 1}^{m} w_{j} x_{i j}, i = 1, \dots, n

(11)

$z_{i j}$ is the normalized weight of each indicator; $e_{j}$ is the information entropy of the j-th indicator; $p_{i j}$ is the proportion of the i-th sample value in the j-th indicator; $d_{j}$ is the information entropy redundancy; $w_{j}$ is the weight of the j-th indicator; $I$ is the spatial vitality of the grid.

Table 4.

Indicators of spatial vitality.

Indicator (Weight)		Mean	S.D.
Population density (0.858)	Weighted in-degree centrality (-)	4032.45	9129.03
Spatial interaction intensity (0.142)	PageRank centrality (0.300)	0.0005	0.0009
	Closeness centrality (0.036)	0.45	0.07
	Betweenness centrality (0.635)	0.0004	0.0014
	Local clustering coefficient (0.029)	0.53	0.24

Nonlinear impacts of the configuration elements of life services on spatial vitality

Explanatory variables

Table 5 shows the IWOM and spatial layout of life services, ultimately selecting 20 indicators across five dimensions as explanatory variables.

(1) IWOM of life services reflects the richness of information on spaces, products, and manual services that residents obtain from virtual platforms, influencing or determining individuals’ activity destinations (Zhang and Long, 2024); (2) Spatial accessibility and diversity indicate the availability and variety of specific life services within each unit (Gao et al., 2024). Differences in the scale and form of service supply are further distinguished by weighting various points of interest (POIs) with service‑capacity indices (Liu et al., 2023a) (Supplementary Note 1); (3) Transport accessibility is measured using roads density at different levels, bus‑stop coverage, and metro‑station coverage, capturing the commuting efficiency and population‑aggregation capacity of service locations (Xie et al., 2024); (4) Architectural and environmental design is represented by building density (horizontal layout) and floor‑area ratio (vertical layout), together with the proportion of water area as a typical natural element, to assess their role in attracting and concentrating people (Xie et al., 2024); (5) Spatial location is expressed as the distance from each grid to the geometric centers of regional‑level spatial‑organization and resource‑allocation units, reflecting the capacity to attract economic and social activity (Meng et al., 2025).

Table 5.

Description of explanatory variables.

Aspect	Indicator	Description	Mean	S.D.
IWOM of life services (X1)	retail commercial services (X1-1)	Ratio of the comments’ number on second-level POIs of various life services on Dianping to the grid area (number of comments/km²).	0.23	0.43
	catering services (X1-2)		0.13	0.22
	convenience services (X1-3)		0.06	0.10
	medical and health services (X1-4)		0.02	0.03
	leisure and entertainment services (X1-5)		0.05	0.09
Spatial accessibility and diversity of life services (X2)	retail commercial services (X2-1)	Ratio of the number of second-level POIs of various life services and their service capacity indices to the grid area (km⁻²).	15.63	80.07
	catering services (X2-2)		339.93	1608.71
	convenience services (X2-3)		58.43	226.20
	medical and health services (X2-4)		2.50	8.47
	leisure and entertainment services (X2-5)		34.28	133.92
	Diversity of life services (X2-6)	Functional diversity of first-level service facilities in the spatial grid. The Shannon diversity index³ is used, and the formula is $H_{i} = - \sum_{i = 1}^{n} (P_{i} \times \ln P_{i})$	0.64	0.31
Traffic accessibility (X3)	Road density to the urban interior (X3-1)	Ratio of the sum length of trunk roads and second-level roads to the grid area (km⁻¹).	0.42	0.27
	Road density to the regional scale (X3-2)	Ratio of the sum lengths of highways and expressways to the grid area (km⁻¹).	1.50	1.98
	Bus stop coverage rate (X3-3)	Ratio of the overlapping radiation area of all bus stops with a radius of 500 m to the grid area (%).	0.70	0.35
	Subway station coverage rate (X3-4)	Ratio of the overlapping radiation area of all subway stations with a radius of 1 km to the grid area (%).	0.33	0.43
Building and environmental design (X4)	Building density (X4-1)	Ratio of the building base area to the grid area (%).	0.05	0.07
	Plot ratio (X4-2)	Ratio of the total building area to the grid area.	0.68	0.97
	Water area ratio (X4-3)	Ratio of the area of planar water bodies to the grid area (%).	0.03	0.08
Spatial location (X5)	Distance to the geometric center of the urban circle where it is located (X5-1)	Straight-line distance from the geometric center to the geometric center of the urban circle where it is located (km).	7.47	5.19
Spatial location (X5)	Distance to the geometric center of central urban area (X5-2)	Straight-line distance from the geometric center to the geometric center of central urban area (km).	27.15	15.08

Control variables

Focusing on life services, the selection of independent variables is centered on maintenance and leisure activities. However, actual trips also include mandatory activities such as living and commuting (Gong et al., 2020; He et al., 2024). Driven by inelastic demand, these activities are directly dominated by spatial layout elements and not influenced by social media platforms. Accordingly, this study incorporates the densities of POIs related to residence, education and office functions as control variables (Table 6).

Table 6.

Description of control variables.

Indicator	Description	Mean	S.D.
Density of residence	Number of second-level POIs (Supplementary Note 1) categorized as mandatory activities per unit grid area (km⁻²).	1.85	3.28
Density of education and research		1.20	1.42
Density of business and office		1.71	2.84

Nonlinear impacts of the configuration elements of life services

Machine learning models have significant advantages in explaining the impacts of built environment elements on urban development, spatial vitality, and residents’ transportation patterns (Gao et al., 2023; Peng et al., 2023). In this study, Python is used to divide samples into a 70% training set and a 30% test set. GridSearch method is employed for five-fold cross-validation, and model performance is compared across multiple machine learning models using average R² and root mean square error (RMSE). Among ensemble learning methods such as Random Forest (RF), Light Gradient Boosting Machine (LightGBM), and EXtreme Gradient Boosting (XGBoost), the LightGBM model is selected due to its highest average R² (0.63) and lowest RMSE (0.071) for explaining variable relationships.

Compared with traditional linear regression models, LightGBM model offers the following advantages: (1) It can automatically capture complex nonlinear relationships between features and target variables (Xiao et al., 2021); (2) It can handle multicollinearity issues that linear regression cannot address. Compared with other ensemble machine learning models, its advantages lie in: (1) Compared with XGBoost, it uses histogram algorithms and parallel training, resulting in faster operation speed and lower memory usage; (2) Compared with RF, it has stronger capabilities in nonlinear relationship modeling and generalization.

This study uses SHapley Additive exPlanations (SHAP) via Python to interpret the regression results of the trained LightGBM model, where the Shapley values of samples are used to explain the influence degree of explanatory variables on dependent variables (Guo et al., 2023). Specifically, this study first determines the impact of IWOM of life services on spatial vitality; second, analyzes the effective influence thresholds of impactful elements; finally, investigates the interaction effects between IWOM and spatial layout elements to derive optimization methods for the configuration of life services in the online-merge-offline context.

Results

Analysis of spatial vitality based on activity aggregation and connection

Multi-dimensional indicators of spatial vitality

Each dimensional indicator is divided into levels 1 to 4 from low to high using the Natural Breaks method, revealing spatial vitality patterns (Figure 3): (1) For population density representing activity aggregation, the weighted in-degree centrality decreases from the main urban area to integrated and comprehensive development urban circles, with high values concentrated in the main urban area; (2) For interaction intensity representing activity connection, while the spatial patterns of each dimension indicator are discussed in detail in Supplementary Note 2, this section analyzes the spatial pattern of the comprehensive indicator (sum of weighted values): the distribution curve is a left-skewed inverted “U-shaped” when sorted from low to high, meaning most units have few activity connections with other units. Additionally, visual analysis of population flow patterns for maintenance and leisure activities shows that, using township-level administrative regions as origin points and the OD data with the largest traffic to represent dominant activity flows (35 groups of coincident flows are not marked), the dominant flow direction generally shows that people in the “central urban area-main urban area-urban circles” flow toward centers within their respective scopes. In terms of flow scale, the distribution follows an upward concave “L-shaped” curve, indicating significant differences in activity connection scales. Specifically, activity flows within the main urban area mainly correspond to high values of levels 3-5, while each urban circle has at least one activity flow reaching level 3 or above, confirming that second-level centers at the regional level have certain attraction in suburbans.

Figure 3.

Spatial vitality patterns of multi-dimensional indicators.

Spatial vitality patterns based on the weighted indicators

The entropy weight method yields a much higher weight for population density (0.858) than for interaction intensity (0.142), indicating that the ability of units to attract population aggregation per unit area differs more significantly than their ability to facilitate activity flows. In terms of spatial patterns (Figure 3), township-level administrative regions with spatial vitality of levels 3-4 are mainly concentrated within and on the edges of the main urban area, with some forming clustered distributions; level 2 townships are relatively evenly distributed on the edges of the main urban area and at regional centers within urban circles. In terms of data distribution, the it shows an upward concave “L-shaped” curve from low to high, indicating significant regional gaps in spatial vitality.

Nonlinear impacts of the configuration elements of life services on spatial vitality

Relative importance of the configuration elements of life services

First, based on the proportion of absolute Shapley values of explanatory variables, the importance ranking of each dimension is analyzed (Figure 4): spatial accessibility and diversity of life services (46.86%) > IWOM of life services (18.14%) > traffic accessibility (13.20%) > architectural and environmental design (11.63%) > spatial location (10.17%). Specifically, the influence of elements across dimensions is as follows: (1) For spatial accessibility and diversity of life services, the relative importance ranking is: retail commercial services (1/20) > catering services (2/20) > medical and health services (4/20) > leisure and entertainment services (7/20) > convenient services (9/20) > life service diversity (12/20).

Figure 4.

Importance ranking of each dimension of explanatory variables.

Second, regarding the importance ranking and impact directions of different elements (Figure 5), (1) when spatial accessibility of leisure and entertainment services and life service diversity have high values, they exhibit complex effects of both promoting and inhibiting human activities, while other elements show clear positive effects; (2) For IWOM of life services, only IWOM of retail commercial services (5/20) and convenient services (5/20) have relatively significant impacts, and high values of elements in this dimension exhibit obvious dual positive and negative effects; (3) For the other three dimensions, three elements significantly influence spatial vitality: distance to the geometric center of central urban area (3/20), plot ratio (6/20), and road density at the regional scale (8/20).

Figure 5.

Importance ranking and impact directions of explanatory variables.

Threshold effects of the configuration elements of life services

Based on the Shapley values of all grids corresponding to explanatory variables, they are sorted from smallest to largest, and trend lines are fitted to uncover influence thresholds and their spatial patterns. Due to space constraints, this study introduces the top 6 elements by relative importance (Figure 6), with the subsequent 6 elements discussed in detail in Supplementary Note 3:

(1) Spatial accessibility of retail commercial services (X2-1): When this element reaches 1 km⁻² in a unit grid, its impact no longer increases significantly, stabilizing within the range of 0.004-0.006. In terms of impact patterns, grids with level 5 influence are evenly distributed in local centers of each urban circle.

(2) Spatial accessibility of catering services (X2-2): The impact curve of this element first rises slowly, then rapidly, and gradually flattens. When the element is in the range of 0-0.25 km⁻², it has a negative impact on spatial vitality; when approaching 1 km⁻², its impact reaches a threshold of approximately 0.005. In terms of impact patterns, compared to retail commercial services, grids with level 5 influence are significantly more concentrated, showing a pattern of “high concentration in the main urban area-scattered distribution in suburbs”.

(3) Distance to the geometric center of central urban area (X5-2): The impact curve of this element first declines rapidly and then rises gradually. When the element is within 0-8 km of the grid, its impact decreases from 0.015 to 0, after which it has almost no effect. In terms of impact patterns, locations in the central urban area and comprehensive development urban circles have positive effects, while those in integrated improvement urban circles show negative effects.

(4) Spatial accessibility of medical and health services (X2-4): The impact curve of this element shows a slow initial rise followed by a rapid increase. When the element is in the range of 0-0.075 km⁻², the impact increases minimally; however, in the range of 0.075-0.1 km⁻², the impact rises rapidly and reaches a threshold of approximately 0.05. This indicates that the element has an effective impact range with minimum and maximum thresholds. Its spatial layout is similar to retail commercial services, showing a homogeneous impact pattern within the region.

(5) IWOM of retail commercial services (X1-1): The impact of this element exhibits a curve that rises sharply and then plateaus, transitioning from a negative to a positive correlation. When the element reaches approximately 200 reviews/km², it achieves an effective impact threshold of 0.01. In terms of impact patterns, grids in the central urban area show levels 3-5 impacts with positive effects, while most other grids have Level 2 impacts.

(6) Plot ratio (X4-2): When this element is between 0-3.2, it has a negative impact; in the range of 3.2-4, the impact increases rapidly and reaches a threshold of 0.01. Additionally, in terms of impact patterns, grids with level 5 impacts are concentrated in the central urban area, indicating that high plot ratios significantly promote spatial vitality in this area.

Figure 6.

Threshold effects of the configuration elements of life services.

Interaction effects between IWOM and spatial layout elements of life services

To avoid obscured interaction effects due to dimensional differences, independent and dependent variables are first normalized using range standardization. Then, SHAP algorithm is used to evaluate pairwise interactions among the 20 explanatory variables. Finally, the top 25% (100 out of 400 pairs) of interaction effects by average value are selected for analysis, resulting in 7 groups of explanatory variable combinations involving significant interactions between IWOM and spatial layout elements (Figure 7).

Figure 7.

Synergistic effects between IWOM and spatial layout elements of life services.

Overall, plot ratio exhibits significant interaction effects with the IWOM of multiple life services, including retail commerce, medical and health services, and leisure and entertainment services. Additionally, IWOM of retail commercial services interacts with spatial accessibility, plot ratio, distance to the geometric center of central urban area, and road density at the regional scale. Among the five life services, only IWOM of convenient services shows no significant interaction with any spatial layout elements.

Specifically: (1) When both IWOM of leisure and entertainment, medical and health services (X1-4, X1-5) and building plot ratio (X4-2) in a grid exceed certain values, they exhibit a clear positive interaction; (2) When IWOM of retail commercial services (X1-1) in a grid exceeds a certain value, it interacts with spatial accessibility (X2-1) and plot ratio (X4-2); and within a certain range, it shows clear interaction with road density at the regional scale (X3-2); when below a certain value, it significantly interacts with distance to the geometric center of central urban area (X5-2).

Discussion

Impact and threshold effects of IWOM of life services on spatial vitality

This study finds that IWOM of life services significantly influences spatial vitality, with explanatory power second only to spatial accessibility and diversity. The result aligns with the broader shift of social interaction under digital transformation from a geography‑determined toward a digital visibility‑driven paradigm (Park and Kim, 2024). Specifically, IWOM enhances spatial vitality through two pathways: 1) Spatial activation: by weakening the constraints of location and transport (Park and Kim, 2025; Wang et al., 2025), it offers low‑cost online visibility to small and medium‑sized merchants in disadvantaged areas, promoting a more balanced geography of vitality (Jia et al., 2024; Zhang et al., 2025); (2) Spatial interaction: it converts online attention into offline gathering, fostering multi‑layered spatial patterns, such as “core business districts-backstreet economy-building economy”, and intensifying agglomeration in traditionally strong locations. Nevertheless, the role of IWOM has limitations: (1) Groups highly dependent on social media platforms are prone to being influenced by algorithmic recommendations and preference filtering, falling into information cocoons composed of homogeneous information (Chen et al., 2025; Yan et al., 2025), which may deepen social segregation in physical space and undermine vitality balance; (2) Its core role is to enhance service perception, but differences in digital literacy among various groups lead to gaps in service cognition and screening capabilities (Liu et al., 2025), exacerbating inequities in resource access and reducing the inclusiveness of spatial vitality.

Notably, spatial layout factors of life services still dominate spatial vitality, including the accessibility of commercial, catering, and medical services, as well as the location advantages of central urban areas. This indicates that the independent impact of IWOM on spatial vitality is limited, which is consistent with the research conclusion of Zhang and Long (2024) on leisure consumption spaces in Beijing: IWOM has a specific distribution tendency, i.e., aggregating in locations with macro advantages but micro disadvantages (e.g., inside residential buildings near large urban business districts). Therefore, the integrated virtual‑physical model enhances spatial vitality in traditional core areas while activating surrounding disadvantaged ones, fostering a more balanced distribution of urban vitality and improving service quality.

Second, regarding IWOM of different life services, the impact ranking is: retail commercial services (5/20) > convenient services (11/20) > catering services (13/20) > medical and health services (16/20) > leisure and entertainment services (18/20), with retail commercial services’ IWOM playing a dominant role. This is because they occupy an absolute advantage in the life service system, so their IWOM greatly influences residents’ activity decisions. Additionally, convenient and catering services correspond to high usage frequencies and comprehensive supply forms, so IWOM mainly provides supplementary information on spatial environments and service quality, with limited influence on residents’ behavioral habits. In contrast, medical and health services provide relatively scarce manual services, while leisure and entertainment services offer fixed spatial experiences, so their spatial layouts restrict consumers’ choice ranges, resulting in limited roles for their IWOM in shaping spatial vitality. Wu et al. (2024b) noted differentiated effects of virtual amenities on residential mobility, attributing this to varying needs for face-to-face contact among services under the influence of ICT. This study further reveals significant differences in the capacity of digital-era life service configuration to shape spatial vitality. The findings suggest that giving priority to the online-merge-offline configuration of commercial and convenience services represents an effective approach to enhancing spatial vitality.

Finally, focusing on the threshold effects of IWOM of retail commercial and convenient services on spatial vitality: both exhibit significant promotion of spatial vitality when number of service reviews exceeds a minimum threshold. This differs from the threshold effects of physical elements like building density and intersection density, which show significant impacts only within specific value ranges (Li et al., 2022; Xie et al., 2024; Xiao et al., 2021). This provides refined guidance for promoting spatial vitality through retail commercial and convenient service configurations: leveraging the minimum threshold role of virtual platforms to activate spatial vitality while selecting space layout elements adapted to different scenarios. In terms of impact patterns, IWOM of convenient services generally enhances vitality across the entire urban area, while that of retail commercial services only has positive effects in the central urban area. This further reflects spatial differences in IWOM’s role in promoting spatial vitality: for retail commercial services, IWOM enhances agglomeration effects, exacerbating “service level differentiation” between central and peripheral areas; for convenient services, it plays a protective and wide-coverage role, narrowing regional gaps in service accessibility.

Interaction effects between IWOM and spatial layout elements of life services

First, except for convenient services, plot ratio interacts significantly with IWOM of retail commerce, medical and health, and leisure and entertainment services. This reflects that IWOM drives life services to penetrate vertical spaces, forming new gathering points of spatial vitality in the digital era. This result confirms that, beyond indoor leisure services (e.g., nail salons, gyms, private cinemas) and takeout catering services, other service types are also adopting the layout trend of “online virtual clustering and offline vertical distribution” (Jia et al., 2024; Talamini et al., 2022). On one hand, this provides a theoretical basis for activating the vitality of high-rise office buildings and idle commercial spaces by introducing more life services vertically at the block scale. Planning strategies for urban renewal should therefore incorporate vertical space evaluation and optimize vertical transportation connections: the former aims to enhance the functional richness and flexibility of high-rise buildings by introducing or dynamically adjusting life service types; the latter improves vertical accessibility by ensuring connectivity between high-rise buildings and public transport nodes. On the other hand, using IWOM to achieve precise resource matching and population aggregation has become a key direction for digital-era life service configuration, such as forming virtual clusters through recommendation algorithms while guiding business aggregation to meet diverse needs and sustain spatial vitality in vertical spaces.

Additionally, under the interaction of IWOM of retail commercial services with high resource accessibility, proximity to the central urban area, appropriate external traffic road density, and high plot ratio, spatial vitality is more effectively promoted. This not only indicates that retail commercial services are more adaptable to the online-merge-offline configuration compared to other life services but also provides specific guidelines for their spatial layout. The synergistic effects of superior location, traffic connectivity, and IWOM at the macro scale align with the conclusion that online attractiveness further amplifies agglomeration effects in highly accessible areas (Zhang et al., 2025), confirming that the configuration of retail commercial services with “high accessibility and online exposure” is a crucial source of spatial vitality.

Conclusion

This study focuses on resource configuration in the digital era and analyzes the nonlinear impacts of IWOM and spatial layout elements of life services on spatial vitality. On one hand, regarding the role of IWOM: (1) Overall, it significantly drives spatial vitality, reshaping the distribution logic of urban spatial vitality by shifting life service usage from “location determinism” to “digital visibility-driven approach” through spatial activation and interaction; (2) Specifically, spatial layout elements remain the primary drivers of spatial vitality, emphasizing the limited standalone effect of IWOM and the necessity of integration with physical spaces; (3) IWOM impacts vary across life services, with retail commerce and convenient services being more effective. Additionally, both exhibit a “minimum threshold” role in activating vitality, though their impact patterns differ spatially. On the other hand, regarding interaction effects between IWOM and spatial layout: (1) The significant interaction between plot ratio and IWOM of multiple life services not only demonstrates the trend of IWOM driving life services into vertical spaces but also highlights the potential of using IWOM for precise resource matching and population aggregation; (2) Compared to other life services, the online-merge-offline configuration of retail commercial services is the most effective in promoting vitality, characterized by “high geographic accessibility and online exposure”.

These findings hold significant implications: Theoretically, they reveal the impacts of IWOM itself and its interactions with spatial layout elements on spatial vitality, enriching the connotations of spatial vitality shaping in the digital era and providing new support for exploring new life service configuration methods. Practically, they emphasize the necessity of vertical space penetration of life services under IWOM influence and spatial layout transformation methods for retail commercial services, helping to enhance spatial vitality in the digital era and promote sustainable urban development.

However, this study has certain limitations. (1) Existing indicators used to measure spatial vitality are incomplete, future research should optimize activity flow analyses from the complex network perspective to accurately depict activity connection characteristics. (2) Mobile signaling data adopted in this study has inherent user base bias, tending to favor younger, more digitally connected groups while underrepresenting elderly residents and low-frequency consumers, which may affect the comprehensiveness of spatial vitality assessment. (3) Due to data constraints, this study characterizes IWOM only by the volume of online reviews. While this reflects the availability of service options in individual decision-making, it overlooks other diagnostic dimensions, such as review valence (emotional tone) and variance (perceptual heterogeneity), that also shape final choices (Babić Rosario et al., 2016; Li et al., 2025; Zhang et al., 2025). Future work could integrate multimodal large-language models to analyze textual and visual content of reviews, capturing dimensions like sentiment polarity and review focus (e.g., product, space, or service). This would support more refined, digitally informed service-allocation models and ultimately contribute to spatial vitality and urban quality.

Supplemental Material

sj-docx-1-tus-10.1177_27541231261426518 – Supplemental material for Nonlinear impacts of the configuration elements of life services on spatial vitality in the online-merge-offline context: A case study of Shanghai, China

Supplemental material, sj-docx-1-tus-10.1177_27541231261426518 for Nonlinear impacts of the configuration elements of life services on spatial vitality in the online-merge-offline context: A case study of Shanghai, China by Jing He, He Zhang, Linghong Ke, Wenpei Zhou and Jingyi Peng in Transactions in Urban Data, Science, and Technology

Footnotes

Author Note

We declare that this manuscript is original, has not been published before and is not currently being considered for publication elsewhere.

We confirm that the manuscript has been read and approved for submission by all the named authors.

ORCID iDs

He Zhang

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Natural Science Foundation of China [grant number 52078328].

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Supplemental material

Supplemental material for this article is available online.

Notes

Author biographies

Jing He, female, PhD candidate at the School of Architecture, Tianjin University. Her research focuses on smart city and community. E-mail: hej981008_@tju.edu.cn.

He Zhang, male, PhD, Professor at the School of Architecture, Tianjin University. His research interests include low-carbon urban planning, coastal zone planning, smart city and community. E-mail: ccazh2000@tju.edu.cn.

Linghong Ke, female, Master candidate at the School of Architecture, Tianjin University. Her research focuses on smart city and community. E-mail: kelinghong@tju.edu.cn.

Wenpei Zhou, female, Master candidate at the School of Architecture, Tianjin University. Her research focuses on smart city and community. E-mail: 2422505071@qq.com.

Jingyi Peng, female, PhD candidate at the School of Architecture, Tianjin University. Her research focuses on low-carbon urban planning. E-mail: pengjingyi@tju.edu.cn.

References

Babić Rosario

Sotgiu

De Valck

, et al. (2016) The effect of electronic word of mouth on sales: A meta-analytic review of platform, product, and metric factors. Journal of Marketing Research 53(3): 297–318.

Cao

Zhen

, et al. (2022) Urban temporal vibrancy mode and its influencing factors based on mobile signaling data: A case study of Nanjing, China. Human Geography 37(6): 109–116.

Chen

Zhao

Xiao

, et al. (2022) Investigating the spatiotemporal pattern between the built environment and urban vibrancy using big data in Shenzhen, China. Computers, Environment and Urban Systems 95: 101827.

Chen

Yeh

AG-O

(2021) Effects of built environment on activity participation under different spacetime constraints: A case study of Guangzhou, China. Travel Behaviour and Society 22: 84–93.

Chen

Qiu

(2025) The information cocoon paradox: Fostering unity or fueling divergence? Humanities and Social Sciences Communications 12(1): 859.

Dadashpoor

Yousefi

(2018) Centralization or decentralization? A review on the effects of information and communication technology on urban spatial structure. Cities 78: 194–205.

DataIntelo (2023) Online to offline commerce market report | Global forecast from 2025 to 2033. Available at: https://dataintelo.com/report/global-online-to-offline-commerce-market (accessed 31 Jaunary 2025).

Gan

, et al. (2024) Spatial vitality detection and evaluation in Zhengzhou’s main urban area. Buildings 14(11): 3648.

Gehl

(1992) Life between Buildings. China Architecture & Building Press.

10.

Gong

Jin

Liu

, et al. (2020) Identifying urban residents’ activity space at multiple geographic scales using mobile phone data. ISPRS International Journal of Geo-Information 9(4): 241.

11.

Gao

Sun

, et al. (2024) Exploring the relationship between urban vibrancy and built environment using multi-source data: Case study in Munich. Remote Sensing 16(16): 1107.

12.

Gao

Chong

H-Y

Zhang

, et al. (2023) Nonlinear effects of public transport accessibility on urban development: A case study of mountainous city. Cities 138: 104340.

13.

Guo

Jiang

Qiao

, et al. (2023) The nonlinear relationship between the active travel behavior of older adults and built environments: A comparison between an inner-city area and a suburban area. Sustainable Cities and Society 99: 104961.

14.

Chen

, et al. (2024) Exploring the activity-travel patterns of multi-purpose commuters on workdays based on activity chains and time allocation: Evidence from Kunming, China. ISPRS International Journal of Geo-Information 13(12): Article 12.

15.

Iimedia Research (2023) 2023-2024 China local life service industry market monitoring report. Available at: https://www.iimedia.cn/c400/92616.html (accessed 31 Jaunary 2025).

16.

Jacobs

(1961) The Death and Life of American Cities. Random House.

17.

Jia

Sun

(2024) Exploring factors influencing the agglomeration of implicit consumption space: A comparison with explicit consumption space. Applied Geography 169: 103340.

18.

Zhou

Zheng

X (Kevin)

, et al. (2025) Restaurant survival prediction using machine learning: Do the variance and sources of customers’ online reviews matter? Tourism Management 107: 105038.

19.

Jia

, et al. (2022) The six dimensions of built environment on urban vitality: Fusion evidence from multi-source data. Cities 121: 103482.

20.

Zhen

Zhang

(2023) Review and prospects of urban residents’ smart mobility. Urban Planning International 38(1): 110–116.

21.

Liao

Wang

Liu

(2024) Digitalization era: Investigating the spatial interplay between cyber human activity and economy with a hierarchical framework. Applied Geography 167: 103284.

22.

Lin

Guoqing

Yiwen

, et al. (2023) Spatializing the emerging geography of urban system in China: Based on live streaming commerce. Cities 143: 104613.

23.

Liu

Gao

(2023) Dianping Empowers Public Consumption: How Does the Online Rating System Alleviate Information Mismatch? Available at: https://mp.weixin.qq.com/s/Ut0m1OYG0nCGr7qm0Xn2HQ (accessed 31 Jaunary 2025).

24.

Liu

Gou

Xiong

(2022) Vital triangle: A new concept to evaluate urban vitality. Computers, Environment and Urban Systems 98: 101886.

25.

Liu

Onega

Moen

, et al. (2025) Digital divides in telehealth accessibility for cancer care in the United States. NPJ Digital Medicine 8(1): 534.

26.

Liu

Meng

Shi

(2023a) A multi-activity view of intra-urban travel networks: A case study of Beijing. Cities 143: 104634.

27.

Liu

Meng

, et al. (2023b). Exploring public transportation supply-demand structure of Beijing from the perspective of spatial interaction network. ISPRS International Journal of Geo-Information 12: 213.

28.

Luo

Zhen

(2019) How to evaluate public space vitality based on mobile phone data: An empirical analysis of Nanjing’s parks. Geographical Research 38(7): 1594–1608.

29.

Meng

Zeng

Huo

(2025) How do subway stations encourage the vitality of urban consumption amenities in Shanghai: A perspective on agglomeration. Journal of Transport and Land Use 18(1): 175–196.

30.

Niu

Zou

Chen

, et al. (2023) The second supporting role of virtual and real integration on decentralized activities in suburban new towns. Urban Development Studies 30(8): 38–44.

31.

Oldenburg

Brissett

(1982) The third place. Qualitative Sociology 5(4): 265–284.

32.

Park

Kim

(2024) Digitally accessible space: Discovery of the distinctive locational characteristics of restaurants with online reservation. Technological Forecasting and Social Change 209: 123788.

33.

Park

Kim

(2025) Does digital space shape physical space? Unraveling spatial characteristics of Third Places distinguished by online platform using explainable AI. Cities 165: 106144.

34.

Peng

Zhang

, et al. (2023) Nonlinear, threshold and synergistic effects of first/last-mile facilities on metro ridership. Transportation Research Part D: Transport and Environment 121: 103856.

35.

Pyrialakou

Gkritza

Fricker

(2016) Accessibility, mobility, and realized travel behavior: Assessing transport disadvantage from a policy perspective. Journal of Transport Geography 51: 252–269.

36.

Shanghai Municipal People’s Government (2021) 14th Five-Year Plan for Comprehensively Promoting Urban Digital Transformation in Shanghai. Available at: https://app.sheitc.sh.gov.cn/rddbta/690861.htm (accessed 31 Jaunary 2025).

37.

Shanghai Municipal Planning and Natural Resources Bureau (2021) Shanghai’s recent territorial spatial plan (2021-2025). Available at: https://ghzyj.sh.gov.cn/ghjh/20210901/678d1f5f707943fca0e746f4f58ea041.html (accessed 31 Jaunary 2025).

38.

Shao

Derudder

Yang

, et al. (2023) The association between transit accessibility and space-time flexibility of shopping travel: On the moderating role of ICT use. Journal of Transport Geography 111: 103661.

39.

Sun

Dai

Liu

, et al. (2022) Commuting carbon emissions in the perspective of spatial organization model: The case of Guangdong Hong Kong Macao Greater Bay Area. Urban Planning Forum 2022(7): 18–28.

40.

Talamini

(2022) From brick-and-mortar to location-less restaurant: The spatial fixing of on-demand food delivery platformization. Cities 128: 103820.

41.

Wan

Jha

Kazantsev

, et al. (2023) Online-to-offline platforms: Examining the effects of demand-side usage on supply-side decisions. Information & Management 60(2): 103757.

42.

Wang

Liu

, et al. (2025) Exploring urban activity intensity with Gravity-Based Betweenness Centrality and social media evaluation: A case study of Beijing. Sustainable Cities and Society 134: 106898.

43.

Wang

Liu

, et al. (2024) Identification of 71 factors influencing urban vitality and examination of their spatial dependence: A comprehensive validation applying multiple machine-learning models. Sustainable Cities and Society 108: Article 105491.

44.

Wei

Zhen

(2023) Characteristics and influencing factors of residents’ non-commuting travel in Nanjing metropolitan area. Scientia Geographica Sinica 43(10): 1740–1750.

45.

Williams

Pocock

(2010) Building “community” for different stages of life: Physical and social infrastructure in master planned communities. Community, Work & Family 13(1): 71–87.

46.

Niu

Ajiabibula

, et al. (2024a) Exploration of online and offline community life circle with virtuality-reality fusion: Theoretical advances and planning practices. Urban Planning Forum 2024(2): 25–33.

47.

Niu

Liang

, et al. (2024b) Impacts of virtual amenities on residential relocation and decentralization of megacity: An empirical study in wuhan, China. Applied Geography 167: 103273.

48.

Xia

Yeh

AG-O

Zhang

(2020) Analyzing spatial relationships between urban land use intensity and urban vitality at street block level: A case study of five Chinese megacities. Landscape and Urban Planning 193: Article 103669.

49.

Xiao

Liu

, et al. (2021) Nonlinear and synergistic effects of TOD on urban vibrancy: Applying local explanations for gradient boosting decision tree. Sustainable Cities and Society 72: 103063.

50.

Xiao

(2022) Human spatiotemporal behavior in the digital society: Logic change and research prospects. Progress in Geography 41(1): 86–95.

51.

Xie

Zhang

, et al. (2024) Integrating multi-source urban data with interpretable machine learning for uncovering the multidimensional drivers of urban vitality. Land 13(12): 2028.

52.

Xinhua News Agency (2023) Overall Layout Plan for the Construction of Digital China. Available at: https://www.gov.cn/zhengce/2023-02/27/content_5743484.htm (accessed 31 Jaunary 2025).

53.

Yan

(2017) “Homogenization-reaggregation”: The impact of online take-out service on restaurant location characteristics: A study based on open data in Shanghai inner city. Urban Planning Forum (4): 113–119.

54.

Yan

Wang

Jiao

(2025) Promoting or inhibiting? The impact of the information cocoon on customer stickiness in e-commerce platforms: The moderating role of decision-making style. Electronic Commerce Research and Applications 71: 101505.

55.

Yang

Wang

Han

(2023) Nonlinear and synergistic effects of the built environment on street vitality: The case of Shenyang. Urban Planning Journal 2023(5): 93–102.

56.

Zhang

Long

(2025) Uncovering how online attractiveness mediates and moderates the impact of multi-scale transportation accessibility on non-work mobility. Travel Behaviour and Society 39: 100991.

57.

Zhang

Long

(2024) The rise of hidden spaces: A study of leisure consumption spaces in Beijing using Dazhong Dianping data. Tourism Tribune 39(4): 16–27.

58.

Zhang

Liang

, et al. (2025) Why restaurants survive? A perspective of word-of-mouth dispersion. Tourism Management 109: 105132.

59.

Zhou

Yang

Chen

, et al. (2023) Commuting versus consumption: The role of core city in a metropolitan area. Cities 141: 104495.

60.

Zou

(2025) Transcoding a wanghong city: Mediatized culturalization of urban places in China. Media, Culture & Society 47(1): 22–40.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

2.27 MB