Evaluating the Quality of High-Frequency Pedestrian Commuting Streets: A Data-Driven Approach in Shenzhen

Highlights

What are the main findings?

• A mismatch between usage and quality: only 2.15% of high-frequency pedestrian commuting streets in Shenzhen perform well across all three evaluation dimensions, indicating a substantial gap between street quality and usage frequency.
• Regional disparities in street distribution and quality: Approximately 70% of these streets are concentrated in the northern areas, where the overall quality is relatively poor. In contrast, the southern region has slightly better street quality but still requires improvements.

What is the implication of the main finding?

• These findings highlight the need for targeted urban planning and resource allocation to optimize street environments. By addressing the mismatch between usage frequency and street quality, and by improving the overall quality of high-frequency pedestrian commuting streets, urban experiences can be significantly enhanced. This study provides valuable insights for policymakers and urban planners to prioritize interventions in areas with the highest pedestrian activity and to reduce regional disparities in street quality.

Abstract

Streets, as critical public space nexuses, require synergistic quality–utilization alignment—where quality without use signifies institutional inefficiency, and use without quality denotes operational ineffectiveness. Focusing on high-frequency pedestrian commuting streets (HFPCSs) that not only crucially mediate metropolitan mobility patterns but also shape citizens’ daily urban experiences and satisfaction, this study proposes a data-driven diagnostic framework for street quality–utilization assessment, integrating multi-source urban big data through a case study of Shenzhen. By integrating multi-source urban big data, we identify HFPCSs using LBS data and develop a multi-dimensional evaluation system that incorporates 1.07 million Points of Interest (POIs) for assessing convenience, utilizes DeepLabv3+ for the semantic segmentation of street view imagery to evaluate comfort, and leverages 15,374 km of road network data for accessibility analysis. The results expose dual mismatches: merely 2.15% of HFPCSs achieve balanced comfort–convenience–accessibility benchmarks, while over 70% of these are clustered in northern districts, exhibiting systematically inferior quality metrics across dimensions. Diagnostic analysis reveals specific planning and spatial configurations contributing to these disparities, informing targeted retrofitting strategies for priority street typologies. This approach establishes a replicable model for megacity street renewal, deploying supply–demand diagnostics to synchronize infrastructure upgrades with pedestrian flow realities. By bridging data insights with human-centric urban improvements, this framework demonstrates how smart city technologies can concretely address the quality–utilization paradox—advancing sustainable urbanism through evidence-based street transformations.

1. Introduction

Streets are a fundamental component of urban public space, serving as indispensable venues for daily human interaction with the city [1]. They also play a vital role in shaping citizens’ everyday urban experiences and perceptions [2]. In recent years, the improvement of street quality has gained increasing attention from researchers and urban policymakers [3], leading to the widespread implementation of street beautification projects in many cities [4].

The study of street quality has evolved into a critical issue in the discourse on urban spatial optimization and sustainable development, attracting growing attention from both academia and planning practice. Research perspectives have evolved from classical humanistic spatial theories, such as Lynch’s concept of “imageability” [5] and Gehl’s emphasis on the “human scale” [6], toward multi-dimensional evaluation frameworks. These frameworks incorporate the interplay between physical spatial form [7], functional activity patterns [8], and socio-perceptual experiences [2]. For instance, Carmona et al. [2] proposed a comprehensive “place dimension” model, emphasizing the integration of morphological, functional, perceptual, and control dimensions in urban space quality. Recent empirical studies have further expanded this perspective; Wang et al. [9] investigated how various streetscape elements, such as lighting, greenery, and sidewalk width, influence pedestrians’ perceptions of safety and aesthetic appeal. Doiron et al. [10] utilized Google Street View images and deep learning techniques to assess urban environmental features that predict walking-to-work rates, highlighting the role of streetscape characteristics in promoting active transportation. Research methodologies have also evolved significantly with technological advancement. Traditional field observation [11] has been augmented by space syntax analysis [8], GIS-based spatial modeling, and large-scale urban data analytics [12]. Recent developments in artificial intelligence, such as street view image recognition [13], and perceptual experiments using VR simulations and eye-tracking technologies [14] have further enhanced the precision and temporal dynamics of street environment assessments. At the same time, the focus of street research has shifted from static spatial descriptions to dynamic social responses, increasingly addressing the impacts of climate change [15], emergent crises, and social equity concerns [16] on street usage. This shift has emphasized the importance of data-driven governance [17] and inclusive urban design. Although the majority of existing studies grounded in Western urban contexts have insufficiently addressed the particularities and complexities of high-density Asian cities, technology-enabled street research is evolving at an unprecedented pace.

However, the continuous refinement of street research contrasts with the widespread imbalance in actual street quality. The persistent gap between theory and practice remains a central challenge in this field. This disconnection stems from two fundamental dilemmas. First, the mechanisms used to select streets for improvement lack a human-centered orientation. Urban streets account for 10% to 20% of city land area, making comprehensive upgrades ideal but difficult to operate. Existing selection criteria, whether based on geographic zones or hierarchical classifications, tend to reflect top-down, automobile-oriented perspectives, rather than actual pedestrian usage patterns. As a result, cities often exhibit a paradoxical coexistence of poorly maintained high-use streets and excessively beautified low-activity streets, leading to a misallocation of public resources. Second, there is a critical disjunction between strategy formulation and implementation. Although many urban design guidelines provide general principles for street optimization, they often fail to identify specific “high-frequency but low-quality” streets or distill the core issues they face. This limits their applicability for delivering effective, context-sensitive interventions. Therefore, there is an urgent need to reframe street renewal from a usage-based, human-centered perspective. This includes systematically identifying the spatial patterns of high-frequency pedestrian streets under different technical frameworks, assessing their current quality status, and uncovering common deficiencies. Such an approach can provide both methodological tools and analytical insights for targeted, evidence-based street upgrading strategies in the era of urban stock optimization.

Targeted research on street usage frequency remains limited. Most existing studies treat usage frequency as a supplementary variable within broader street quality evaluations. For instance, Zhang et al. [18] and Wang [19] applied discrete choice models to reveal pedestrian route preferences, but did not link these preferences to the quality characteristics of frequently used streets. Similarly, Huang [20] constructed a walking path model for transit-oriented travel, focusing on the connection strength between transit stations and the street network, without delving into the spatial shortcomings of high-frequency streets. Ye et al. [21] combined sDNA-based accessibility analysis with street view evaluations to identify streets with renewal potential, yet failed to quantify the specific issues faced by frequently used streets. To address these gaps, we [22] proposed a tripartite selection framework in 2023 driven by behavioral, mobility, and spatial logics to identify high-frequency pedestrian streets. Building upon the existing literature on street quality quantification, we developed the PEP (People–Event–Place) index system, which evaluates the alignment between pedestrian frequency (People), functional density (Event), and greening level (Place). The results revealed a significant spatial mismatch in Shenzhen: only 1.53% of all streets demonstrated an alignment of high frequency and high quality. Similarly, Liu et al. [23] used path planning APIs to identify high-frequency streets in Xi’an and found that just 6.95% of them exhibited both high environmental quality and functional diversity, even in a historically rich city. These findings suggest that supply–demand mismatches in street environments may be a widespread phenomenon. It is important to note that high-frequency pedestrian streets are often closely tied to residents’ daily commuting activities. While these streets are typically considered individually in terms of their transportation function, once they carry a threshold level of pedestrian flow, they evolve into critical urban corridors used multiple times per day by large volumes of people. At this point, such streets function not only as transportation infrastructure but also as vital public spaces that aggregate social interaction, reflect urban character, and significantly shape everyday urban experiences. Although these streets constitute a small fraction of the overall street network, their elevated exposure and frequency of use generate a “frequency–perception” multiplier effect, making them a key determinant of perceived urban quality and residents’ well-being.

Therefore, this study addresses a fundamental research question: “What are the spatial distribution patterns, environmental quality characteristics, and functional classifications of high-frequency commuting streets in high-density urban areas, and how can precision design interventions enhance their spatial performance?” Building upon the logical framework established in previous research, this study further focuses on high-frequency commuting streets as critical spatial carriers of everyday urban life. Leveraging large-scale Location-Based Services (LBS) data across the entire city, we conduct continuous temporal monitoring of commuting behaviors, enabling real-time identification and quantification of high-frequency pedestrian flows. This approach extends our theoretical framework into an empirical model grounded in real-world behavioral statistics. Then, we integrate multiple sources of urban big data, including POI (Point of Interest) datasets, street view imagery, road network data, and building footprints, to construct a comprehensive, multi-dimensional street quality diagnostic system. Compared to the original two indicators (functional density and greening rate), the revised framework includes six evaluation indicators across three major dimensions: functional diversity and functional density (convenience), green view index (GVI), sky openness, and street enclosure (comfort), as well as walkability (accessibility). Through functional clustering analysis, the study identifies the common weaknesses shared across all types of high-frequency commuting streets, as well as the structural contradictions unique to key street types. Based on these findings, we propose a targeted renewal strategy rooted in a “demand-tiered, type-responsive” approach. This study thus offers both methodological tools and data-driven evidence to support the creation of high-quality pedestrian street environments and enhance residents’ daily urban experiences.

2. Materials and Methods

2.1. High-Frequency Pedestrian Commuting Street Selection Mechanism

High-frequency pedestrian commuting streets refer to urban roads that carry a substantial volume of regular pedestrian commuting activity during weekday morning and evening peak hours, characterized by stable and significant foot traffic density associated with daily round-trip travel. The current methods for measuring pedestrian street usage frequency can be broadly categorized into two approaches: field surveys and interview-based counting methods, and big data-driven quantitative analysis. Field surveys and questionnaire-based methods offer high controllability and reliability, allowing for the collection of in-depth data beyond verbal and written responses. These methods enable researchers to better understand respondents’ attitudes and needs. However, they are labor-intensive, time-consuming, and limited in spatial and temporal coverage, making them impractical for large-scale street monitoring.

With the advancement of big data, many studies have adopted various data sources to assess pedestrian street usage [24,25], including social media data, such as Weibo (e.g., Twitter) analytics [26,27,28,29,30], mobile signaling data [31], heatmap data [32], and LBS data generated by mobile internet services [27,33,34,35]. While big data may provide less detailed information compared to traditional survey methods and often struggles to distinguish individual behavioral differences, it offers significant advantages in terms of broad spatial coverage, high temporal granularity, and computational efficiency. These characteristics make it particularly effective for identifying group behavioral patterns and spatial distribution trends [36].

In this study, LBS data were selected for their ability to measure street usage across an entire city while also capturing temporal variations in pedestrian activity. Additionally, LBS data can achieve GPS accuracy at the 10 m scale [14], allowing for an approximation of street usage at a human-centered scale. Drawing on prior research that has employed LBS data to analyze urban pedestrian activity [36,37], we constructed a measurement framework for assessing street usage frequency based on LBS data.

The methodology consisted of three steps. First, for each selected date, we calculated the pedestrian density per unit street length at hourly intervals during commuting hours to obtain the street usage frequency at specific time points ($P_{ij}$). Second, we computed the average street usage frequency ($P_i$) by averaging the values of the same time point across multiple dates. Finally, we aggregated the street usage frequency across all commuting time points to derive the overall street usage frequency ($P$) for the commuting period. The specific measurement method is as follows:

• Street Usage Frequency at a Specific Time Point on a Given Date (P_ij)

The street usage frequency at a specific time point on a given date is determined by the pedestrian density per unit length of the street. The calculation formula is as follows:

$$P_{ij}={{\displaystyle \frac{Po{p}_{ij}}{L}}}^{\prime }$$

where

$P_{ij}$ represents the street usage frequency at time point i on day j for a specific street, measured in occurrences per meter;

$Po{p}_{ij}$ denotes the number of pedestrians observed on the street during hour i on day j (i.e., the total usage count);

L represents the length of the street in meters;

j refers to the day in the LBS dataset (j = 1, 2, 3…);

i represents the time point within a day (i = 1, 2, 3, …, 24).

2.

Street Usage Frequency at a Specific Time Point (P_i)

The street usage frequency at a specific time point is calculated from a temporal perspective, capturing the dynamic changes in street usage over different time slices. It reflects the relative frequency of street usage at different times. The calculation formula is as follows:

$$Pi={\displaystyle \frac{1}{n}}\ast {\displaystyle \sum }_{j=1}^{n}Pij$$

(1)

where

Pi represents the street usage frequency at a specific time point;

n is the number of sample days of LBS data in the study.

3.

Street Usage Frequency Over a Specific Time Period (P)

To quantify the street usage frequency over a selected time period (such as the commuting period), we introduce the concept of cumulative street usage frequency for a specific time range. The formula is as follows:

$$P={\displaystyle \sum }_{i=m}^{M}Pi$$

where

P represents the street usage frequency over the selected time period;

m and M denote the start and end time points of the chosen period;

Pi represents the street usage frequency at each time point within the selected period.

2.2. Street Quality Evaluation System

As shown in Table 1, this study conducted a literature review on the CNKI and Web of Science platforms, retrieving and filtering publications from 2013 to 2024 that focused on “street spatial quality”. The most frequently mentioned street quality evaluation dimensions included comfort, convenience, accessibility, safety, diversity, and aesthetics. Based on these findings, and adhering to the principles of comparability, quantifiability, data availability, scientific rigor, and validity, this study selected comfort, convenience, and accessibility as the primary evaluation dimensions. Further analysis of existing research identified key indicators with significant influence within each dimension. Specifically, we identified the following:

• Comfort is primarily assessed through street greenery, sky openness, and street scale [38,39].
• Convenience is evaluated based on the number and diversity of Points of Interest (POIs) [38,39].
• Accessibility is determined by the degree of pedestrian network connectivity [38,39].

Ultimately, as shown in

Table 2. Street evaluation variable attributes.

Aspect	Category	Variable	Measurements	Description	Source	Weights
Street Usage by People	Street Usage Frequency	Street Usage Frequency (Pi)	$Pi={\displaystyle \frac{1}{n}}\ast {\displaystyle {\displaystyle \sum }_{j=1}^n}Pij$	The relative usage frequency of streets at different times is represented by the average street usage frequency at a specific time point over a given period.	LBS	/
Street Quality	Convenience (Cvqi)	POI Business Density (Pdi)	$Pdi={\displaystyle \frac{Poi num}{L_{road length}}}$	This measures the density of business establishments along the street. Poi num represents the number of business establishments on the street, with the unit being counted; $L_{road length}$ represents the length of the street in meters.	Baidu POI	0.3947
		Poi Business Diversity (Ldi)	$Ldi=-{\Sigma }_{i=1}^{s}Pi\mathit{ln}Pi$	This uses the Shannon index to measure the diversity of business establishments. S represents the total number of Poi types, an Pi d represents the proportion of the i-th type of Poi to the total number. When there is only one type of Poi on the street, the Shannon index, which measures Poi diversity, is at its minimum value of 0.	Baidu POI	0.6053
	Comfort (Cqi)	Street Green View Index (Gvi)	$\mathit{Gvi}={\displaystyle \frac{S_{\mathit{green}}}{S_{\mathit{total}}}}$	This indicates the proportion of green vegetation visible in street view images. $S_{green\_n}$ represents the pixels occupied by greenery in the street view image, with the identifier “n”, and $S_{total}$ represents the total number of pixels in the entire street view image.	Baidu Street View (2021)	0.5524
		Sky Openness Index (Svi)	$Svi={\displaystyle \frac{S_{sky}}{S_{total}}}$	This indicates the proportion of the sky visible in street view images. $S_{sky}$ represents the pixels occupied by the sky in the street view image and $S_{total}$ represents the total number of pixels in the entire street view image.	Baidu Street View (2021)	0.218
		Street Enclosure Index (Sei)	$Sei={\displaystyle \frac{S_{building}+S_{column}}{S_{total}}}$	This measures the extent to which the street is enclosed by buildings and columns. $S_{building}$ represents the pixels occupied by buildings in the street view image, $S_{column}$ represents the pixels occupied by columns in the street view image, and $S_{total}$ represents the total number of pixels in the entire street view image.	Baidu Street View (2021)	0.2295
	Accessibility (Aqi)	Walkability (Rai)	$Rai=Betweenness800$	This measures walkability based on the betweenness centrality at an 800 m distance.	sDNA	1

, a quantitative analysis of street quality was conducted using six key indicators across the three selected dimensions: comfort, convenience, and accessibility.

Table 1. Evaluation dimensions of street spatial quality.

Indicator	Reference
Comfort	Wang et al., 2023 [40]; Liu et al., 2024 [23]; Rui et al., 2023 [41]; Li et al., 2024 [42]; Cui et al., 2023 [43]; Guo et al., 2024 [44]; Wang et al., 2023 [40]; Wang et al., 2023 [45]; Fruin, 1971 [46]; Zeng et al., 2022 [47]; van den Berg et al., 2017 [48]; Lo, 2009 [49]; Arellana et al., 2020 [50]; Kim et al., 2014 [51]; Bivina et al., 2019 [52]; Alfonzo et al., 2008 [53]; Ball et al., 2001 [54]; Hu et al., 2020 [55]; Ye et al., 2019 [56]; Tang et al., 2017 [57]; Shao et al., 2022 [58]; Huang et al., 2023 [59]; Zhou et al., 2022 [60]; Ye et al., 2019 [61]; Fan et al., 2019 [62]; Xu et al., 2017 [63]; Hu et al., 2020 [38]; Si et al., 2021 [39]
Convenience	Liu et al., 2024 [23]; Li et al., 2024 [42]; Cui et al., 2023 [43]; Guo et al., 2024 [44]; Wang et al., 2023 [45]; Zeng et al., 2022 [47]; van den Berg et al., 2017 [48]; Lo, 2009 [49]; Arellana et al., 2020 [50]; Kim et al., 2014 [51]; Bivina et al., 2019 [52]; Asadi-Shekari et al., 2014 [64]; Owen et al., 2004 [65]; Humpel et al., 2002 [66]; Humpel et al., 2004 [67]; Ball et al., 2001 [54]; Hu et al., 2020 [55]; Shao et al., 2022 [58]; Huang et al., 2023 [59]; Fan et al., 2019 [62]; Xu et al., 2017 [63]; Di et al., 2021 [68]; He et al., 2022 [69]; Hu et al., 2020 [38]
Accessibility	Harvey et al., 2017 [70]; Batty, 2012 [71]; Xu et al., 2024 [72]; Guo et al., 2024 [44]; Wang et al., 2022 [73]; Ma et al., 2024 [74]; Alfonzo et al., 2008 [53]; Owen et al., 2004 [65]; Humpel et al., 2002 [66]; Hu et al., 2020 [55]; Ye et al., 2019 [56,75]; Ye et al., 2019 [61]; Fan et al., 2019 [62]; Xu et al., 2017 [63]; Hu et al., 2020 [38]; Di et al., 2021 [68]
Safety	Ma et al., 2024 [74]; Wang et al., 2023 [40]; Cui et al., 2023 [43]; Li et al.,2023 [76]; Wang et al., 2022 [73]; Wang et al., 2024 [77]; Wang et al., 2023 [45]; Ma et al., 2024 [74]; Zeng et al., 2022 [47]; van den Berg et al., 2017 [48]; Lo, 2009 [49]; Arellana et al., 2020 [50]; Bivina et al., 2019 [52]; Asadi-Shekari et al., 2014 [64]; Alfonzo et al., 2008 [53]; Hu et al., 2020 [55]; Shao et al., 2022 [58]; Huang et al., 2023 [59]
Diversity	Cui et al., 2023 [43]; Shahideh, 2013 [78]; Zeng et al., 2022 [47]; Fan et al., 2023 [79]
Aesthetics	Cui et al., 2023 [43]; Li et al., 2023 [76]; Wang et al., 2024 [77]; Owen et al., 2004 [65]; Humpel et al., 2002 [66]; Humpel et al., 2004 [80]; Humpel et al., 2004 [67]; Ball et al., 2001 [54]

Table 1. Evaluation dimensions of street spatial quality.

Indicator	Reference
Comfort	Wang et al., 2023 [40]; Liu et al., 2024 [23]; Rui et al., 2023 [41]; Li et al., 2024 [42]; Cui et al., 2023 [43]; Guo et al., 2024 [44]; Wang et al., 2023 [40]; Wang et al., 2023 [45]; Fruin, 1971 [46]; Zeng et al., 2022 [47]; van den Berg et al., 2017 [48]; Lo, 2009 [49]; Arellana et al., 2020 [50]; Kim et al., 2014 [51]; Bivina et al., 2019 [52]; Alfonzo et al., 2008 [53]; Ball et al., 2001 [54]; Hu et al., 2020 [55]; Ye et al., 2019 [56]; Tang et al., 2017 [57]; Shao et al., 2022 [58]; Huang et al., 2023 [59]; Zhou et al., 2022 [60]; Ye et al., 2019 [61]; Fan et al., 2019 [62]; Xu et al., 2017 [63]; Hu et al., 2020 [38]; Si et al., 2021 [39]
Convenience	Liu et al., 2024 [23]; Li et al., 2024 [42]; Cui et al., 2023 [43]; Guo et al., 2024 [44]; Wang et al., 2023 [45]; Zeng et al., 2022 [47]; van den Berg et al., 2017 [48]; Lo, 2009 [49]; Arellana et al., 2020 [50]; Kim et al., 2014 [51]; Bivina et al., 2019 [52]; Asadi-Shekari et al., 2014 [64]; Owen et al., 2004 [65]; Humpel et al., 2002 [66]; Humpel et al., 2004 [67]; Ball et al., 2001 [54]; Hu et al., 2020 [55]; Shao et al., 2022 [58]; Huang et al., 2023 [59]; Fan et al., 2019 [62]; Xu et al., 2017 [63]; Di et al., 2021 [68]; He et al., 2022 [69]; Hu et al., 2020 [38]
Accessibility	Harvey et al., 2017 [70]; Batty, 2012 [71]; Xu et al., 2024 [72]; Guo et al., 2024 [44]; Wang et al., 2022 [73]; Ma et al., 2024 [74]; Alfonzo et al., 2008 [53]; Owen et al., 2004 [65]; Humpel et al., 2002 [66]; Hu et al., 2020 [55]; Ye et al., 2019 [56,75]; Ye et al., 2019 [61]; Fan et al., 2019 [62]; Xu et al., 2017 [63]; Hu et al., 2020 [38]; Di et al., 2021 [68]
Safety	Ma et al., 2024 [74]; Wang et al., 2023 [40]; Cui et al., 2023 [43]; Li et al.,2023 [76]; Wang et al., 2022 [73]; Wang et al., 2024 [77]; Wang et al., 2023 [45]; Ma et al., 2024 [74]; Zeng et al., 2022 [47]; van den Berg et al., 2017 [48]; Lo, 2009 [49]; Arellana et al., 2020 [50]; Bivina et al., 2019 [52]; Asadi-Shekari et al., 2014 [64]; Alfonzo et al., 2008 [53]; Hu et al., 2020 [55]; Shao et al., 2022 [58]; Huang et al., 2023 [59]
Diversity	Cui et al., 2023 [43]; Shahideh, 2013 [78]; Zeng et al., 2022 [47]; Fan et al., 2023 [79]
Aesthetics	Cui et al., 2023 [43]; Li et al., 2023 [76]; Wang et al., 2024 [77]; Owen et al., 2004 [65]; Humpel et al., 2002 [66]; Humpel et al., 2004 [80]; Humpel et al., 2004 [67]; Ball et al., 2001 [54]

2.3. Street Classification

There are various methods for classifying streets, including transportation function [81], functional type [82], and spatial type [83,84]. This study focused on street classification based on the functional composition of adjacent buildings. Therefore, we categorized streets according to five core building functions: residential, office, school, commercial, and hospital. By analyzing the adjacent buildings along both sides of each street, we identified different combinations of these functions and classified streets accordingly. The frequency distribution of these street types was then ranked to determine the predominant functional street types in Shenzhen. Based on this classification, we conducted feature analysis and problem identification for the key street types.

2.4. Analytical Framework and Research Design

This study followed a three-step research process. First, high-frequency pedestrian commuting streets across the city were identified using large-scale LBS (Location-Based Services) data during commuting hours. Second, the selected streets were evaluated based on three key dimensions—comfort, convenience, and accessibility—to assess their overall quality. Finally, a comprehensive quality assessment was conducted for all functional street types, with a detailed analysis of key street types. The study summarizes patterns and common issues, providing targeted optimization strategies to improve street environments and enhance urban pedestrian experiences (Figure 1).

2.5. Study Scope and Data Processing

2.5.1. Study Scope

This study focused on the entire urban area of Shenzhen, which consists of nine administrative districts and one new district. Shenzhen is the pioneering city of China’s reform and opening-up. A total of 108,963 streets with available street view imagery were selected as the research objects. The study specifically targeted high-frequency pedestrian commuting streets during peak commuting hours on weekdays. Based on prior research on commuting behavior patterns in Shenzhen [85,86], the study defined the commuting peak periods as 7:00–10:00 AM and 5:00–8:00 PM. Data were collected over 14 consecutive days (16 June–29 June 2022), covering both weekdays and weekends under consistent weather conditions (mild temperatures and clear skies). The data collection range was set to a 55 m buffer on both sides of the street centerline [31,87].

2.5.2. Data Processing

The datasets used in this study support high-frequency pedestrian commuting street identification, street quality evaluation (accessibility, convenience, and comfort), and street classification. The data sources included LBS data, road network data, POI data, Baidu Street View imagery, and Shenzhen building data. The road network dataset was from Baidu Maps 2021 and consisted of 173,296 street segments with a total length of 15,374 km. To mitigate the seasonal bias, a total of 324,896 Baidu Street View images (1600 × 800 pixels) in spring and summer were processed [88], covering 108,963 street segments, including a substantial number of segments within urban villages. Street usage frequency was calculated based on high-precision mobile LBS data from Jike Company, covering the study period of 16 June–29 June 2022. Jike by iDream Network Technology employs spatiotemporal signal fingerprint location technology to accurately track mobile phones by analyzing unique environmental signals such as Wi-Fi, Bluetooth, and base stations. The system gathers various types of data, including GPS, Scene Recognition Points, and environmental signals, achieving high-precision location identification, with POIs accurate to within 2–3 m. The SDK is integrated into mobile phones, enabling real-time signal scanning and location verification. Convenience indicators were derived from 2021 Baidu POI data, which include approximately 1.07 million POIs classified into eight categories, such as residential communities, government institutions, and healthcare facilities [89,90] (as shown in

Table 3. Classification of POI data.

Types of POIs	Baidu POI Primary Categories	Number	Proportion
Residential Areas	Residential	41,972	3.91%
Government Institutions	Government Institutions and Social Organizations	21,846	2.03%
Medical Facilities	Healthcare Services	22,928	2.13%
Commercial	Dining Services/Shopping Services/Financial and Insurance Services/Motorcycle Services/Automobile Services/Automobile Repair/Automobile Sales/Life Services/Sports and Leisure Services/Accommodation Services/Public Facilities/Scenic Spots	568,481	52.91%
Science, Education, and Culture	Science, Education, and Culture Services	40,824	3.80%
Transportation	Transportation Facilities Services	45,585	4.24%
Companies	Companies and Enterprises	154,618	14.39%
Others	Roadside Facilities/Geographical Names and Address Information/Events and Activities/Indoor Facilities/Passage Facilities/Null	178,206	16.59%

). POI density and diversity per unit street length were calculated to assess convenience [31,91]. Comfort indicators were extracted using DeepLabv3+, a state-of-the-art semantic segmentation architecture [31,87]. The model was pre-trained on the Cityscapes dataset, which is specifically designed for urban street scenes, achieving a mean Intersection over Union (mIoU) of 81.3% on the Cityscapes validation set. This ensured high accuracy in segmenting key visual elements such as vegetation, sky, and built structures in complex urban environments. Their proportions were computed and spatially analyzed using ArcGIS. The specific calculation formulas are shown in Table 2. Street classification was conducted using Shenzhen building data provided by the Shenzhen Urban Planning and Design Institute in 2016, where adjacent buildings along streets were categorized into five functional types: residential, office, school, commercial, and hospital.

To standardize the raw data for each indicator, normalization was applied to convert all values into dimensionless numbers. The Analytic Hierarchy Process (AHP) [92] was used to statistically determine the weights of each indicator factor. That is, three hierarchical analysis models were constructed from three dimensions—combined with the 1–9 scale method proposed by T.L. Saaty [93]—the factors in each dimension were compared two by two, a consistency test was performed on the judgment matrix, the geometric mean was taken as the judgment value of the matrix, and the weights of the index factors were derived comprehensively. Specifically, in the convenience-oriented street spatial hierarchical model, pairwise comparisons were conducted between POI category density and POI category diversity; in the comfort evaluation model, pairwise comparisons were made among street greenery view ratio, sky openness, and street enclosure degree. Twenty industry professionals were invited to conduct expert evaluations for this study. The final composite scores for convenience, comfort, and accessibility were obtained through weighted calculations (Table 2). Street usage frequency, comfort, convenience, and accessibility were categorized into six levels using a geometric interval classification method, where levels 1–3 represented relatively low values and levels 4–6 indicated higher values. This method is particularly effective for handling skewed or long-tailed distributions, as it focuses on the relative differences between data points rather than absolute intervals. This approach is especially useful for urban indicators, such as street quality or pedestrian flow, which often have uneven distributions [61,68]. For commuting street usage frequency, streets with values between 0 and 23.84 persons per meter per day were classified as low-frequency commuting streets, while those between 23.85 and 11,368.61 persons per meter per day were categorized as high-frequency pedestrian commuting streets. A high-frequency pedestrian commuting street was defined as a street where the average total pedestrian usage during the 6 h peak commuting periods exceeds 2385 persons per 100 m per day, with the highest recorded value reaching approximately 1.14 million persons per 100 m. These high-frequency pedestrian commuting streets accounted for 41.19% of all streets in Shenzhen, aligning well with real-world commuting patterns (see Figure 2).

3. Results

3.1. Spatial Distribution and Quality Characteristics of High-Frequency Pedestrian Commuting Streets in Shenzhen

3.1.1. Spatial Distribution Characteristics of High-Frequency Pedestrian Commuting Streets in Shenzhen

As shown in Figure 3, among the 108,963 streets in Shenzhen, 44,879 streets are identified as high-frequency pedestrian commuting streets, accounting for 41.19% of the total. Analyzing the distribution across different districts reveals that these streets tend to cluster in groups and are concentrated in areas with high building density. Specifically, the three districts with the highest total length of high-frequency pedestrian commuting streets are Bao’an, Longhua, and Longgang, which together account for 75.06% of the total. These are followed by Nanshan, Futian, and Luohu (21.12%), and Guangming, Yantian, and Pingshan, which have the lowest share (3.77%) (

Table 4. High-frequency pedestrian commuting street data for each administrative district in Shenzhen.

Administrative District	High-Frequency Pedestrian Commuting Streets (Streets)	Length of High-Frequency Pedestrian Commuting Streets (m)	Proportion of All High-Frequency Pedestrian Commuting Streets in Shenzhen (%)	High-Frequency Pedestrian Commuting Streets per Square Kilometer (Streets/km2)	Length of High-Frequency Pedestrian Commuting Streets per Square Kilometer (m/km2)	Length of Each High-Frequency Commuting Street per Square Kilometer (m/Street × km2)
Futian	3359	144,529.68	7.48%	42.63	1834	43.02
Luohu	2418	119,363.14	5.39%	30.86	1523	49.35
Nanshan	3703	145,729.32	8.25%	20.35	801	39.36
Bao’an	14,360	615,434.94	32.00%	36.05	1545	42.86
Longgang	7601	321,352.98	16.94%	19.60	829	42.30
Longhua	11,724	434,466.22	26.12%	66.77	2474	37.05
Guangming	1236	68,432.02	2.75%	7.95	440	55.35
Yantian	253	12,204.87	0.56%	3.50	169	48.29
Pingshan	208	6627.55	0.46%	1.25	40	32.00

). In Bao’an, Longhua, and Longgang, most high-frequency pedestrian commuting streets are concentrated within urban villages [94], such as Shajing and Konggang New Town in Bao’an District. A smaller proportion is found in administrative and commercial centers, such as Bao’an Central Business District, while some are scattered across principal arterials, minor arterials and collector roadway segments. Similarly, in Nanshan, Futian, and Luohu, many high-frequency pedestrian commuting streets are found in urban villages [94], such as Shawei and Xinzhou in Futian District. However, a considerable proportion is also located in administrative and commercial areas, including Nanshan Central District, Chegongmiao, and Huaqiangbei in Futian. Additionally, some of these streets appear in a dispersed pattern along principal arterials, minor arterials, and collector roadway segments.

This analysis suggests a strong correlation between the number of high-frequency pedestrian commuting streets and the presence of urban villages; districts with a higher number of urban villages tend to have a greater concentration of these streets. Contrary to common assumptions, the most frequently used pedestrian commuting streets are not concentrated in the city’s well-publicized core areas or along primary roads. Instead, they are densely clustered within urban villages, followed by administrative and commercial districts, and to a lesser extent, principal arterials, minor arterials, and collector roadway segments.

The spatial distribution of high-frequency pedestrian commuting streets in urban villages exhibits a dense grid-like or tree-branch pattern along collector roadway segments. In administrative and commercial centers, these streets tend to either branch along collector roadway segments or, in some cases, form densely packed grid-like structures. Meanwhile, the scattered distribution of high-frequency pedestrian commuting streets is primarily found along principal arterials, minor arterials, and collector roadway segments.

3.1.2. Comfort Evaluation and Spatial Distribution Characteristics of High-Frequency Pedestrian Commuting Streets in Shenzhen

The average comfort level of streets in Shenzhen is 0.26, with 47.0% (51,205 streets) classified as high-comfort streets. Among high-frequency pedestrian commuting streets, the average comfort level is slightly lower at 0.25, with only 31.8% (14,285 streets) meeting the high-comfort threshold—a proportion lower than the citywide average.

As shown in Figure 4, the spatial distribution of comfort levels in high-frequency pedestrian commuting streets exhibits regional differences, with higher comfort levels in the southern districts (southern Bao’an, Nanshan, Futian, and Luohu) and lower levels in the northern districts (northern Bao’an, Longhua, and Longgang). In the southern districts, high-comfort pedestrian commuting streets are concentrated in urban villages, administrative centers, and commercial hubs, while some are scattered throughout the area. Among them, high-comfort streets in urban villages are mainly distributed along surrounding city streets in a branch-like diffusion pattern, whereas the internal streets within urban villages tend to have low comfort levels—for example, the Xinzhou area in Futian District and Daxin area in Nanshan District.

In administrative and commercial centers, high-comfort streets follow a linear pattern along urban streets, while low-comfort streets are typically found in areas connected to subway stations, such as Bao’an Center Metro Station, Nanshan Metro Station, and Nanyou Metro Station. Within grid-like distributed administrative and commercial areas, the proportions of high- and low-comfort streets are relatively balanced, both exhibiting a branch-like distribution pattern along roads. Scattered high-frequency pedestrian commuting streets along major roads, secondary roads, and local streets predominantly exhibit higher comfort levels.

These findings indicate a strong correlation between comfort levels and street hierarchy. High-frequency pedestrian commuting streets along major transportation corridors tend to have better greenery coverage, which enhances comfort. This trend aligns with Shenzhen’s urban greening initiatives, including the integration of large urban green spaces, road greening networks, and green corridors as part of the Forest City development strategy [45]. These efforts have significantly improved the city’s overall green coverage and enhanced greenery along major motorized roads, reinforcing the city’s visual appeal from a vehicular perspective.

However, it is important to note that these high-comfort streets constitute only a small proportion of all high-frequency pedestrian commuting streets. A large number of internal streets used for daily high-frequency commuting—especially those within urban villages—suffer from low greenery levels due to the lack of systematic urban planning and greening efforts. This mismatch between urban greening distribution and pedestrian commuting demand highlights a significant gap in Shenzhen’s current street planning approach.

3.1.3. Evaluation of Convenience and Spatial Distribution Characteristics of High-Frequency Pedestrian Commuting Streets in Shenzhen

The average convenience level of streets in Shenzhen is 0.10, with 39.9% (43,479 streets) classified as high-convenience streets. Among high-frequency pedestrian commuting streets, the average convenience level is 0.08, and 36.8% (16,500 streets) are categorized as high-convenience streets, a proportion lower than the citywide average.

As shown in Figure 5, the spatial distribution of convenience levels among high-frequency pedestrian commuting streets exhibits regional variations, with higher convenience levels in the early-developed southern districts (southern Bao’an, Nanshan, Futian, and Luohu) and lower levels in the later-developed northern districts (northern Bao’an, Longhua, and Longgang). Across districts, high-convenience pedestrian commuting streets are mainly concentrated in urban villages, administrative centers, and commercial hubs.

In urban villages, those in southern Shenzhen (Nanshan, Futian, and Luohu) generally exhibit higher convenience levels, with a grid-like or branch-like distribution along urban streets. In contrast, urban villages in northern districts (Bao’an, Longhua, and Longgang) show greater variability in convenience levels, largely influenced by local functional land use patterns. Many urban villages in these areas feature a single-function land use structure, resulting in lower convenience levels for their high-frequency pedestrian commuting streets. These streets often exhibit a branch-like distribution, as seen in Xinhe Village and Fuhai West Urban Village.

In administrative and commercial centers, high-frequency pedestrian commuting streets that follow a grid-like distribution generally have high convenience levels. However, those exhibiting a branch-like or linear distribution tend to have high convenience levels only in certain segments, while others remain low. Additionally, scattered high-frequency pedestrian commuting streets located along major roadways generally exhibit low convenience levels.

These findings highlight the strong correlation between convenience levels and a street’s development period, functional land use, and road network layout. In early-developed areas with diverse functions and dense street networks, high-convenience pedestrian commuting streets are more prevalent. Conversely, newly developed districts and those along major transportation corridors tend to have lower convenience levels due to insufficient functional facilities and a lack of commercial and service establishments.

3.1.4. Evaluation of Accessibility and Spatial Distribution Characteristics of High-Frequency Pedestrian Commuting Streets in Shenzhen

The average pedestrian accessibility of streets in Shenzhen is 0.01, with 19.3% (21,065 streets) classified as high-accessibility streets. Among high-frequency pedestrian commuting streets, the average accessibility level is also 0.01, consistent with the citywide average. However, 28.5% (12,776 streets) of high-frequency pedestrian commuting streets exhibit high accessibility, a proportion higher than the citywide average.

As shown in Figure 6, overall pedestrian accessibility remains relatively low across high-frequency pedestrian commuting streets in Shenzhen. Analyzing their spatial distribution, we find that high-accessibility streets tend to be located in central clusters and densely populated areas, often appearing in a scattered pattern. These areas include the Luohu central area, administrative and commercial centers in Nanshan and Futian, and urban villages in Bao’an, as well as urban villages and some industrial areas in southern Longhua.

In urban villages, high-accessibility pedestrian commuting streets are distributed along surrounding urban streets in a branch-like pattern, while some are found within secondary roads inside urban villages. Their accessibility is closely linked to the density of the road network and street length.

In administrative and commercial centers, nearly half of the grid-patterned high-frequency pedestrian commuting streets exhibit high accessibility, as seen in Luohu central area. In contrast, for branch-like pedestrian commuting streets, only some exhibit high accessibility, primarily following a linear distribution along collector roadway segments. Scattered high-frequency pedestrian commuting streets, particularly those in southern Nanshan and Luohu, exhibit very limited accessibility.

3.1.5. Summary of Spatial Distribution and Quality Characteristics of HFPCSs

Our spatial distribution analysis of high-frequency pedestrian commuting streets in Shenzhen reveals that these streets are predominantly concentrated in the northern districts, including Bao’an, Longhua, and Longgang, which collectively account for 75.06% of the total. In contrast, the southern districts—Nanshan, Futian, and Luohu—contain a significantly smaller proportion, at only 21.12%. In the northern districts, most high-frequency pedestrian commuting streets are clustered within urban villages, whereas in the southern districts, these streets are more evenly distributed among urban villages, administrative centers, and commercial hubs. Specifically, high-frequency pedestrian commuting streets in urban villages exhibit grid-like or branch-like distribution patterns, while those in administrative and commercial centers are mainly branch-like, with some exhibiting a grid-like layout. Additionally, a subset of high-frequency pedestrian commuting streets is scattered across various urban areas.

From a street quality perspective, the three core indicators—comfort, convenience, and accessibility—show that the overall quality of high-frequency pedestrian commuting streets is either lower than or roughly equivalent to the citywide average. Compared to the citywide proportion of high-quality streets, high-comfort (31.8%) and high-convenience (36.8%) pedestrian commuting streets account for a lower share than the citywide averages (47.0% and 39.9%, respectively). In contrast, the proportion of high-accessibility streets (28.5%) is relatively higher than the citywide average (19.3%). These findings highlight a notable mismatch between comfort levels and high-frequency street usage, while streets with high accessibility are relatively few. In terms of spatial distribution, the southern districts generally have higher comfort and convenience levels compared to the northern districts. High-comfort pedestrian commuting streets are mainly located along collector roadway segments, while high-convenience pedestrian commuting streets are concentrated in urban villages. High-accessibility pedestrian commuting streets, on the other hand, are primarily located in urban clusters and densely populated areas, particularly within urban villages, which serve as the main hubs for these streets.

3.2. Integrated Analysis of High-Frequency Pedestrian Commuting Street Types and Quality Characteristics

By integrating the three quality evaluation indicators—convenience, comfort, and accessibility—high-frequency pedestrian commuting streets are classified into eight quality types, such as high convenience/high comfort/high accessibility (1,1,1) and high convenience/high comfort/low accessibility (1,1,0). As shown in Figure 7, streets that score high in all three quality dimensions account for only 2% of the total, closely aligning with the 1.53% proportion found in prior theoretical studies. Among the 44,879 high-frequency pedestrian commuting streets in Shenzhen, more than 75% are either streets with only one good quality attribute (49%) or streets with poor performance across all three dimensions (28%), making these the dominant characteristics of Shenzhen’s pedestrian commuting streets. In contrast, only 21% of these streets exhibit two good quality attributes, while just 2% achieve high quality across all three indicators.

Examining the spatial distribution of these streets (Figure 8) reveals that the southern districts (Nanshan, Futian, and Luohu) generally have better overall quality compared to the northern districts (Bao’an, Longhua, and Longgang). The southern districts contain a higher proportion of high-frequency pedestrian commuting streets with at least two quality attributes, whereas in the northern districts, most streets only meet one quality criterion or fail across all three dimensions.

These findings highlight a significant mismatch between different street quality attributes, accompanied by distinct regional disparities. Specifically, the three key quality dimensions—comfort, convenience, and accessibility—do not exhibit strong synergy, resulting in suboptimal coordination among them. In Shenzhen’s economically active southern districts, such as Nanshan, Futian, and Luohu, high-frequency pedestrian commuting streets generally perform better in terms of quality metrics. In contrast, the northern districts (Bao’an, Longhua, and Longgang), which accommodate 75.06% of the city’s pedestrian commuting traffic, suffer from notably lower levels of comfort, convenience, and accessibility, forming distinct “quality gaps” in pedestrian infrastructure.

3.3. Dominant Street Types and Quality Analysis of Key Street Categories in Shenzhen

Building on the citywide analysis, this study further refines the classification and statistical analysis of street functional compositions to identify key street types that have the most significant impact on urban pedestrian experiences. Based on the classification framework in Section 2.3, the study examines the combinations of five core building functions adjacent to both sides of streets across Shenzhen. This analysis identifies 31 distinct street types with valid data, among which residential streets account for the largest share, followed by mixed residential–office streets and office streets, as illustrated in Figure 9. These findings align with existing research and data, which indicate that residential land use occupies the largest proportion of urban construction land (25.0–40.0%) [45]. A deeper review of the urban land use literature suggests that residential areas in Shenzhen can be further categorized into four subtypes [95]. Among them, Category II residential land (new residential communities)—dominated by mid-rise and high-rise housing developments—has experienced rapid expansion in recent years, significantly increasing its share. Simultaneously, urban villages (Category IV residential land), which are densely populated areas with privately built housing, continue to constitute a significant portion of Shenzhen’s residential land use [95,96]. Given the prominence of these two residential typologies, this study prioritizes the quality assessment of high-frequency pedestrian commuting streets within both new residential communities and urban villages, where new residential communities refer to residential zones uniformly developed by real estate developers on state-owned construction land [97], and urban villages denote residential areas located on original rural collectively owned land that has not been fully transformed during the urbanization process and still retains its existing state [98].

For key street type analysis, this study focuses on Futian District, the administrative and financial hub of Shenzhen, which features a well-developed and diverse residential landscape. The study selects 9 newly developed residential communities and 23 urban villages in Futian to further examine the quality characteristics of high-frequency pedestrian commuting streets in residential-dominated areas. Additionally, to capture the unique structure of urban village road networks and pedestrian activity patterns, the study incorporates three high-density urban villages from other districts as supplementary cases, allowing for a comparative discussion of different urban village street environments.

3.3.1. Type Analysis

As shown in

Table 5. The number, length, and proportion of three types of high-frequency commuting street areas in Futian District, Shenzhen (0 indicates the absence of an item and 1 indicates the presence of an item).

	Residential Quarters				Urban Villages
Category (Convenience/Comfort/Accessibility)	Quantity	Quantity Ratio	Length	Length Ratio	Quantity	Quantity Ratio	Length	Length Ratio
0,0,0	42	15.1%	1507.4	11.1%	173	22.1%	7277.0	13.4%
0,0,1	14	4.4%	598.1	3.5%	19	2.4%	599.8	1.1%
0,1,0	91	28.4%	3114.4	18.9%	61	7.8%	3105.6	5.7%
0,1,1	22	6.9%	649.9	4.7%	35	4.5%	702.0	1.3%
1,0,0	32	10.0%	2411.2	12.5%	325	41.5%	29,035.3	53.5%
1,0,1	7	2.2%	438.8	3.6%	28	3.6%	2115.6	3.9%
1,1,0	62	19.4%	4416.4	36.2%	116	14.8%	9449.8	17.4%
1,1,1	9	2.8%	562.1	9.6%	27	3.4%	1960.8	3.6%
Total number	279	87.2%	13,698.2	100.0%	784	100.0%	54,245.9	100.0%

, residential areas in Futian District exhibit distinct patterns in pedestrian commuting street usage. In new residential community clusters, there are 485 streets, of which 279 (57.5%) are classified as high-frequency pedestrian commuting streets. In contrast, in urban village clusters, there are 859 streets, with 784 (91.3%) identified as high-frequency pedestrian commuting streets.

These findings indicate that residential areas generally exhibit a higher proportion of pedestrian commuting streets compared to the citywide average (41.19%). Specifically, nearly 60% of streets in new residential communities are classified as high-frequency pedestrian commuting streets, demonstrating relatively high pedestrian activity. However, in urban villages, over 90% of streets fall into this category, highlighting an exceptionally high level of street usage in these areas. This suggests that urban villages play a particularly crucial role in supporting pedestrian commuting activity, far exceeding the levels observed in newly developed residential communities.

3.3.2. Quality Analysis of Key Street Types

(a)

Newly Built Communities

Among the 279 high-frequency pedestrian commuting streets in new residential communities, comfort is the most dominant quality, with 57.5% of the streets classified as high-comfort. This is followed by convenience (34.4%) and accessibility (16.3%). In the integrated quality analysis, the most common type of street is “high comfort only” (0,1,0), accounting for 28.4%, followed by “high convenience and high comfort” (1,1,0) at 19.4%, and “low quality across all three dimensions” (0,0,0) at 15.1%. Only 2.8% of high-frequency pedestrian commuting streets in new residential communities achieve high quality across all three dimensions. These findings indicate that high comfort is the dominant characteristic of pedestrian commuting streets in new residential communities, while convenience and accessibility are relatively weaker.

Examining the spatial distribution of street quality within new residential communities, we find that high-comfort pedestrian commuting streets are mainly located along internal roads and surrounding collector roadway segments. High-convenience pedestrian commuting streets are concentrated along roads with clustered commercial establishments, while high-accessibility streets are primarily found near community entrances and central internal roads. These patterns align with the planned development and infrastructure design of new residential communities. The unified landscape and green space planning in these communities enhances comfort, while commercial infrastructure is often concentrated in specific areas, and the overall road network density remains relatively low.

Further analysis reveals that only a very small proportion of high-frequency pedestrian commuting streets in new residential communities exhibit high quality across all three dimensions, while the proportion of streets with two good quality attributes is also limited. This highlights a significant mismatch among convenience, comfort, and accessibility, indicating that, while new residential communities excel in providing comfortable pedestrian environments, they lack a balanced integration of accessibility and convenience (Figure 10).

• (b) Urban Villages

Among the 784 high-frequency pedestrian commuting streets in urban villages, convenience is the dominant characteristic, with 63.3% classified as high-convenience streets. This is followed by comfort (30.5%) and accessibility (13.9%). In the integrated quality analysis, the most common type of street is “high convenience only” (1,0,0), accounting for 41.5%, followed by “low quality across all three dimensions” (0,0,0) at 22.1%, and “high convenience and high comfort” (1,1,0) at 14.8%. Streets that score high in all three quality dimensions (1,1,1) are extremely rare, making up only 3.4% of the total. These results indicate that urban village pedestrian commuting streets are primarily characterized by high convenience, while comfort and accessibility remain relatively weak.

The accessibility level of urban village streets is partially influenced by the availability of street network density data in different regions. Taking urban villages in Futian and Bao’an Districts as examples, Futian’s urban villages have extremely high building densities, with building gaps as narrow as 1 to 3 m. Due to the narrow width of these roads, many are not recorded in official street network datasets, leading to an artificially lower calculated street network density in these areas. In contrast, urban villages in Bao’an District tend to have a more loosely arranged building layout, with building gaps of 5 m or more, making it easier for streets to be recognized and recorded in datasets, resulting in higher measured street network density. This discrepancy suggests that northern urban villages may appear to have higher pedestrian accessibility based on calculations, but the actual accessibility conditions still require further validation.

Analyzing the spatial distribution of high-frequency pedestrian commuting street quality in urban villages reveals distinct patterns across different quality dimensions. High-convenience streets are primarily located within the internal grid-like core roads of urban villages and along the peripheral roads adjacent to urban villages. In contrast, comfort levels in high-frequency pedestrian commuting streets within urban villages are generally low. High-comfort streets are mostly found on the outskirts of urban villages, adjacent to secondary urban roads with better greenery, while they are less common within the urban villages themselves. Comparing different districts, urban villages in Futian District exhibit lower overall accessibility compared to those in Bao’an District and other northern areas. High-accessibility pedestrian commuting streets in urban villages are primarily found along roads that directly connect to collector roadway segments or at intersections with urban streets. This variation suggests that accessibility is influenced by differences in urban village street network density and the integration of village streets with the broader city road system.

From the perspective of commercial facility distribution, high-frequency pedestrian commuting streets along the core internal roads of urban villages and those adjacent to urban streets generally feature dense clusters of commercial establishments, resulting in higher convenience levels. However, significant variation exists in high-frequency pedestrian commuting streets adjacent to tertiary roads. Specifically, streets near gated new residential communities or inactive industrial zones tend to have fewer commercial establishments, leading to lower convenience levels. The high convenience levels in urban villages are closely tied to their bottom-up, informal development patterns, which have organically created dense networks of commercial services catering to daily pedestrian activity. However, due to the lack of unified planning and management, the overall comfort levels of high-frequency pedestrian commuting streets in urban villages remain relatively low (see Figure 11).

3.3.3. Summary of Type Analysis

In summary, high-frequency pedestrian commuting streets in new residential communities exhibit relatively high comfort levels, while their convenience and accessibility levels remain low. In contrast, high-frequency pedestrian commuting streets in urban villages show high convenience levels, but lower comfort and accessibility levels. This indicates significant differences in the quality characteristics of high-frequency pedestrian commuting streets across different urban environments. For new residential communities, their building layout and landscape infrastructure are typically planned in a top-down manner, with well-developed public systems and relatively high green coverage. However, the distribution of commercial facilities is constrained by zoning regulations, often leading to clustered rather than linearly distributed commercial establishments along adjacent streets, resulting in lower convenience levels for pedestrian commuting streets. Additionally, gated community management further reduces overall accessibility, making pedestrian connectivity highly insufficient. In contrast, urban villages have developed organically through bottom-up processes, leading to a stronger alignment between function and pedestrian movement patterns. As a result, high-frequency pedestrian commuting streets in urban villages tend to offer high convenience levels. Moreover, their open street networks and high-density layouts enhance accessibility. However, due to the lack of coordinated planning and the intensive use of land, urban villages often experience high building density, insufficient public infrastructure, and lower comfort levels. Therefore, for high-frequency pedestrian commuting streets in new residential communities, attention should be given to improving commercial facilities along pedestrian routes and increasing pedestrian network density to enhance walkability and accessibility for residents. Meanwhile, for high-frequency pedestrian commuting streets in urban villages, interventions should focus on strengthening public space systems, integrating greenery where possible, improving streetscape quality, and optimizing the overall pedestrian environment to enhance walking comfort for urban village residents.

4. Discussion

4.1. Summary

This study integrates LBS positioning data to identify pedestrian behavior and extract high-frequency walking commuter streets, and combines POIs, street view imagery, and other multi-source datasets to construct a comprehensive street quality evaluation system. The analysis reveals a significant mismatch between the spatial distribution and quality attributes of high-frequency walking commuter streets in Shenzhen, presenting a structural contradiction of “high-frequency but low-quality”.

First, there is a spatial mismatch between supply and demand for high-frequency walking commuter streets: There is a clear imbalance in quality provision across high-frequency walking commuter streets between the northern and southern regions. While the northern region accounts for 75% of such streets, their scores in environmental comfort, functional convenience, and transportation accessibility remain low. In contrast, streets in the southern region perform better across all quality dimensions but only represent 21% of high-frequency commuter streets. This reveals a typical pattern of spatial supply imbalance: “high proportion–low quality” in the north versus “low proportion–high quality” in the south.

Second, there are systemic deficiencies in the multi-dimensional quality of high-frequency walking commuter streets: High-frequency commuting streets exhibit systemic weaknesses across multiple quality dimensions. The three core indicators—comfort (32%), convenience (36.8%), and accessibility (29%)—all have compliance rates below 40%, indicating widespread underperformance.

• A mismatch in comfort-related supply and demand: High-frequency commuter streets are sparsely distributed in areas with concentrated green infrastructure, such as major arterial roads. In contrast, urban villages, where high-frequency streets are densely located, tend to have high building density and limited public space, resulting in generally low levels of green coverage and a mismatch between resource allocation and pedestrian demand.
• Convenience is constrained by development stage and planning mechanisms: Older, well-developed areas with high functional diversity (e.g., Luohu old town) perform relatively well in terms of convenience due to well-established street-level commercial amenities. In contrast, high-frequency walking streets located in newly developed districts, mono-functional industrial zones, and urban villages, as well as those scattered along arterial roads, face significant deficits in convenience due to the fragmented distribution of supporting facilities.
• Accessibility is restricted by both street network density and connectivity: Many high-frequency walking commuter streets in urban villages with poor connections to the main road network, as well as in areas with sparse road density, suffer from low accessibility. Although core urban areas and densely populated districts exhibit locally high accessibility due to dense street networks, the lack of effective pedestrian linkages with surrounding zones leads to an “island-like” clustering effect. As a result, the overall walkability of the street network remains insufficient.

Third, there is a severe imbalance in the coordination of quality elements on high-frequency walking commuter streets: Our results show that only 2% of high-frequency walking commuter streets across the city perform well in all three core dimensions. In contrast, 75% fall into the categories of “only one dimension performs well” (49%) or “poor performance in all three dimensions” (28%), highlighting a significant misalignment between street usage frequency and quality provision. This result reveals a structural disconnect between the spatial provision of high-quality streets in current planning practices and the actual demand for high-frequency usage, underscoring the urgent need to restructure quality supply logic through demand-oriented, integrative renewal strategies.

Lastly, high-frequency walking commuter streets vary significantly across different residential typologies: Typological analysis reveals notable quality differences between high-frequency walking commuter streets in two distinct residential forms. In planned gated communities, streets generally exhibit higher comfort levels (58%) due to well-designed green spaces. However, concentrated commercial layouts and enclosed management models result in lower performance in convenience and accessibility. In contrast, urban villages, which have evolved through self-organized development, perform better in terms of convenience (63%) due to their organically formed commercial and functional amenities. Nevertheless, high building density and insufficient public infrastructure lead to poor comfort levels.

Finally, in comparison with existing studies [22,23], this study further highlights both shared findings and novel contributions. In terms of its findings, the results of this study are consistent with prior research in the following aspects:

• There is a significant challenge in the coordinated optimization of high-frequency pedestrian commuting streets. In all three studies, the proportion of streets that are both high-frequency and high-quality remains below 10%, while more than 50% of high-frequency streets exhibit either partial quality performance or poor performance across all dimensions.
• Each study evaluates street-level functional facilities and environmental indicators, and all reveal similar spatial mismatches between usage demand and quality provision.

Compared to the existing literature [22,23], this study contributes in several ways:

• It expands the analysis of accessibility, revealing that even high-accessibility streets tend to exhibit fragmented and island-like distribution patterns.
• It identifies a clear supply–demand mismatch between the spatial distribution of high-frequency streets and the provision of street quality.
• It quantitatively demonstrates that residential streets are the dominant street type in Shenzhen and reveals systematic differences in quality between planned residential areas (gated communities) and self-organized areas (urban villages).
• Methodologically, this study develops a four-stage progressive optimization framework based on a “demand-tiered and type-responsive” approach, establishing a research-to-practice closed-loop decision chain for street renewal. This provides an actionable and transferable pathway for fine-grained urban street optimization.

4.2. The Four-Stage Progressive Optimization Framework: A “Demand-Tiered and Type-Responsive” Strategy for Coordinated Street Research and Renewal

Based on the results of the multi-source data analysis and empirical validation, this study proposes a “demand-tiered and type-responsive” renewal strategy to guide both research into and the practical implementation of the optimization of urban streets. We establish a four-stage progressive optimization framework to facilitate evidence-based and context-sensitive interventions:

• Precise Diagnosis of Street Usage Frequency: A data-driven identification model is developed using fused multi-source spatiotemporal data, incorporating urban commuting behavior patterns to detect and select high-frequency pedestrian streets with high accuracy.
• Multi-Dimensional Quality Assessment: A comprehensive and quantitative evaluation system is constructed, encompassing physical environmental conditions, functional facilities, and network accessibility, thereby enabling the systematic assessment of street quality across multiple dimensions.
• Overlay Analysis for Identifying High-Frequency but Low-Quality Streets: By overlaying usage frequency metrics with quality evaluation results, streets with high pedestrian flow but significantly lower-than-average quality are identified as priority targets for intervention.
• Category-Specific Responsive Implementation: Streets are classified based on their functional roles and regional characteristics. Tailored optimization strategies are then matched to corresponding street typologies, ensuring that renewal efforts are aligned with local needs and typological patterns.

This framework establishes a closed-loop system of “data diagnosis–demand tiering–type-specific response–targeted implementation”, breaking away from the traditional one-size-fits-all model of street renewal. It introduces a quantifiable and dynamic, human-centered street classification mechanism based on usage frequency, coupled with typology-specific optimization strategies. The framework offers an end-to-end decision-making support system, from macro-level identification to micro-level intervention, enabling the coordinated improvement of both functional performance and environmental quality for high-frequency pedestrian streets.

4.3. Optimization Strategies for High-Frequency Commuting Streets in Shenzhen

4.3.1. Citywide Optimization Strategies

(a)

Comfort Optimization: From Motor Vehicle Orientation to a Human-Centered Perspective

Research findings indicate that a large number of high-comfort streets are currently located along major vehicular roads and suburban green areas (such as Shennan Avenue and Tanglangshan Greenway). However, high-frequency commuting primarily occurs within internal urban spaces, including alleyways in urban villages, metro transfer zones, administrative centers, and commercial hubs. This mismatch stems from Shenzhen’s strong emphasis on greening major arterial roads. By reallocating greening resources toward high-frequency pedestrian areas, the structural contradiction between landscape investment and commuting demand can be resolved.

(b)

Convenience Optimization: From Coarse-Grained Provision to Targeted Matching

Results show a significant north–south disparity in convenience. Southern areas, developed earlier, have organically formed grid-like commercial ecosystems within urban villages, while the newer northern districts exhibit linear and fragmented patterns due to their top-down planning. Additionally, many high-frequency commuting streets along major roads suffer from excessive building setbacks, creating “service vacuum zones” disconnected from surrounding urban functions.

Therefore, the newer northern districts should focus on targeted supplementation. For areas like northern Bao’an and Longgang University Town, the implementation of “300-meter radius community service zone” is recommended, prioritizing the provision of basic service facilities such as convenience stores, grocery shops, and pharmacies. At the same time, a government–enterprise data-sharing platform should be established to dynamically match enterprise commuting heatmaps with facility layout, improving allocation accuracy.

For arterial roads, commercial stitching strategies should be applied. Along roads such as Shennan Avenue and Bao’an Avenue, the “linear service belt + vitality node” model is proposed: clustering convenience stores and breakfast vendors around metro exits and peak commuting hotspots to create vitality nodes. Dynamic monitoring mechanisms based on POI heatmaps should be established to optimize commercial mix and spatial layout in real time. Improving vitality along major roads requires breaking the car-centric mindset and transitioning toward human-centered spatiotemporal precision interventions.

(c)

Accessibility Optimization: From Road Network Expansion to Precision Weaving

Our findings indicate that the average walkability value of high-frequency commuting streets in Shenzhen is only 0.01, consistent with the citywide average. However, significant spatial variation exists within this group: only 28.5% of high-frequency commuting streets (12,776 segments) have high accessibility, slightly higher than the city average of 19.3%, and these are mostly scattered around cluster centers and densely populated areas. This “locally efficient but structurally inadequate” pattern leads to mismatches between commuting demand and pedestrian network capacity. Therefore, a precision weaving approach is recommended to guide spatial repair in a tiered manner:

• In urban village areas, enhance lateral connectivity through micro-street infill to stitch branch-like street networks to the urban arterial system.
• In scattered areas, utilize metro station catchment zones to connect fragmented pedestrian paths, gradually achieving citywide pedestrian network upgrades through regional-scale integration.

4.3.2. Targeted Optimization Strategies for Key Street Types

Typological analysis reveals significant differences in the quality characteristics of high-frequency walking commuter streets under two distinct residential forms. Accordingly, differentiated optimization strategies are proposed:

• Optimization Strategy for High-Frequency Walking Commuter Streets in Planned Gated Communities. To address the spatial contradiction of “high comfort–low convenience–low accessibility” commonly found in planned residential developments, this study proposes an integrated “network–function–system” optimization framework:
  • Topological Optimization of the Pedestrian Network: From Isolation to Integration. High-frequency commuting streets serve as catalysts for restructuring the topological relationship between enclosed residential compounds and the broader urban network:
    • Prioritize pedestrian linkages to high-demand destinations such as metro stations and schools to improve overall network connectivity.
    • Activate the peripheral micro-scale street network of residential compounds through corridor-based expansion, forming a hierarchical pedestrian topology consisting of “high-frequency main corridors and permeable side branches”, thereby resolving the asymmetrical disconnection between enclosed communities and surrounding urban space.
  • Functional Gradient Activation of Streets: From Segregation to Symbiosis. Using high-frequency commuting streets as leverage, this strategy aims to reconfigure the functional value of street spaces:
    • Boundary Reproduction: Transform enclosed street edges into linear service interfaces by utilizing setback spaces along compound walls to introduce flexible commercial belts. Through the linear distribution of micro-scale commercial units, functional permeability is achieved along previously closed boundaries.
    • Functional Gradient Activation: Embed small-scale commercial and social facilities along pedestrian corridors in a hierarchical manner. This includes a combination of basic convenience functions (e.g., convenience stores, breakfast kiosks) and vitality-enhancing amenities (e.g., shared book kiosks, flower shops). Such a gradient in spatial layout reconstructs both consumer engagement and social attraction, contributing to the creation of composite experiential street systems that integrate “commuting corridors” with “everyday living networks”.
• Optimization Strategy for High-Frequency Walking Commuter Streets in Self-Organized Urban Villages. To address the spatial contradictions of “high-convenience but low-comfort” and “internal efficiency but external disconnection” in urban village streets, this study proposes a three-pronged strategy:
  • Low-Intervention Environmental Upgrading. For key high-frequency commuter streets within urban villages, adopt incremental improvements through micro-scale interventions (e.g., shading, paving enhancements) to improve thermal comfort while preserving the existing advantages of functional diversity. This approach avoids disrupting the inherent spatial resilience formed through self-organized development.
  • Construction of Adaptive Public Space Networks. Implement infill-based renewal along key high-frequency streets in urban villages by embedding small-scale public facilities into the existing high-density built environment. This aims to form a public space system that aligns with residents’ behavioral patterns and everyday spatial practices.
  • The Coordinated Governance of the Pedestrian System. Strengthen the connectivity between key high-frequency streets in urban villages and the city’s main road network. Focus on addressing the internal–external disconnection along village boundaries to enhance local convenience while promoting synergy with citywide accessibility.

By adopting differentiated and context-specific strategies for planned residential compounds and urban villages, this study emphasizes the role of high-frequency commuting streets as an entry point for street renewal. Ultimately, it promotes the transformation of streets from mere traffic corridors into walkable, consumable, and sociable public spaces for everyday urban life.

4.4. Limitation

This study is subject to certain limitations, primarily due to constraints in data availability, granularity, and research scope, which may impact the accuracy of the findings. The data used in this study were collected under normal weather conditions, excluding extreme or irregular weather scenarios that could influence pedestrian behavior. Additionally, while big data provides broad coverage, it does not account for individuals without mobile devices, which may result in an underestimation of pedestrian movement patterns. Furthermore, the study does not capture the specific reasons behind pedestrian stops or movement choices, making it necessary to integrate more micro-scale field research to gain deeper insights into pedestrian behavior.

Future research should aim to incorporate a wider range of data sources, including field observations and pedestrian movement surveys, to refine the analysis and develop more precise strategies for optimizing high-frequency pedestrian commuting streets. A more comprehensive understanding of pedestrian behavior, particularly under varying environmental and socio-economic conditions, will provide stronger empirical support for urban street improvement initiatives and policy recommendations.

5. Conclusions

In conclusion, this study proposed a data-driven screening framework for identifying high-frequency pedestrian commuting streets (HFPCSs) by integrating multiple sources of urban big data and conducted both comprehensive and typology-specific assessments of street quality. The findings reveal a dual mismatch between street quality and usage frequency in Shenzhen—both in terms of spatial distribution and supply–demand alignment—highlighting the structural dilemma of “high-frequency but low-quality”. Specifically, only 2.15% of HFPCSs in Shenzhen exhibit good performance across all three dimensions: comfort, convenience, and accessibility. In contrast, over 70% of HFPCSs are located in the northern districts, where the overall quality tends to be lower, reflecting a pronounced spatial imbalance between pedestrian demand and environmental supply. Multi-dimensional analysis further indicates that comfort is constrained by the mismatch in greening resources, convenience is shaped by development stage and functional diversity, and accessibility is limited by inadequate road network connectivity and density, leading to a fragmented, “island-like” street system. The study also finds systematic differences between planned gated communities and self-organized urban villages, underscoring the profound influence of residential typology on street space production.

In response, this study introduces a four-stage progressive optimization framework based on a “demand-tiered and type-responsive” strategy. By precisely identifying HFPCSs with low quality and matching them with appropriate intervention strategies, the approach facilitates a transition from homogeneous upgrades to human-centered precision interventions. At the citywide level, these optimizations focus on reallocating greening resources, improving the spatial match of commercial functions, and weaving together fragmented pedestrian networks. At the typology-specific level, targeted renewal strategies are proposed to address the contrasting problems faced by gated communities and urban villages, respectively, advocating for “network–function–system” integrated improvements and low-intervention, incremental upgrades. These strategies not only offer actionable decision support for Shenzhen’s street renewal efforts but also establish a closed-loop “data diagnosis–policy response” model that provides both a theoretical framework and technical paradigm for addressing the quality–usage paradox in high-density cities worldwide. Future research may further explore dynamic monitoring systems and multi-stakeholder coordination mechanisms, enabling a shift from static planning to adaptive governance and supporting the sustainable transformation of urban public space.