Priority Intervention Evaluation of Community Regeneration in Megacities Based on the Business Improvement District (BID) Model: A Case Study of Tianjin, China

Abstract

This study aims to start from the macroscale of the megacity, explore a top-down operational technical path from the identification of community regeneration units as the evaluation objects to the BID priority intervention evaluation, and then propose differentiated community regeneration strategies based on the BID model. In the post-epidemic era, it is necessary for global megacities to take new measures in urban regeneration to respond to worldwide changes and challenges. As an innovative tool to promote central city revitalization, the BID model has played an important role in community transformation. In the context of the continuous decentralization of population and industry in China’s megacities, it is urgent to explore the technical path to apply the BID model to local community regeneration. Given the shortcomings of existing studies in the method to identify the scope of BID implementation and evaluate intervention priorities, this study takes Tianjin, a megacity in China, as an example and uses DBSCAN (density-based spatial clustering of applications with noise) and service area analysis to define the community regeneration units with commercial agglomeration characteristics as the objects. Then, the BID priority intervention evaluation system is constructed from the two measurement aspects of the intervention potential and the necessity of community regeneration to classify the BID intervention priorities of community regeneration units. The main conclusions are as follows: 1. When the DBSCAN analysis parameters take the minimum number of elements as 30 and the search distance as 120 m, the result is most suitable for identifying community units with commercial agglomeration of the study area; 2. Population vitality, especially working and residential population density, is the key factor affecting BID intervention potential, while road network density is an important indicator for determining the necessity of community regeneration; 3. Community regeneration units with high BID priority levels need to develop differentiated regeneration strategies combining their BID intervention potential, regeneration necessity characteristics, and location attributes. These conclusions can provide references for the governments of megacities to screen and establish BIDs.

1. Introduction

Business improvement district (BID), as an interactive regeneration model between business and community, plays an important role in preventing the decline of downtown and promoting community development [1,2]. Especially as the COVID-19 pandemic forces the world to adapt in countless ways and businesses and communities face unprecedented challenges, Western cities such as New York and London have built networks of innovative partnerships through BID and are on the road to rapid economic recovery [3]. The BID model is being transferred to more and more countries or regions, and so far, the number of establishments in the world has reached thousands [4], which highlights business as a new force in urban regeneration [5]. Specifically, it is gradually recognized and practiced to regard retailing as a pillar rather than a subsidiary role in the urban regeneration process [6], and BID is precisely the most effective retail business-led urban regeneration method [7]. In the face of realistic pressure from decentralization and digitalization to disrupt the vitality of traditional business ecosystems [8], while BID is controversial in terms of social justice and equity [9,10], it has been shown to have a positive effect on enhancing commercial real estate values, site attractiveness, urban safety resilience, etc. [11]. China has to gradually face the problem of population decentralization that many developed countries have experienced in the middle and later stages of urbanization. Data from the National Bureau of Statistics show that between 2010 and 2020, the population of the central areas of most megacities has decreased significantly. In this context, the focus of China’s urban regeneration has gradually shifted toward the regeneration of community commerce and other service facilities to boost economic vitality [12]. China’s megacities generally face a large number of older community regeneration needs, shortage of funds, difficulties for businesses to save themselves, etc. In view of the fact that BID has successfully carried out international circulation in different social space and political system environments [13], this model undoubtedly could provide an innovative operational framework for Chinese communities to solve financing difficulties and achieve sustainable operations [14].

Through the authority and constraints of the government in this specific area named BID, commercial owners, the public sector, residents, and non-profit institutions collaborate to transform the environment and reform governance, attracting flowing capital and investment, promoting economic development, and revitalizing declining communities [15]. In the establishment process of most BIDs, their initiations are determined by the local council, which petitions and defines the area within a certain boundary that requires policy intervention, even if the BID requires support from the government [16,17]. A growing body of research suggests that BID model development is not universally applicable to every community, even in the same city [5,18]. Established ways of setting up make it difficult for economically distressed areas and communities with large immigrant populations to operate BID, and wealthy communities may not be those most in need of development and revitalization [19]. BID should target areas where there is already a strong commercial base [20]. In other words, it can fully play its role in communities with commercial agglomeration characteristics [2]. Although the scale of BID does not stipulate the correct size [17], relevant practical evaluation studies point out that during the initial establishment, scientifically delineating the scope of BID has an important impact on ensuring the effective allocation of resources and investment and promoting urban sustainable revitalization [21,22]. However, most of the existing BID-related research consists of compilations and implementation effect evaluations of practical cases [23,24]. There is a lack of study on scientifically identifying the objects before BID implementation and BID priority intervention evaluation. At the same time, there is also a lack of theoretical and practical exploration of China’s localized BID intervention.

Therefore, given the non-universality and common initiation method of BID, and the benefits of accurately identifying the initiating object and its scope mentioned above, it is of great significance to address the issue of how to macroscopically determine what the scope of communities in megacity have more potential and necessity for BID intervention, as well as what targeted regeneration strategies should be involved in the guidance for these communities. To fill the research gaps, this study started at the macro-scale of the megacity, explored a top-down operational technical path from the identification of community regeneration unit objects to the BID priority intervention evaluation, and then proposed differentiated community regeneration strategies based on the BID model. Specifically, this study takes Tianjin as the study area. Based on the residential-commercial characteristics of the BID model, this paper uses DBSCAN (density-based spatial clustering of applications with noise) analysis and service area analysis, combined with the age characteristic of residential construction, to establish a top-down method to identify the community regeneration units as the objects of priority intervention evaluation. Then, two index systems are established to measure the BID intervention potential and the community regeneration necessity of the designated units. Finally, according to the comprehensive comparison of results from two measurements, the priorities of embedding the BID model into different units are classified, and differentiated community regeneration strategies are proposed.

The subsequent sections of this paper are organized as follows: Section 2 provides a literature review to support the selection of the research area, the identification methods of evaluation objects within the area, and the construction of the BID priority intervention evaluation system; Section 3 describes the method path of using multi-source big data to identify research objects within the research area and conducting BID priority intervention evaluation through measurements of intervention potential and regeneration necessity. Section 4 illustrates the distribution of research objects obtained from the optimal results of DBSCAN analysis, and then classifies these research objects according to BID priority intervention based on two aspects of measurement. The key influencing factors affecting the priority intervention level of different types of community regeneration units are analyzed; Section 5 offers an in-depth discussion of the computational outcomes and identifies the study’s significance and limitations; Section 6 concludes the research and projects potential future trajectories for the field.

2. Literature Review

2.1. Research on the Scope of BID Intervention Object

BIDs are place-based private–public partnerships created for the purpose of enhancing local economies through the provision of member-defined services and infrastructure [25,26]. With the changing paradigm of the government’s functions and the increasing private sector participation in urban regeneration, the delineation method of BID has become increasingly flexible [27]. The scope of community regeneration projects largely depends on the initiator and initiation process [6]. The unilateral dominance of public power demonstrates their limitations on the ability to scope their boundaries in the initiation process of sustainable community regeneration [28]. Top-down community regeneration concepts, such as the complete community [29] and the integrated community [30], led by the government, promote community regeneration projects based on the service radius of the concentrated and standardized living service facilities. In addition, there are some community regeneration projects that fall within the scope of community administrative jurisdiction. A study by Rabbiosi, C. shows that it is more suitable for sustainable community development to obtain the scope of regeneration based on the conditions of the community itself [31], instead of rough interference that is detached from the actual situation of the community. Stratton-Rayner, J.M. explained that BID should determine its scope of action and organizational structure based on community context, through the analysis of two BID cases in Philadelphia, US [32]. Rezart Prifti and Fatma Jaupi conducted a study on the scope of BID initiation by manually identifying areas of commercial agglomeration on community maps [2]. Therefore, from the perspective of the scope of BID intervention, the research on the demarcation of BID mainly consists of two aspects at the microscale. One is about the commercial gathering judgment before the implementation of BID in a specific community. Another is also targeted at specific cases, but it is the experience summary of the scope after the implementation of BID in a specific community for a period of time. There is a lack of research exploring the priority intervention objects of BID from a macroscale perspective.

2.2. Research on DBSCAN Method for Commercial Agglomeration

DBSCAN, as a density-based spatial clustering method, is widely used on large amounts of positional data [33]. Many studies have shown that DBSCAN is an effective tool to identify areas with a fixed minimum-scale commercial agglomeration. DBSCAN differs from other common density-based spatial clustering methods such as HDBSCAN and OPTICS in that the main variables are the minimum number of elements (N) and search distance (D). Methods such as HDBSCAN and OPTICS, due to their adaptive variable distance [34] and multi-scale density variables [35], may have uncontrollable results when identifying community businesses with a determined minimum number of shops and service radius. The research conducted by Sheng Q. in Beijing indicated that there is a general consensus among the public regarding the minimum number of shops and service radius in community commercial vitality centers in their respective cities [36], proving the effectiveness of DBSCAN in identifying business clusters. Michalis Pavlis et al. applied the DBSCAN method for the identification of local retail agglomerations within London [37]. There are studies comparing other data sources to the DBSCAN results. A study on China combines DBSCAN with comments data from the Dianping app to explain the formation mechanism of catering space clusters, while also demonstrating the objectivity and effectiveness of comments data in the Dianping app [38]. In addition, a study by T. Luo et al. comparing the K-means and DBSCAN clustering methods for urban functional POIs showed that the clustering shape of DBSCAN can be more concave and convex [39]. Other clustering methods such as K-means, spectral clustering, mean shift clustering, etc., based on their computational principles, tend to form spherical distribution results, which have limitations in the clustering analysis of commercial POIs that are mainly irregularly and densely distributed along urban streets. However, existing research on DBSCAN for identifying business clusters is based on determined N-D parameters and lacks research on how the parameters themselves are valued.

2.3. Theory Fundations of BID Intervention Potential and Necessity

The content of the BID intervention potential measurement is affected by the agglomeration effect theory, business district theory, and consumer behavior theory. Measurement of BID intervention potential could be constructed from the dimensions of commercial compactness, population vitality, and spending power. The agglomeration effect theory holds that commercial agglomeration is the most basic phenomenon in urban commercial development, and urban commerce relies on urban space for layout and agglomeration. The higher the degree of business agglomeration, the stronger the attraction to consumers, and the greater the probability of sustainable development and successful operation after BID intervention [40]. To a certain extent, BID is about designating business districts that adopt special policies. According to the main content of the business district theory, demographic status is the most direct factor affecting business districts, including population size, age, education level, occupational composition, consumption habits, population change trends, etc. Among them, population size is an important parameter to measure the demand for business districts [41,42]. In addition, representative scholars of consumer behavior theory, Berry B.J.L. and Garrison, affirmed the impact of factors such as consumer demand for goods, spending power, and travel ability on commercial space, especially the income level of the service area population and the diversity of consumer products [43,44]. In terms of evaluating the necessity of community regeneration, the paper mainly establishes an index system from three aspects: street environment, traffic conditions, and facility supply. These three aspects are the main practical activities carried out after the establishment of BID, focusing on the goals of creating a commercial space atmosphere and improving the value of regional properties [45,46]. It is also the foundational work that needs to be carried out to leverage communities to achieve commercial revitalization and transformation. This means that if a community already has a good neighborhood environment quality, it is not necessary to introduce the BID model to promote multi-party fundraising to improve the environment. To sum up, this paper sets measurement indicators from multiple perspectives to evaluate the priority intervention of community regeneration units.

3. Materials and Methods

3.1. Study Area

In recent years, cases of multi-party governance in older communities that are very similar to BID have been spontaneously formed in southern China’s megacities, such as Chengdu, Suzhou, and Shanghai, which have more active market economies and private-driven forces for urban regeneration. These practices may be called business and residential alliance [47] or community governance alliances [48], but they do not have a unified name currently. Research shows that the performances of these cases are uneven, the majority of failed cases are due to the lack of public power support and mandatory constraints that the BID model possesses, making it difficult for participants to reach a cooperative consensus on community regeneration and achieve long-term sustainable operation [47,49]. Tianjin, which is one of the 21 megacities where the urban permanent population of more than 10 million as defined by the State Council of China, second only to Beijing among northern cities in terms of population, economy, and urban scale, went through earlier than the aforementioned southern cities of the rapid industrialization, urbanization, and industrial population aggregation from the 1970s to the early 21st century [50]. Therefore, Tianjin is the representative city in China, facing the most significant demand for urban industrial transformation and the largest number of old residential areas built in an era before the end of 2000 where quality was neglected. In other words, the severe challenges in the decline of Tianjin’s central urban area give prominence to the significance of the BID intervention discussion. This paper takes the central urban area of Tianjin as the study area (bounded by the outer ring road, with a total area of 286 square kilometers), to conduct data collection and analysis (Figure 1).

3.2. Data Source and Processing

The required data in the research scope of Tianjin are shown in

Table 1. Data and data sources.

ID	DATA Category	DATA Remarks	DATA Source
A	Residential AOI data	Reflecting the location and boundaries of all residential areas	Baidu map open platform “https://lbsyun.baidu.com/ (accessed on 25 June 2023)”
B	POI data of residential information	Reflecting the basic information of the residential areas, such as the year of construction, total number of households, floor area ratio, etc.	Lianjia company real estate agency platform “https://tj.lianjia.com/ (accessed on 25 June 2023)”
C	Community business POI datasets	Reflecting the location, type of the store, and the business district to which the store belongs.	Baidu map open platform; Dianping Public comment platform “https://lbsyun.baidu.com/; https://dianping.com/ (accessed on 3 July 2023)”
D	Open Street Map (OSM) data	Urban road alignment element data classified according to China’s road traffic standards.	CSDN developer community “https://csdn.net/ (accessed on 15 June 2023)”
E	Questionnaire data *	Obtain partial indicator data in the measurement system through on-site visits and questionnaire surveys.	Field investigation (accessed on August 2023)
F	Public service facility POI data	reflecting the spatial distribution of public service facilities	Baidu map open platform “https://lbsyun.baidu.com/ (accessed on 3 July 2023)”
G	Street view images (SVI) semantic segmentation data *	Obtain view images and semantic segmentation data of community streets through Segnet machine learning, reflecting the visual proportion of environmental elements.	Baidu map open platform “https://lbsyun.baidu.com/ (accessed on July 2023)”
H	Satellite remote sensing data	Reflecting community green coverage through NDVI calculation	“https://sentinel-hub.com/explore/eobrowser/ (accessed on July 2023)”
I	Population portrait raster data	Reflecting the characteristics of frequent phone users within the three months before the date of information collection, including income level, consumption preferences, residence and work location, etc.	Aurora Mobile mobile developer data service platform “https://jiguang.cn/en (accessed on 11 September 2023)”

*: The acquisition scope of Data E and G is based on the delineated community units obtained in Section 4.1. Except for these, all other data is obtained within the study area.

. According to the purpose of the data, it can be divided into four parts:

(1) This study conducted one-to-one information matching between POI data of residential information and residential AOI data (Data A, B) to jointly constitute the Tianjin residential dataset in ArcGIS Pro software (version 3.0.2, Esri, Redlands, CA, USA); (2) community business POI datasets (Data C) have two sources, namely Baidu and Dianping website. The former can fully cover all community commercial points after data cleaning, which is the main commercial data used in this article. The latter is only used as reference data for commercial district judgment in research; (3) Open Street Map data (Data D) are used for service area analysis and community unit indicator measurement; (4) data other than those mentioned above are only used for measurement of community regeneration units (Data E, F, G, H, and I). Questionnaire data (Data E) are obtained through a brief questionnaire of 30 residents familiar with the community in each delineated community unit, with one question used to ensure residents’ familiarity. Another question is to measure residents’ positive views on the specific indicator using a Likert scale ranging from 1 to 5 points. Public service facility POI data (Data F) are used to measure the density of public service facilities such as public toilets, free outdoor courts, and open spaces. SVI image semantic segmentation data (Data G) are used to evaluate the visual proportion of street features; satellite remote sensing data is used to calculate the green space coverage of the research object; population portrait (Persona) data (200 m × 200 m raster data) are from the Aurora data service platform and depict the personal characteristics of a large number of mobile phone users by integrating real samples provided by deep partners of the AURORA platform and “TAG tags + app behavior characteristics” provided by AURORA developers for model training and prediction.

3.3. Methodological Framework

The methodological framework is shown in Figure 2. The DBSCAN analysis and the clustering service area analysis are the preliminary research for BID priority intervention evaluation. In this process, the spatial clustering result matrix is constructed by taking different values of the two key parameters of the minimum number of elements (N) and search distance (D). Then the community regeneration units are delimited through service area analysis with the year built attributes of the neighborhoods. After that, the index system is constructed from the aspects of BID intervention potential and community regeneration necessity and calculates the scores. Finally, the BID intervention priority level of community regeneration units was classified by mathematical statistics. Combining the correlation analysis and spatial location characteristics, the main factors affecting the BID intervention potential of community regeneration units were also explained.

3.4. Method of Evaluation Objects Identification

3.4.1. Delineation of Community Units with Commercial Agglomeration

The study uses the DBSCAN spatial clustering method and service area analysis to delineate community regeneration units based on commercial service connectivity. In terms of specific operations, it first uses community-level commercial POIs such as retails, supermarkets, and restaurants as the data basis and conducts DBSCAN analysis on ArcGIS Pro 3.0.2 to obtain community-level commercial clusters. The key to DBSCAN analysis is to determine two parameters: the minimum number of cluster elements and the search distance. For DBSCAN analysis, if the minimum features per cluster parameter value can be found within the search distance from a particular point, that point will be marked as a core point and included in a cluster, along with all points within the core distance. A border point is a point that is within the search distance of a core point but does not itself have the minimum number of features within the search distance. Each resulting cluster is composed of core points and border points, where core points tend to fall in the middle of the cluster and border points fall on the exterior. If a point does not have the minimum number of features within the search distance and is not within the search distance of another core point (in other words, it is neither a core point nor a border point), it is marked as a noise point and not included in any cluster.

Then proceed to the next step after obtaining the optimal result of DBSCAN analysis. This study uses the road traffic network data set constructed from Open Street Map (OSM) data as the basis for service area network analysis. The road classification involved in constructing the road traffic network data set refers to the four-level urban roads such as expressways, main roads, secondary roads, and branch roads in the “Project code for urban road and transportation engineering” [51], and the pedestrian paths such as sidewalk roads, bicycle paths, and internal roads were included in the study. Different from previous studies that ignored road complexity and traffic light factors that hindered walking ability, this study included the impact of road intersections when assigning walking speeds on these roads. The study found that the average walking speed on auxiliary roads of urban expressway, urban main roads, and urban secondary roads is 50 m/min; the average walking speed on urban branch roads is 70 m/min; and the average walking speed on sidewalks and bicycle lanes without traffic lights is 80 m/min [52,53]. Then, excluding non-walkable roads such as highways and main roads of the Expressway, the walking speed assignment of each road section and the road traffic network data in the central urban area of Tianjin are completed. Through analysis, the main service areas of each commercial cluster within a 10 min walking radius are obtained. Finally, by combining commercial continuity (to determine a single cluster unit or multi-cluster unit) and road network data, it fits the boundaries of community regeneration units based on business district services to establish a BID. The operation flow for identifying community units with commercial agglomeration characteristics is shown in Figure 3.

3.4.2. Filtering of Community Regeneration Units

This study defines the community regeneration unit from the aspect of the average year built of residential buildings in the unit. A large number of residential buildings have been built during China’s rapid urbanization process in recent decades. At the policy level, row-type residential communities before 2000 are generally defined as old communities by the Chinese government. The average year built of the unit is calculated with the number of households in the unit as the weight. The formula of the community unit average year-built index is as follows:

$$f\left(x\right)={\displaystyle \frac{{\sum }_{n=1}^{\mathrm{\infty }}\left(H_n\times Y_n\right)}{{\sum }_{n=1}^{\mathrm{\infty }}\left(H_n\right)}}$$

(1)

Among them, $x$ ($x$ = 1, 2, 3, 4…) represents the code of community units, $n$ is the number of residential blocks in each unit, $Y_n$ is the year built of each residential block, and $H_n$ is the number of households in it. If the result is $f\left(x\right)$ ≤ 2000, then this unit will be defined as a community regeneration unit for BID intervention evaluation.

3.5. Intervention Priority Evaluation System Establishment

3.5.1. Measurement Index System of BID Intervention Potential

• Dimension of commercial compactness

Commercial space is the spatial carrier of community street vitality, and its capacity and density directly affect the viability of BID [54]. This paper selects commercial POI density and proportion of clustered POI to evaluate commercial compactness. The reason for taking these two indicators at the same time is that a commercial area may have high commercial POI density, but if commercial POIs are scattered throughout the area without obvious gathering points, then the commercial compactness may be relatively low. Relevant literature also proves that these two indicators are effective indicators for evaluating commercial or urban compactness [55].

• Dimension of population vitality

Some studies used population density indicators, including residence and workplace, in their studies to study the population vitality of communities [56,57,58]. Piña, based on population flow data, used high-frequency visitor population density to measure urban population vitality [59]. Ahas et al. and Azmi et al. selected the surrounding radiation residential population index data among the population flow change factors to analyze the spatiotemporal distribution of China’s urban population vitality [60,61]. Therefore, this article uses the density of four types of population as an indicator to evaluate the population vitality of the community business.

• Dimension of spending power

This study combines existing literature and selects three indicators: the proportion of the high-income population, the proportion of the high-consumption population, and the consumer product diversity to evaluate the level of spending power. Relevant supporting studies include Zhao et al. and Blyde et al. using indicators such as the proportion of high-income people and household consumption expenditure in different regions of China to assess differences in spending power levels between regions [62,63]. Varian and Hana et al. used consumer product-level indicators to explore how they affect consumers’ spending power and cost of living [64,65,66].

Table 2. The contents of the BID Intervention potential measurement index system.

	Dimension	Code *	Indicator	Calculation	Related Data
BID Intervention Potential Measurement	Commercial compactness	P1	Commercial POI density	Total amount of commercial POI/unit area	Commercial POI data
		P2	Proportion of clustered POI	Clustered POI/total amount of POI	Commercial POI data + Clustering results
	Population vitality	P3	Frequent population density	high-frequency (all-day) phone user population/unit area	Population portrait raster data
		P4	Residential population density	Residential (nighttime active) phone user population/unit area	Population portrait raster data
		P5	Working population density	Working (daytime active) phone user population/unit area	Population portrait raster data
		P6	Number of surrounding residents	Number of user population within 5 km radius of unit centroid	Population portrait raster data
	Spending power	P7	Proportion of high-income population	Number of middle- and high-income users/residential phone user population	Population portrait raster data
		P8	Proportion of high-consumption population	Population with medium and high consumption habits/residential phone user population	Population portrait raster data
		P9	Consumer product diversity	Number of non-mass consumer users/residential phone user population	Population portrait raster data

*: All indicators are positive.

shows the contents of the BID Intervention potential measurement index system, including index levels, index calculation methods, and sources of data obtained.

3.5.2. Measurement Index System of Community Regeneration Necessity

• Dimension of street environment

The realization of “clean, green, and safe” is the main goal of BID street environment quality creation [67]. These three aspects can be analyzed through objective composition analysis of environmental factors and subjective ratings of users [68]. Cleanliness is an important indicator for evaluating the quality of the street sanitation environment [69], which can be obtained through a questionnaire survey. In addition, unlike previous studies that mainly used the green coverage rate or street green-looking ratio as a single indicator to evaluate the green environment, this paper uses these two indicators to evaluate and obtain more comprehensive research and analysis results. Existing studies have shown that as indicators for comprehensive measurement of urban green environments, the two will have an impact on social interaction, walking behavior, and other activities [70]. In terms of the safety environment, the study refers to the relevant research on the relationship between urban environmental factors and crime and chooses to use the street enclosure index to evaluate it using SVI semantic segmentation data [71].

• Dimension of facility supply

In terms of public facilities’ environmental measurement, this paper selects the number of public toilets within a unit area for comparison, which is an indicator that is crucial to the quality of life and convenience of residents and tourists [72,73].

• Dimension of traffic conditions

Improving the pedestrian experience is the main goal of improving traffic conditions within the BID [44]. Referring to existing literature, the study selected road network density, motor vehicle occurrence rate, and proportion of street pedestrian paving to comprehensively evaluate the traffic environment. The relevant support mainly includes: Shoup (2005) used the street car appearance rate as an important indicator for evaluating the quality of the urban traffic environment [74]. Owen et al. (2004) explored the impact of environmental characteristics, including pedestrian pavement ratio, on walking quality [75]. Ewing et al. (2006) also provided indicator information for measuring the pedestrian environment, involving road network density, street pedestrian pavement ratio, etc. [76].

All mentioned indicators of this part are shown in

Table 3. The contents of the community regeneration necessity measurement index system.

	Dimension	Code	Indicators	Calculation	Related Data
Community Regeneration Necessity Measurement	Street environment	N1	Environmental cleanliness	Scoring and assignment	Questionnaire data
		N2	Street interface enclosure degree	Average visual proportion of walls and buildings	SVI semantic segmentation data
		N3	Green coverage rate	NVDI green coverage area/unit area	Satellite remote sensing data
		N4	Green-looking ratio	Average visual proportion of trees, grass, and other plants	SVI semantic segmentation data
	Facility supply	N5	Public facility density	Total amount of public facility POI/unit area	Public service facility POI data
	Traffic conditions	N6	Road network density	Unit road length/unit area	OSM data
		N7 *	Motor vehicle occurrence rate	Average visual proportion of car and motorcycle	SVI semantic segmentation data
		N8	Proportion of street pedestrian paving	Average visual proportion of sidewalk and pavement	SVI semantic segmentation data

*: N7 is a positive indicator, all other indicators are negative.

.

3.6. Calculation Based on Entropy Value and Linear Weighting Method

To calculate the potential and necessity scores of community regeneration units from indicator values, this study uses a combination of the entropy method and the linear weighting method. The entropy value method normalizes the indicators to measure the degree of dispersion of the indicators and calculate their weight, which can reduce the uncertainty caused by subjective weight assignment and eliminate the influence of positive and negative indicators. By combining the scores of various indicators and the weights obtained from the entropy method, the results of potential and necessity scores can be calculated, respectively, using the linear weighting method. The detailed calculation steps are listed in Appendix B.

4. Results

4.1. Optimal Result Selection of DBSCAN

Since the value range of the minimum number of elements (referred to as N in the article) cannot be determined before the calculation process, this paper refers to relevant studies in which the total floor area of community-level commercial space is generally more than 2500 square meters, and most community commercial shops range from 30 to 80 square meters [77]. Based on this, it was initially determined that the minimum number of elements would be set within the range of 2500/30 = 83.3 (upper limit relaxed to 90) to 3000/80 = 37.5 (lower limit relaxed to 30 pieces) to complete the simulation investigation. The value of the search distance (referred to as D in the article) refers to the formulation of 800 m as a ten-minute community walking life circle in the “Urban Residential Area Planning and Design Standards” [78]. The walking search distance is accumulated based on the 80 m walk in 1 min. That is D = 80 m (1 min), 120 m (1.5 min), 160 m (2 min), 240 m (3 min), 320 m (4 min), and 400 m (5 min) to calculate.

Table 4. Statistical on N-D spatial clustering results matrix.

Value	N = 30	N = 45	N = 60	N = 75	N = 90
D = 80 (1 min)	Noises: 29,198 Clusters: 119 NPLC: 69	Noises: 33,331 Clusters: 39 NPLC: 285	Noises: 35,364 Clusters: 11 NPLC: 149	Noises: 36,070 Clusters: 3 NPLC: 141	Noises: 36,233 Clusters: 1 NPLC:137
D = 120 (1.5 min)	Noises: 18,993 Clusters: 191 NPLC: 1560	Noises: 27,415 Clusters: 90 NPLC: 1306	Noises: 31,859 Clusters: 35 NPLC: 1097	Noises: 33,856 Clusters: 16 NPLC: 648	Noises: 34,829 Clusters: 8 NPLC: 373
D = 160 (2 min)	Noises: 10,807 Clusters: 181 NPLC: 5089	Noises: 18,272 Clusters: 95 NPLC: 2942	Noises: 25,299 Clusters: 76 NPLC: 1518	Noises: 29,889 Clusters: 37 NPLC: 1422	Noises: 32,324 Clusters: 20 NPLC: 1344
D = 240 (3 min)	Noises: 3726 Clusters: 95 NPLC: 14,062	Noises: 7239 Clusters: 95 NPLC: 7006	Noises: 11,879 Clusters: 81 NPLC: 5406	Noises: 15,885 Clusters: 78 NPLC: 4464	Noises: 20,935 Clusters: 45 NPLC: 3891
D = 320 (4 min)	Noises: 1676 Clusters: 45 NPLC: 25,256	Noises: 3234 Clusters: 50 NPLC: 19,599	Noises: 4769 Clusters: 52 NPLC: 14,913	Noises: 7246 Clusters: 60 NPLC: 9402	Noises: 9927 Clusters: 56 NPLC: 7341
D = 400 (5 min)	Noises: 864 Clusters: 28 NPLC: 29,709	Noises: 1779 Clusters: 21 NPLC: 27,894	Noises: 2554 Clusters: 24 NPLC: 25,524	Noises: 3639 Clusters: 29 NPLC: 20,119	Noises: 4908 Clusters: 35 NPLC: 17,373

and

Table A1. N-D spatial clustering result matrix.

Value	N = 30	N = 45	N = 60	N = 75	N = 90
D = 80 (1 min)
D = 120 (1.5 min)
D = 160 (2 min)
D = 240 (3 min)
D = 320 (4 min)
D = 400 (5 min)

Adjacent POIs of the same color belong to the same cluster.

(Appendix A) show the quantitative statistics and the clustering spatial distribution of the N-D clustering result matrix, respectively, which mainly refers to the data of the Dianping App to judge the rationality of different value results. Since the Dianping App is the most popular and recognized Internet platform for evaluating offline merchants in China, it covers 73 business districts outlets in Tianjin. Referring to the POI amount of the largest business district in Dianping APP, the upper limit of the number of POIs in the largest cluster (NPLC) should be 2000. When N = 30,45, D ≥ 160 m, or N takes any value and D ≥ 240 m (red part of Table 4), the result is invalid with NPLC exceeding 2000. The matrix results in the blue part of Table 4 represent invalid results where the proportion of clustering noise points exceeds 70%. Only a few cluster points can appear in a few city-level business districts; the corresponding parameters can only be applied to identify the most densely populated areas of urban commerce. Considering that there may be multiple clusters in the same business district, the number of clusters should be greater than the number of Dianping business districts. The matrix results in the yellow part of Table 4 indicate that only city- and district-level business districts are mainly displayed, and the coverage of business districts is not full.

More details of the N-D spatial clustering results are shown in Appendix A. In summary, when N = 30 and D = 120, the spatial clustering result is most in line with the current situation of community-level commercial agglomeration in the central urban area of Tianjin (Figure A1). This conclusion forms the basis for the subsequent research.

4.2. Scope of Evaluation Objects

Based on clustering results (N = 30, D = 120) and the service area network after different walking speed assignments shown in Figure 4a, this study performed a service area analysis in ArcGIS Pro. The time cost value is T = 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 5.0…10.0 (min), and the calculation result of the block range that can be covered by walking from 0.5 to 10 min from any cluster point is shown in Figure 4b.

Based on residential community distribution and road networks in Tianjin obtained above, this study combined community boundaries, road network conditions, and community business district scale to delineate community regeneration units based on walkable commercial service. The specific operation of community unit disassembly and merging is mainly based on the isochronous cover range obtained by T = 10 min service area analysis. If the isochronous cover range intersects, the two ranges will be merged. If the result is other than tangent, the unit will be divided, and at the same time, fine adjustments will be made based on the road network and the block outlines. The unit division results obtained by this method are shown in Figure 4c. The area of 93 delineated units ranges from 26 to 280 hectares, approximately covering a total of 1,409,060 households in 2083 residential blocks, and does not cover 896,733 households in 1270 residential blocks.

After the calculation of the average year built of 93 community units, the 57 selected community regeneration units as the objects for further measurement and evaluation are shown in Figure 4d. The commonality of these obtained 57 community regeneration units is that they have both the characteristics of commercial agglomeration and relatively low internal living environment quality.

4.3. Measurement of Intervention Potential and Regeneration Necessity

For calculating the indicator values for each unit, this study processed the basic data according to the scope of 57 community regeneration units. As shown in Figure 5a, this study used ArcGIS Pro software to project raster data into 57 community regeneration units and cut the raster at the unit boundaries. In the raster cut at the boundary, absolute value-type data such as the population is subtracted according to the proportion of the cut area. Proportion-type data, such as the proportion of high-income people, use the original raster data to avoid data errors due to raster cutting in subsequent calculations. On this basis, absolute value data of the community unit are obtained by adding the raster data within the unit. P3, P4, P5, and P6 indicator values are calculated based on the absolute value type of population portrait data and unit area. P7, P8, and P9 indicator values are based on the proportion type of population portrait data, which are weighted average results calculated from the original proportion data of the rasters in the unit with the residential population as the weight. The N2, N4, N7, and N8 indicator values are based on SVI semantic segmentation data, with road length as the weight, and calculate the weighted average result of the visual proportion of all road segments in the unit. An example of the green-looking ratio is shown in Figure 5b.

The evaluation of the BID intervention priority in this study is determined by both the potential of BID intervention and the necessity of community regeneration. The measurement results details are shown in Appendix B, including the scores of BID intervention potential (

Table A2. Index values and comprehensive scores for BID intervention potential measurement.

ID and Name	Location *L	F1 *1	F2 *2	F3 *3	F4 *3	F5 *3	F6 *4	F7 *2	F8 *2	F9 *2	Potential Score
26 Duolun Road	A	6.18	81.02	1681	1056	1129	229	44.64	48	3.08	0.766
37 Weidi Road	B	3.26	67.48	1674	1156	1108	199	52.13	53.07	2.79	0.6941
19 Wujiayao	B	3.15	71.55	1415	948	945	218	42.92	51	4.1	0.6752
30 Haiguang Temple	A	4.15	63.75	1738	917	1249	227	39.03	40.68	3.12	0.6725
28 Xikang Road	B	4.78	43.92	1422	921	963	232	44.98	49.28	2.99	0.631
29 Anshan Road	B	3.13	42.32	1715	968	1186	231	39.49	42.89	2.86	0.6144
31 Wandezhuang	B	4.27	70.7	1249	765	851	220	42.61	47.12	3.34	0.604
33 Machang Road	B	2.96	56.06	1405	869	971	210	44.2	47.46	3.24	0.592
35 South Tower	B	3.85	71.32	1126	710	756	193	45.01	48.07	2.9	0.5373
34 Concert Hall	B	4.55	64.45	1216	594	897	228	32.91	37.06	3.21	0.5312
62 Dabei Temple	B	3.57	68.82	870	607	563	219	53.07	53.94	2.77	0.515
84 Tangkou	B	2.33	76.77	830	628	521	208	59.77	58.41	2.67	0.5098
65 Wangchuan	B	3.24	62.19	799	600	521	191	62.55	58.05	2.49	0.4939
20 Baidi Road	B	2.94	70.54	899	593	601	198	46.68	50.66	3.2	0.4919
38 Tong Tower	C	2.88	55.47	1149	656	802	204	41.21	43.41	2.94	0.4831
82 Zhangshan Gate	C	2.9	86.35	700	537	443	152	64.6	59.24	2.6	0.4822
1 Wangdingdi	C	3.07	64.66	786	589	483	144	57.46	57.99	3.05	0.4728
56 North Dingzigu	C	2.98	52.24	936	685	621	143	64.7	57.28	2.07	0.4653
32 Changjiang Road	B	2.47	40.48	999	662	666	202	49.58	50.6	2.91	0.4642
71 Minquan Gate	C	2.32	47.78	938	719	617	141	62.79	60.39	2.3	0.4634
6 Tractor Factory	C	2.45	71.65	698	533	436	161	60.78	58.86	2.9	0.461
86 Xiangyang Tower	C	1.46	44.12	849	666	506	172	59.7	60.38	3.12	0.4605
80 Shiyijing Road	B	2.84	33.33	1105	626	759	227	38.61	43.46	2.94	0.4486
85 Wanda Square	C	3.7	80.53	641	430	422	182	50.59	51.44	2.91	0.4463
36 Culture Center	C	2.59	25	1198	721	819	182	45.04	46.6	2.63	0.4442
73 Yibai Road	C	3.82	74.67	635	488	414	114	67.15	60.21	2.19	0.4414
88 North Wanxin	D	3.37	62.78	757	559	496	128	63.63	57.74	2.18	0.4366
55 South Dingzigu	B	3.22	65.56	650	478	424	157	59.65	57.41	2.57	0.4337
49 Xiyuehui Plaza	D	3.12	73.29	804	543	539	136	56.01	52.49	2.35	0.4296
47 East Jiangjing	D	2.6	69.78	720	562	452	120	66.89	60.84	2	0.4236
21 Pingshan Road	C	2.57	63.13	815	514	543	210	44.02	47.49	2.84	0.4234
50 Luofu Bridge	D	1.9	51.56	767	605	493	120	72.1	61.21	2.14	0.4207
76 Yueya River	C	2.21	32.69	855	642	571	149	66.7	58.61	2.1	0.4195
51 Xiying Gate	C	2.29	45.71	741	545	482	161	63.07	57.12	2.49	0.4173
54 West Rail Station	B	3.46	73.44	754	457	514	182	50.47	47.13	2.32	0.415
52 Shaogong	C	1.72	40	738	568	474	157	68.08	59.54	2.43	0.4136
83 Fumin Bridge	C	1.6	67.86	596	468	374	149	65.55	60.8	2.65	0.4125
44 Univ of Finance	D	2.19	67.07	832	588	545	94	49.17	55.01	2.94	0.4044
43 Huashanli	D	2.08	36.57	836	632	540	104	58.4	58.93	2.73	0.3987
13 Tianjin TV Tower	C	2.87	71.72	704	430	467	194	42.7	47.06	2.8	0.3963
41 Weishan Road	D	1.93	48.07	713	566	446	85	60.94	61.66	2.97	0.3952
11 Sports Collage	C	2.91	30.06	761	554	492	148	55.28	56.37	2.68	0.3947
12 Sports Collage (N)	C	1.99	21.59	995	606	687	184	47.34	47.07	2.65	0.3877
46 Tiantuobei	C	2.16	22.26	704	525	465	167	64.94	58.02	2.36	0.3821
42 Steel Factory	D	2.03	53.33	720	556	464	94	66.99	59.97	2.11	0.3795
68 Jiangdu Road	C	2.29	49.9	591	426	392	169	60.77	55.78	2.35	0.3723
81 Jianfu Temple	B	2.06	39.64	648	445	428	185	53.82	53.09	2.62	0.3626
90 Bridge No.1	D	2.1	91.07	538	346	347	117	46.62	49.08	3.26	0.3582
92 Bridge No. 2	D	2.6	40.95	684	517	431	96	62.45	59.34	2.21	0.357
89 South Wanxin	D	1.6	56.65	626	446	410	127	57.86	55.64	2.56	0.3512
57 Liuyuan East	D	1.85	51.06	524	407	330	90	65.59	60.33	2.54	0.3379
17 Xinye Plaza	D	1.62	39.13	729	475	494	148	50.53	50.29	2.6	0.3327
93 Zhangguizhuang	D	1.12	48.78	534	419	322	87	63.53	60.87	2.91	0.3324
3 Huayuan East	D	1.19	36.04	547	421	333	122	56.58	58.44	3.12	0.3257
61 Nancang North	D	2.66	60.26	627	426	421	54	57.72	53.57	2.4	0.3222
40 Tucheng-Badali	C	1.76	49.57	624	399	410	156	49.98	49.6	2.23	0.3028
60 Nancang south	D	1.94	55.38	558	374	370	63	53.51	52.3	2.4	0.2719

*^L Location: A—within the inner ring road, B—inner ring road-middle ring road, C—middle ring road-expressway, D—expressway-outer ring road; *¹ number/hectare, *² percent, *³ number of people/square kilometers, *⁴ number of people.

) and community regeneration necessity (

Table A3. Index values and comprehensive scores for community regeneration necessity measurement.

ID and Name	Location *L	N1 *1	N2 *2	N3 *2	N4 *2	N5 *3	N6 *4	N7 *2	N8 *2	Necessity Score
47 East Jiangjing	D	24	31.51	13.18	0.47	1.87	3.07	1.53	2.33	0.88
52 Shaogong	C	30	28.31	32.77	4.65	1.11	8.19	0.69	2.27	0.81
89 South Wanxin	D	33	29.02	9.79	2.55	1.85	9.08	2.86	1.62	0.79
90 Bridge No.1	D	31	19.57	19.98	5.33	3.75	11.26	0.41	3.04	0.79
50 Luofu Bridge	D	29	33.49	28.14	1.32	2.97	10.49	0.46	3.09	0.78
49 Xiyuehui Plaza	D	27	19.73	22.26	3.36	11.64	8.36	0.66	1.18	0.76
13 Tianjin TV Tower	C	39	30.23	20.86	5.97	0	8.63	1.96	2.53	0.75
57 Liuyuan East	D	28	24.3	16.13	5.68	4.3	11.41	2.3	3.07	0.75
81 Jianfu Temple	B	33	23.07	20.04	7.98	5.76	8.99	1.8	1.71	0.74
31 Wandezhuang	B	32	33.23	21.78	5.63	2.82	8.31	3.8	1.49	0.72
32 Changjiang Road	B	36	26.34	32	7.86	0.86	9.85	3.29	1.68	0.72
40 Tucheng-Badali	C	30	23.08	20.36	4.92	7.51	9.76	1.08	3.31	0.72
1 Wangdingdi	C	32	30.51	33.54	3.23	7.39	7.51	1.76	2.12	0.7
6 Tractor Factory	C	30	36.02	20.19	4.97	4.5	8.31	2.45	3.34	0.7
41 Weishan Road	D	30	29.81	30.54	5.08	0.68	9.8	1.95	5.65	0.7
56 North Dingzigu	C	33	27.24	31.07	9.79	4.44	7.25	0.83	3.82	0.7
86 Xiangyang Tower	C	33	33.44	29.33	4.49	1.43	7.85	2.87	3.67	0.7
36 Culture Center	C	40	30.49	20.24	3.54	4.32	11.6	2.55	2.16	0.68
73 Yibai Road	C	30	31.2	17.68	5.37	2.19	11.38	4.29	3.94	0.68
83 Fumin Bridge	C	38	30.2	34.55	6.36	1.9	8.25	3.49	2.04	0.68
82 Zhangshan Gate	C	35	23.45	24.68	8.89	7.64	7.55	3.13	1.5	0.67
93 Zhangguizhuang	D	34	31.07	24.92	8.21	1.37	11.8	4.4	2.41	0.67
20 Baidi Road	B	34	34.85	25.76	7.55	3.16	9.45	3.06	3.45	0.66
51 Xiying Gate	C	26	25.28	24.48	7.01	10.09	9.64	2.1	3.35	0.66
92 Bridge No. 2	D	32	35.9	25.06	7.97	0	9.35	5.35	2.84	0.66
17 Xinye Plaza	D	34	24.79	35.61	5.17	9.89	10.52	1.34	2.45	0.65
21 Pingshan Road	C	37	38.92	36.25	6.3	1.44	7.91	4.79	2.18	0.63
46 Tiantuobei	C	25	35.66	19.78	2.23	5.57	10.02	2.8	7.23	0.63
68 Jiangdu Road	C	30	30.45	21.17	7.03	7.25	8.95	2.46	4.89	0.63
65 Wangchuan	B	32	37.84	22.32	5.66	4.48	13.17	3.81	3.64	0.62
85 Wanda Square	C	30	23.55	26.93	6.37	13.85	10.16	2.61	1.43	0.62
37 Weidi Road	B	35	33.39	25.47	5.04	7.96	11.72	3.42	2.36	0.61
62 Dabei Temple	B	38	39.82	19.62	4.83	5.77	10.89	2.45	3.88	0.61
88 North Wanxin	D	32	34.72	33.54	3.43	3.52	11.93	4	4.36	0.61
76 Yueya River	C	34	29.9	32.88	8.64	6.37	9.68	0.75	6.32	0.6
80 Shiyijing Road	B	37	35.48	17.4	4.94	7.15	12.37	3.05	3.76	0.6
3 Huayuan East	D	32	30.98	39.64	9.25	1.07	16.6	4.91	3.91	0.59
30 Haiguang Temple	A	33	37.31	13.2	5.16	10.39	11.48	3.87	2.54	0.59
42 Steel Factory	D	30	31.68	42.86	7.7	2.46	9.48	2.46	7.69	0.59
54 West Rail Station	B	31	27.68	22.4	4.6	18.9	16.41	0.57	1.04	0.59
55 South Dingzigu	B	32	35.65	33.06	7.78	6.29	10.89	3.57	3.64	0.59
61 Nancang North	D	32	41.06	22.79	4.69	1.13	14.56	3.66	6.68	0.59
28 Xikang Road	B	31	38.28	25.79	6.86	5.28	13.01	3.75	4.78	0.58
38 Tong Tower	C	33	34.92	23.53	8.06	10.5	10.33	3.04	2.71	0.58
84 Tangkou	B	29	23	34.47	16.29	9.4	17.56	2.3	2.62	0.58
60 Nancang south	D	33	39.45	18.17	6.82	3.86	17.28	4.1	5.43	0.56
11 Sports Collage	C	33	25.65	31.93	10.8	10.7	10.75	2.37	4.48	0.55
19 Wujiayao	B	36	24.11	38.47	21.18	5.44	8.09	3.86	3.65	0.55
44 Univ of Finance	D	33	24.71	38.76	9.25	13.35	7.65	1.1	4.84	0.55
33 Machang Road	B	35	40.09	20.74	5.21	11.22	11.2	3.74	2.88	0.54
35 South Tower	B	39	34.94	23.73	6.45	9.46	11.3	3.09	3.85	0.54
12 Sports Collage (N)	C	34	31.48	45.8	9.94	4.52	11.21	2.66	6.67	0.53
71 Minquan Gate	C	31	21.8	36.16	11.36	7.74	9.15	2.17	8.64	0.53
29 Anshan Road	B	34	39.79	22.33	6.36	11.75	13.04	4.15	3.49	0.5
26 Duolun Road	A	36	50.06	19.04	4.8	7	14.78	3.89	5.45	0.49
43 Huashanli	D	33	30	33.23	9.98	9.53	9.94	3.35	6.48	0.49
34 Concert Hall	B	39	38.16	20.3	5.05	11.15	18.12	3.35	4.82	0.47

*^L Location: A—within the inner ring road, B—inner ring road—middle ring road, C—middle ring road—expressway, D—expressway—outer ring road; *¹ score, *² percent, *³ number/hectare, *⁴ meter/hectare.

) and the values of 9 potential indicators and 8 necessity indicators for 57 urban regeneration units. The key parameters generated in the process of calculating them by the entropy method are shown in

Table 5. Indicator weights and other process parameters of the entropy method.

Part 1—Potential measurement
	MAX	MIN	DV	ej	dj	W
P1	6.9605	0.6801	6.2804	0.9616	0.0384	0.1389
P2	0.9271	0.2159	0.7112	0.9754	0.0246	0.0890
P3	1738.0755	288.4548	1449.6207	0.9638	0.0362	0.1309
P4	1156.3143	171.6266	984.6876	0.9751	0.0249	0.0901
P5	1248.9031	193.9208	1054.9823	0.9570	0.0430	0.1554
P6	2,404,759	539,152	1,865,607	0.9703	0.0297	0.1074
P7	0.7210	0.3159	0.4051	0.9717	0.0283	0.1022
P8	0.6166	0.3706	0.2460	0.9779	0.0221	0.0799
P9	0.0410	0.0192	0.0218	0.9707	0.0293	0.1061
Part 2—Necessity measurement
	MAX	MIN	DV	ej	dj	W
N1	40.0000	24.0000	16.0000	0.9761	0.0239	0.0836
N2	0.5006	0.1717	0.3290	0.9745	0.0255	0.0892
N3	0.5209	0.0979	0.4230	0.9772	0.0228	0.0796
N4	0.2118	0.0047	0.2071	0.9746	0.0254	0.0888
N5	22.8715	0.0000	22.8715	0.9152	0.0848	0.2966
N6	18.5024	3.0694	15.4330	0.9811	0.0189	0.0662
N7	0.0567	0.0041	0.0525	0.9635	0.0365	0.1274
N8	0.0906	0.0104	0.0802	0.9518	0.0482	0.1686

, including the weights (W) of various indicators. Among the potential measurement indicators, P5 (working population density) has the highest degree of value dispersion with W = 0.1554, while P8 (proportion of high-consumption population) has the lowest degree with W = 0.0799. Correspondingly, the indicators with the highest and lowest degree of dispersion in necessity measurement are N5 (public facility density) and N3 (green coverage rate), respectively.

4.4. BID Priority Intervention Classification of Units

This study drew a scatter plot about the results of BID intervention potential and community regeneration necessity of units to explore the degree of mutual influence (Figure 6). The linear regression result shows that the R² value is 0.13, and there is no obvious correlation between potential and necessity scores, which means that these two sets of measurement index systems are relatively independent of each other. By observing the linear trend of the regression line of y = 0.72 + −0.424 x, there is a slight negative correlation.

The potential and necessity scores of 57 community regeneration units are statistically distributed according to the standard normal distribution as shown in Figure 7a,b. For potential and necessity scores, the average value is 0.4504 and 0.3612, and the standard deviation is 0.1031 and 0.0889, respectively. This paper uses the average value and positive and negative standard deviations of potential and necessity scores to divide the 57 community regeneration units into four levels separately. The spatial distribution of community regeneration units with different levels of potential and necessity is shown in Figure 7c,d.

Considering that some community regeneration units span the ring roads that are used to define the location, the location definition of the units is based on their spatial centroid. The spatial distribution of the four potential levels of units is shown in Figure 7c. It can be found that the 13 units with the highest potential level are all located within the middle ring road in Tianjin, while the other 13 units with the lowest potential level, shown in Figure 7d, are basically located outside the expressway. However, the units with high necessity levels are more distributed in the area close to the periphery of the city, although the distribution pattern of necessity levels is not as clear as that of potential levels. Based on this finding of location divergence in the potential and necessity scores of community renewal units, in the following text, we will further explore the location characteristics of community units. By arranging and combining the potential and necessary levels for community regeneration units, the quantity statistics and spatial distribution of the unit classification are shown in Figure 8a,b.

According to the principle that the higher the level of potential and necessity, the higher the priority of BID intervention, this study divided community regeneration units into five priority levels. Priority—I is the highest level, including P4N3 type unit 31 and P3N4 type unit 49. The Sankey diagram in Figure 9 and

Table 6. Statistics of community regeneration unit types based on priority level and location range.

Location Ranges	Priority-I		Priority-II		Priority-III		Priority-IV		Priority-V
A. Within the inner ring road			P4N2	1	P4N1	1
B. Inner ring road— middle ring road	P4N3	1	P4N2	2	P4N1	5	P2N2	1
			P3N3	2	P3N2	6
			P2N4	1
C. Middle ring road— expressway			P3N3	7	P3N2	2	P3N1	1	P2N1	1
			P2N4	2	P2N3	2	P2N2	5
							P1N3	1
D. Expressway— outer ring road	P3N4	1	P2N4	5	P3N2	1	P2N2	3	P2N1	1
					P2N3	4			P1N2	2
Total		2		20		20		11		4

shows the distribution of different types of community regeneration units grouped by priority level in urban location ranges. From the perspective of unit type distribution, the “P4-” and “P3-” types of units with higher potential levels are mainly concentrated within the middle ring road. On the contrary, the “-N4” and “-N3” types of units with higher necessity levels are mainly distributed outside the middle ring road.

4.5. Main Influencing Factors and Location Differentiation

To judge the key factors that affect the BID intervention potential and community regeneration necessity of 57 units in Tianjin, this paper conducts regression analysis using the values of each indicator as the independent variables and the potential and necessity scores as the dependent variables (

Table 7. Regression analysis of potential and necessity scores and corresponding indicator values.

Part 1—Potential measurement
	P1	P2	P3	P4	P5	P6	P7	P8	P9
R2-Value	0.57 **	0.11	0.80 ***	0.82 ***	0.76 ***	0.53 **	0.21	0.17	0.26
Part 2—Necessity measurement
	N1	N2	N3	N4	N5	N6	N7	P8
R2-Value	0.14	0.16	0.09	0.20	0.24	0.31 *	0.20	0.27

*: 0.5 > R² ≥ 0.3 (low goodness-of-fit), **: 0.7 > R² ≥ 0.5 (moderate goodness-of-fit), ***: R² ≥ 0.7 (high goodness-of-fit).

).

For the potential measurement, the indicator values of the population vitality dimension have a more significant goodness-of-fit with the potential score. Among them, the high-frequency population density index (P3) is the statistical union of the residential and working population density (P4 and P5). It has a significant impact on the BID potential scores of the community to almost the same extent as the residential population density index (p4). The working population density (P5) contributes relatively less to the community potential score than the residential population density (P4). This indicates that the population living in the community is the most critical determinant, followed by the working population. The surrounding population (P6) also contributes to the BID potential scores of the community, but it is significantly lower than the three indicators mentioned above (P3, P4, and P5). There is a moderate goodness-of-fit between commercial POI density (P1) and potential score, which indicates that the commercial capacity of the community will affect the BID intervention potential, but it is not a significant decisive factor. The clustering POI proportion index (P2) and the three indicators of the spending power dimension (P7, P8, and P9) have a poor fit with the potential score, indicating that residents’ income, consumption habits, and the diversity of consumption levels are not the key factors currently affecting the BID potential level.

For the necessity measurement, only road network density (N6) has a low goodness-of-fit with necessity scores among all necessary indicators; other necessity indicator values do not have a good fit with necessity scores.

To further study the differences in the impact of influencing factors in different location ranges, this study grouped the community regeneration units by location ranges and conducted another two linear regression analyses for potential and necessity measurement systems. There are only two samples (26# and 30# units) within the location range A, which is not sufficient for linear regression analysis. Therefore, unit samples of location ranges A and B will be merged for statistical analysis. In order to facilitate the explanation of urban location, this article follows the classification of location ranges in

Table 8. Regression analysis of potential and necessity scores and indicators grouped by location.

Part 1—Potential measurement
Location Ranges of Units	P1	P2	P3	P4	P5	P6	P7	P8	P9
Urban core area (A + B)	0.19	0.04	0.78 ***	0.88 ***	0.73 ***	0.30 *	0.12	0.03	0.36 *
Urban middle area (C)	0.17	0.15	0.17	0.33 *	0.11	0.00	0.01	0.04	0.14
Urban fringe area (D)	0.32 *	0.09	0.59 **	0.66 **	0.50 **	0.22	0.12	0.10	0.09
Part 2—Necessity measurement
Location Ranges of Units	N1	N2	N3	N4	N5	N6	N7	N8
Urban core area (A + B)	0.09	0.19	0.00	0.01	0.37 *	0.32 *	0.06	0.41 *
Urban middle area (C)	0.00	0.00	0.15	0.39 *	0.27	0.21	0.09	0.35 *
Urban fringe area (D)	0.44 *	0.14	0.44 *	0.61 **	0.06	0.27	0.26	0.50 **

*: 0.5 > R² ≥ 0.3 (low goodness-of-fit), **: 0.7 > R² ≥ 0.5 (moderate goodness-of-fit), ***: R² ≥ 0.7 (high goodness-of-fit).

, which are called urban core area (location ranges of A + B), urban middle area (location range of C), and urban fringe area (location range of D) from inside to outside. The result is shown in Table 8.

For the potential measurement, among the indicators of the commercial compactness dimension, only the units in the urban fringe area have a moderate goodness-of-fit between commercial POI density (P1) and the potential scores, which shows that the BID intervention potential of communities on the fringe of Tianjin will be affected by insufficient commercial volume. Indicators of the spending power dimension have little or no goodness-of-fit with the BID potential score in all locations, indicating that spending power-related residents’ income levels, residents’ consumption levels, and the diversity of consumption habits (P7, P8, and P9) are not key factors that affect BID intervention potential. The indicators of the population vitality dimension have the highest goodness-of-fit in each location range. The most relevant influencing factor for units is residential population density (P4). In the urban middle area, there is no statistical goodness-of-fit between the potential scores and all the indicators except residential population density (P4).

For the necessity measurement, the proportion of street pedestrian paving (N8) has an impact on necessity scores in all locations. In addition, the necessity scores of units in the urban core area are more affected by public facility density (N5) and road density (N6) than other locations, while in the urban fringe area, environmental cleanliness (N1), green coverage rate (N3), and green looking ratio (N4) have a greater impact. In general, the goodness-of-fit among indicator values and necessity scores is significantly better than the regression results in Table 7 after grouping by location. Moreover, similar to the regression results of potential measurement, there are obvious differences in the main influencing factors of different location ranges.

5. Discussion

5.1. The Non-Universality of the BID Model for Community Regeneration

The urban regeneration strategies for old communities need to be customized according to the specific local conditions. Not all of them meet the prerequisites for implementing BID [14,24], and its success is influenced by multiple factors [5]. The more an old community has the elements required for BID success, the more potential the community has to realize sustainable operation and renew [17]. According to this point, this study identified the community regeneration units by characteristics of the year built and commercial agglomeration for the aims of low-cost and precise intervention of BID. The research result of unit scope calculated by using a spatial clustering algorithm and service area analysis method does not cover all residential areas in the study area. These uncovered communities do not have commercial clusters of sufficient scale within the walking time range defined in this study, which provides a method for preliminary screening of communities that can implement the BID model with commercial entry points.

The delineation results of the units in this study are basically in line with the community commercial service scale proposed by Feng Zhiyong, He Jun, and other scholars in their respective studies: the travel distance does not exceed 1000 m, and the vertical walking radius from the residential area to the commercial area (mainly community shopping malls, 24 h convenience stores, restaurants, etc.) is 5–10 min [36,79]. Specifically, when conducting DBSCAN spatial clustering analysis using life service commercial POIs, the minimum number of elements is 30 and the search distance is 120. The results are most consistent with the division of Tianjin community business districts by the network platform (Dianping) with the highest usage rate. At the same time, these two values also indicate that the number of commercial shops in the community is at least about 30, with an area of about 3000 square meters. Thus, if the commercial density within 120 m search distance from the commercial gathering area can still be maintained, this range can be considered to belong to the business district. According to this standard, there are a total of 93 community units in Tianjin with commercial agglomeration characteristics. On this basis, the list of 57 community regeneration units was obtained by overlaying the average year built, which provides a directory for community regeneration practices under the framework of the BID model.

5.2. BID Intervention in Megacities Should Focus on Differentiated Regeneration Strategies

Tianjin, as a monocentric city, is experiencing the impact of population and industrial decentralization [12,80]. This study found that the units in the urban core area still have higher population vitality than the more peripheral urban areas. In addition, population density, especially residential population density, is the most important influencing factor on BID potential in all the location ranges, followed by the working population density, and exhibits high good-of-fit with potential scores throughout the areas except the ring-shaped urban middle area. It is supposed that the units in the urban core area showing higher goodness-of-fit are partly because the urban core area is saturated with population density and has a profound impact on urban vitality [81]. Although the units in the urban middle area are also affected the most by population density, the regression results show that there is no absolutely dominant influencing factor. The units in the urban middle area face more complex situations and may be affected by the main development and expansion directions of the city. The urban fringe area of the city has not yet been fully developed, and the commercial layout and job opportunities are generally not large [82]. Therefore, whether there is large-scale residential development and population agglomeration will become the main factors affecting the BID potential of the community in an urban fringe area. The influencing factors of the commercial compactness dimension, such as the density of commercial facilities and the degree of agglomeration, have a weak impact on BID potential. Spending power-related factors about residents’ income, residents’ spending power, and the degree of diversification of residents’ consumption are not important influencing factors either.

The above research on the influencing factors of BID intervention potential and community regeneration necessity indicates that Chinese megacities should consider location differentiation in strategies for community regeneration units when establishing BIDs in priority order. After urban regeneration actions with a “demolition and reconstruction” style in the past two decades, the area within the inner ring road of the city has instead been mainly composed of communities built after 2000 AD. Therefore, “P4- ” and “P3- ” types of community regeneration units with high BID intervention potential are mainly concentrated between Tianjin’s inner ring road and middle ring road. Taking units with priorities I and II as examples, the BID strategy for P4N3 and P4N2 types of units should focus on enhancing the public facility density, road density, pedestrian paving of the community, etc. P3N4 and P2N4 types of units, which are mainly distributed outside the urban core area, should focus more on strategies such as attracting population, increasing commercial density, improving environmental hygiene, enhancing green environments, etc. Conducting BID site selection based on the priorities judged by this study and formulating differentiated regeneration strategies based on the types of community regeneration units and their locations will help reduce later government financial investment and achieve sustainable BID community autonomy.

5.3. Research Significance and Prospects

The establishment of BID mostly relies on the self-organization of merchants and residents, with a few coming from government initiatives [44]. In the context of China’s “big government” and relatively weak experience of bottom-up public participation, selecting communities with higher BID potential will make the establishment of the low-cost BID tool more feasible against the background of the huge community regeneration needs in China’s megacities and the financial pressure on urban governments. When China is expanding domestic demand and transforming its economic development model, comprehensively promoting consumption has become a necessary measure to cope with medium- and long-term challenges and maintain economic resilience [83]. The existing policy measures for enhancing commercial space are more focused on main commercial streets, mature large commercial districts, and business office areas [84], and there is a clear lack of attention to the upgrading and transformation of “small” community businesses that carry a large amount of urban consumption. Therefore, compared with the types of “Corporate BID” and “Main Street BID”, exploring the localization implementation mechanism, policies, and governance methods of “Community BID” is of great significance for accelerating the construction of China’s circular economy pattern [85]. Although there is relatively little research and practice in the BID model in China, it is necessary to seize the integration points of urban old residential communities and community commerce in terms of development necessity in the policy transfer process and explore a path for developing BID that is different from that in European and American countries.

This study fills the research gap in the application of the BID model in the central areas of China’s megacities. On the one hand, this study established a community regeneration unit delineation method that reflects the core concept of the BID spatial governance model. In particular, the density-based spatial clustering method and the service area analysis method considering road level and traffic obstruction factors were adopted to supplement the shortcomings of existing research in the scope definition of community-level regeneration units. On the other hand, this study demonstrated that when China’s megacities promote community regeneration through the establishment of BID, old communities in urban centers with high population density and job opportunity density are the best choice. The results of this study can not only provide a basic database and information materials for the implementation of the BID model in the old communities of Tianjin but also provide a reference for the research and practice of community regeneration governance in commercial-residential linkage in other megacities in China.

6. Conclusions and Future Works

This paper provides a macroscale operational technical path for megacities, from object identification to priority intervention evaluation of BID. Based on the above analysis, the main theoretical contributions of this study to fill the research gaps are briefly summarized as follows: (1) Constructed a method for identifying the objects and their scope of BID intervention from the macroscale perspective of megacity; (2) established a BID priority intervention evaluation tool consisting of measurements of BID intervention potential and necessity; and (3) proposed the differentiated strategy directions for BID intervention targeting different types of community regeneration units. The above innovations together form the technical path for the priority intervention evaluation of community regeneration in megacities based on the BID model.

Specifically, community regeneration units with commercial agglomeration characteristics in Tianjin were initially delineated based on urban spatial big data, through DBSCAN analysis, service area analysis, and the calculation of average year built. Then, by summarizing the factors that affect the BID intervention potential and community regeneration necessity of communities in existing studies, this study constructed corresponding measurement systems to classify the community regeneration unit types for BID priority intervention. In addition, based on the obvious spatial distribution characteristics of the measurement results, this study supplemented exploring the key influencing factors of BID intervention in China’s megacities, explaining the impact differences in indicators among communities from a location perspective. On this basis, this study proposes differentiated community regeneration strategy directions. The paper’s conclusions can provide references for the governments of megacities to establish BIDs. The main findings are as follows:

(1)

When conducting DBSCAN spatial cluster analysis with the minimum number of elements (N) of 30 and the search distance (D) of 120 m, a community commercial cluster result that is closer to the actual situation according to the comparison of the Dianping data. The service area analysis based on the clustering results can determine the range of community units with commercial agglomeration characteristics.

(2)

Population vitality, especially working and residential population density, is the most critical factor affecting BID intervention potential, followed by community commercial density. Among the indicators that determine the necessity for community regeneration, road network density is the most important one.

(3)

There is obvious location differentiation in the critical influencing factors of both the BID intervention potential and the community regeneration necessity of units. Different types of community regeneration units with high BID priority need to develop differentiated regeneration strategies based on their BID intervention potential, community regeneration necessity characteristics, and location attributes.

This study has potential limitations. It should be pointed out that the research is based on the objective conditions of the communities in China’s megacities that have not yet established a BID. There is currently insufficient discussion on how the BID model can adapt to the operational mode of the Chinese government and civil organizations from a policy perspective. In addition, the macroscale technical path proposed in this article may result in different optimal parameter values for DBSCAN in different megacities, and the factors affecting the priority intervention of community regeneration unit BID may also vary depending on the differences among megacities. In the future, the research object can be horizontally expanded to other Chinese megacities to compare the similarities and differences in BID priority interventions in order to better study community regeneration and governance under the BID model in China. Meanwhile, research can also continue to delve vertically, selecting research objects from the community update units identified in this article and further expanding the methods of microscale BID intervention.