Identification and Redevelopment of Inefficient Residential Landuse in Urban Areas: A Case Study of Ring Expressway Area in Harbin City of China

Abstract

The current efficiency of residential land utilization is witnessing a decline, attributable to accelerated urbanization and inefficient resource allocation, thereby presenting unprecedented threats and challenges to the quality of urban living and the pursuit of high-quality urban development. To enhance residents’ satisfaction and well-being, and to effectively activate existing land resources, it is imperative to accurately identify inefficient residential landuse and its driving factors. While the literature on identifying inefficient urban landuse is expanding, research specifically focusing on residential land, which is closely linked to residents’ lives, remains limited. Furthermore, the factors contributing to inefficient land use are relatively inadequate. Therefore, this study employs a “two-step identification method” to comprehensively identify inefficient residential landuse and utilizes standard deviation ellipses and kernel density assessment methods to analyze the spatial distribution characteristics of such land. Subsequently, the study employs the Random Forest (RF) model to quantitatively analyze factors such as building quality, economic, social, and ecological factors, aiming to provide a scientific basis for subsequent redevelopment initiatives. The findings reveal that inefficient residential landuse is primarily concentrated in city centers, particularly in districts such as Nangang and Xiangfang. In relative inefficient residential areas, aside from Nangang District and Xiangfang District, Songbei District also holds a significant proportion. The intensity of these associations with inefficient residential landuse formation varies depending on urban development history and regional development intensity. In areas other than Songbei District, factors such as aging residential neighborhoods and inadequate green spaces are major contributors to inefficient land use efficiency, whereas in Songbei District, insufficient medical and educational facilities are the primary factors. The RF algorithm, distinguished by its flexibility and accuracy, offers novel perspectives and methods for analyzing issues related to inefficient residential landuse. Moreover, it effectively manages nonlinear relationships between the data, avoiding overfitting and generating precise regression and classification results. Thus, the RF algorithm demonstrates significant promise for widespread application in urban land studies.

1. Introduction

Since the economic and developmental reforms initiated in the late 1970s, China has undergone rapid and sustained urbanization, emerging as a leading developing nation globally [1,2]. However, inadequate long-term planning has precipitated numerous urban challenges, such as urban decay, shrinkage, environmental pollution, traffic congestion, and insufficient facilities [3,4,5]. These issues not only jeopardize the preservation of high-quality farmland, ecological balance, and living environments but also impede the efficient and intensive utilization of land resources [6,7]. Therefore, there is increasing attention on urban renewal within these issues [8]. In Europe, approaches include the Fuzzy model of adaptive reuse of historical structures and the renovation of old communities in Germany’s Ruhr metropolitan area [9,10]. In recent years, the urban construction land expansion rate in China has significantly decelerated [11]. This emphasis on revitalization and utilization of inefficient construction land is highlighted in the Key Tasks for New-Type Township Construction in 2019 and Key Tasks for New-Type Township Construction and Urban-Rural Integration in 2020. In 2023, the Ministry of Natural Resources (MNR) issued a notice on pilot projects to redevelop inefficient land use, explicitly proposing the activation of existing land resources. Urban construction is recommended to prioritize revitalizing and utilizing existing land to improve land use efficiency [12]. Furthermore, this approach promotes the quality of urban development by advocating a transition from superficial scale expansion to substantial quality enhancement [13,14].

Inefficient landuse serves as the fundamental catalyst for diverse urban social issues. The identification and redevelopment of inefficient landuse has garnered substantial scholarly attention. Regarding methods for identifying inefficient landuse, there are currently two main methods in academia: one based on remote sensing imageries [15], utilizing techniques such as multiscale segmentation and object-oriented classification to extract information on inefficient landuse [16], and the other is based on statistical survey analysis [17], constructing an evaluation indicator system and delineating inefficient residential landuse according to evaluation criteria [18,19]. Furthermore, with the development of internet information technology, the empowerment of informatization has gradually been integrated into urban land research. Big data sources such as Baidu heatmaps [20,21], POI data [22,23], mobile signaling [24,25,26], and artificial intelligence [27] provide dynamic and accurate perspectives for urban development, also offering reliable data support for subsequent redevelopment efforts. In terms of redeveloping inefficient residential landuse, scholars have proposed multiple redevelopment strategies from different perspectives such as transformation models [28], management mechanisms [29], policy systems [30,31], and investment financing systems [32]. These strategies address different types of inefficient residential landuse, such as old residential communities [33] and living areas surrounded by “noise” [32], aiming to promote the efficient and intensive use of urban residential land through land consolidation. This ongoing research underscores the critical link between residential land and residents’ lives, reflecting its profound impact on the urban living experience.

Precisely analyzing the factors contributing to inefficiency is crucial for proposing scientifically effective redevelopment measures. Cui found that factors such as road network density and the quality of the living environment significantly impact the efficiency of residential land use [19]. The primary approaches employed to analyze these driving factors include spatial regression models and empirical statistical models. However, empirical models often grapple with overfitting, particularly when handling high-dimensional data sets or a surplus of variables. The Random Forest (RF) algorithm, known for its robustness against overfitting, can handle a diverse array of quantitative and qualitative explanatory variables [34]. Therefore, the RF algorithm not only produces greater interpretability of results but also allows for the extraction of the importance and marginal effects of each predictor [35,36]. Furthermore, unlike other common machine learning approaches, RF can generate reliable results and insights into the contribution of driving factors through its variable importance measure (VIMs) feature, an aspect less effectively addressed by other methods. Consequently, utilizing the importance ranking derived from RF in analyzing the causes of inefficient residential landuse provides a scientifically robust framework that supports effective redevelopment initiatives.

To summarize, current scholars have investigated various aspects of urban land use, including the identification of urban construction land [37], the redevelopment of inefficient construction land [38], and the comprehensive examination of inefficient industrial landuse concerning its identification [39], evaluation [40,41], and redevelopment [42]. However, a systematic focus on the identification and optimization of inefficient residential landuse remains notably sparse. Most studies have not constructed a tailored indicator system for the identification of inefficient residential landuse, reflecting various challenges such as inconsistent assessment criteria for inefficient residential landuse, the difficulty involved in redevelopment, and the significant social impacts associated with these processes. In addition, research on land use in China predominantly targets economically prosperous or high-profile areas such as mega-cities like Beijing [43], Shanghai [44], and Guangzhou [45], with scant attention to the Northeast of China, a region where residents are experiencing severe population decline and significant underdevelopment in terms of quality urban growth. The demographic reductions in this area not only affects local economic development but also directly leads to the emergence of inefficient residential landuse. In response to these challenges, this paper categorizes inefficient residential landuse into two types: absolute inefficient residential landuse and relative inefficient residential landuse. Absolute inefficient residential landuse refers to residential land that fails to comply with the construction regulations or belongs to the old neighborhoods. Conversely, relative inefficient residential landuse encompasses residential neighborhoods beset by suboptimal economic, living, ecological and other conditions, failing to meet modern living standards.

Harbin is an early-established city in the Northeast and has seen various developmental phases that have shaped its residential landscape. During the periods of “the first five-year plan” and “the second five-year plan”, numerous factories were constructed, and extensive residential areas were developed. Some residential neighborhoods in Harbin date back 50 or 60 years ago. The onset of economic reforms and the opening-up policy led to the establishment of residential districts by enterprises and institutions. The urban housing system reforms in the 1990s spurred the expansion of urban residential districts driven by housing commercialization. These developments mark the neighborhoods that have existed for over 20 years. Since 2000, Harbin has been experiencing a notable population exodus, with census data showing an increase in population loss between 2010 and 2020. This phenomenon has further contributed to the inefficient residential landuse.

Therefore, this study targets the Ring Expressway Area of Harbin City, encompassing Nangang District, Daoli District, Daowai District, Xiangfang District, and Songbei District. Residential communities are taken as the fundamental unit of analysis. This study delves deeply into the spatial distribution characteristics of inefficient residential landuse within the Harbin City Roundabout Expressway and seeks to reveal the extent to which various factors contribute to this inefficiency, with the overarching aim of providing a scientific basis for the formulation of relevant redevelopment policies. Specifically, the research objectives of this paper include: (1) To construct absolute and relative indicators for assessing the efficiency of residential land use through the “two-step identification method” which incorporates tools such as Baidu heat map, POI, and other multi-source data to construct an evaluation indicator system. This system employs standard deviation ellipse and kernel density analysis to identify inefficient residential landuse and to examine its spatial distribution characteristics. (2) To analyze the contribution of each factor to inefficient residential landuse in different regions, we use the characteristics of Variable Importance Measures (VIMs) in the RF algorithm. This approach aims to scientifically and effectively analyze the causes of inefficient residential landuse formation. (3) To propose refined redevelopment strategies tailored to different types of inefficient residential landuse, considering regional characteristics and redevelopment objectives, with the expectation of providing robust decision support for the efficient and intensive utilization of urban residential land.

2. Study Area and Data Sources

2.1. Study Area

Harbin City, located in the south of Heilongjiang Province, consists of nine districts and nine counties. It has experienced a notable population decline, with the number of inhabitants decreasing from 5,541,500 to 5,519,300 between 2011 and 2021. Concurrently, the built-up area of Harbin City has expanded from 367 km² in 2011 to 491 km² in 2021, marking an increase of 33.79%. This expansion amidst a declining population highlights a mismatch between the supply of and demand for building land. The focus of this study is confined to five districts within Harbin’s ring expressway area: Nangang District, Daoli District, Daowai District, Xiangfang District and Songbei District within residential communities serving as the primary unit of analysis. At present, some residential communities (including family zones) within the study area are characterized by antiquated buildings and infrastructures, limited functionality within the communities, and substandard living conditions. This inefficiency in residential land use is starkly at odds with the aspirations for new town and city development, profoundly impacting the dynamism of the urban economy and impeding the broader development of the city (Figure 1).

2.2. Data Sources

The data utilized in this study comprise both residential communities vector data and geospatial big data. Firstly, the collection and processing of data from residential communities involved using the Python programming language to scrape vector data from the Anjuke real estate service platform (http://xa.anjuke.com, accessed on 2 September 2023). These data include detailed metrics such as plot ratio, average residential price and green space rate, and so on. Residential communities located outside the designated study area were systematically excluded from the collected data, resulting in a final sample of 2006 records accessed on 12 November 2023. The latitude and longitude coordinates of each community were then acquired and projected into geospatial space utilizing the Baidu Map API interface. Secondly, geospatial big data mainly includes Baidu heat map, POI (point of interest) and road network. The Baidu heatmap crawling using Baidu Map API (http://map.baidu.com, accessed on 5 September 2023). Subsequently, POI data of Harbin City were collected based on Baidu Map API on 6 September 2023, with fields including POI point type, latitude and longitude. Road network data were obtained from Open Street Map (http://download.geofabrik.de/, accessed on 20 May 2023).

3. Methodology

3.1. Fundamental Principles

Philosophically, attributes often embody both absolute and relative dimensions. This duality applies to residential land use efficiency, which encompasses aspects of both absolute and relative inefficiency. Absolute inefficient residential landuse aligns with specific criteria that indicate inefficiency in one or several respects. In contrast, relative inefficient residential landuse is defined as having significantly poorer surrounding facilities compared to other residential areas, thus failing to meet the normal needs of residents. Modern residential communities typically feature higher construction quality and more extensive amenities compared to older or less well-located neighborhoods, which are consequently at risk of becoming outdated and underutilized. Literature on this topic suggests a plethora of evaluation indicators and evaluation models for assessing inefficient residential landuse. However, these often omit critical aspects such as the structure of the road network. The placement and design of the road network within a residential community significantly influence the convenience of residents’ lives and, by extension, urban livability [46]. A well-connected road network enhances accessibility to various services and facilities, fosters community interconnectivity, and supports efficient land use. Moreover, an optimal road network structure enhances commercial activity, thereby boosting employment opportunities. Therefore, the structure of the road network has a significant impact on both the convenience of residents’ lives and the efficiency of residential land use. It is crucial to consider the road network structure into the identification of inefficient residential landuse. In the following evaluation indicator system, indicators such as “distance to high-quality enterprises” and “accessibility to commercial and service facilities” are included in the comprehensive analysis of the road network structure.

Drawing upon existing academic definitions and integrating the “15-min living circle” concept from the Planning and Design Standards for Urban Residential Areas (GB50180-2018) [47], this paper establishes an evaluation indicator system for inefficient residential landuse based on four dimensions: architectural attributes, economic environment, living environment and ecological environment. This evaluation system categorizes inefficient residential landuse into absolute and relative types and employs a “two-step identification method” for its assessment. The first step of the methodology mainly focuses on identifying absolutely inefficient residential landuse. And it involves applying predefined criteria to detect residential areas that meet specific benchmarks of absolute inefficiency. After we assess the other residential areas to see how they stack up. Then, we look at other areas more flexibly, considering architectural quality, economic conditions, living standards, and ecological features. The system calculates efficiency values, delineates varying levels of inefficiency, and ultimately classifies the least efficient areas as relatively inefficient residential landuse. This two-tiered approach allows for a nuanced understanding of inefficiency within residential land use, facilitating targeted urban planning and redevelopment strategies.

Based on the accurate identification of inefficient residential landuse, we conducted spatial analysis using standard deviation ellipse and kernel density estimation methods. Subsequently, we employed the RF algorithm to quantitatively analyze the driving factors of inefficient residential landuse in different regions. Finally, we proposed different redevelopment measures according to various situations. The workflow of this study is shown in Figure 2.

3.2. Construction of Evaluation Indicator System

Drawing from the inherent characteristics of inefficient residential landuse and adhering to principles of scientific rigor, comprehensiveness, practical applicability, and a blend of quantitative and qualitative approaches. This study created a system to evaluate how efficiently residential landuse is used (

Table 1. Evaluation indicator of residential landuse efficiency and its weight.

Type of Indicator	Sub-Dimension Layer	Evaluating Indicator	Level of Evaluation Indicators					Weight Value
			1	2	3	4	5
absolute indicator	building attributes	Plot ratio A1	<0.8					—
		Layout regularity A2	The basic style of residential layout includes determinant type, peripheral type, mixed type and freestyle type, which will not conform to the above four types of judgment as irregular layout					—
		Old residential communities A3	Screening of residential land in the area in accordance with the list of old communities in Harbin City					—
	ecological environment	Green space ratio B1	<30%					—
Relative indicator	building attributes	Building quality(age)/year C1	≤2000	>2000–2005	>2005–2010	>2010–2015	>2015	0.065
	economic environment	Average residential price /(CNY/m2) D1	>3826–5812	>5851–6419	>6421–7303	>7323–8735	>8763–18,590	0.165
		Distance to high-quality enterprises/(m) D2	≤320	>320–640	>640–960	>960–1200	≥1200	0.080
		Population concentration density/(person/hm2) D3	Ki ≤ 10	10 < Ki < 20	20 < Ki < 40	40 < Ki < 60	Ki ≥ 60	0.065
	living environment	Accessibility to commercial and service facilities/(points) E1	≤20	>20–40	>40–60	>60–80	≥80	0.095
		Accessibility to healthcare facilities/(points) E2	≤20	>20–40	>40–60	>60–80	≥80	0.095
		Accessibility to kindergarten and primary education facilities/(points) E3	≤20	>20–40	>40–60	>60–80	≥80	0.100
		Accessibility to public transportation/(points) E4	≤20	>20–40	>40–60	>60–80	≥80	0.075
		quality of public life/(points) E5	≤20	>20–40	>40–60	>60–80	≥80	0.125
	ecological environment	Walking distance from parks /(m) F1	≤320	>320–640	>640–960	>960–1200	≥1200	0.135

), aimed at distinguishing between absolute and relative inefficiencies.

Absolute indicators are comprised of plot ratio, layout regularity, the status of “old residential communities” and green space ratio. The explanations are as follows: (1) The plot ratio is the core indicator of land development intensity control, which plays a controlling role in determining development capacity [48]. A reasonable plot ratio is the key to ensuring balanced development intensity and living comfort. (2) Layout regularity is evaluated based on basic residential layout style in urban design principles. If the layout does not conform to these basic styles, the use of residential land is inefficient. Examples include “single buildings”, “scattered communities” and other scattered forms of residential land. (3) The old residential communities are defined by the government documents. These communities generally have problems such as poorly maintained public spaces, dilapidated infrastructure, and challenges in grid management. (4) The green space rate is an important indicator of the environmental quality of residential communities. According to the planning and design specifications, the green space ratios in new residential communities should not be less than 30%.

The relative indicators include building attributes (building quality), economic environment (average residential price, distance to high-quality enterprises, population concentration density), living environment (accessibility to commercial and service facilities, healthcare facilities, kindergarten and primary education facilities, accessibility to public transportation, and quality of public life), and ecological environment (walking distance from parks). The explanations are as follows: (1) Building Attributes: As residential communities ages, their quality tends to deteriorate. In this paper, the classification is based on construction year intervals: pre-2000 is considered the lowest level, with levels increasing in five-year intervals, and post-2015 being the highest level. (2) Economic environment: The economic environment of a region is a major driving force attracting people. Regions with relatively better economic environments are more attractive to the labor force, thereby enhancing the efficiency of residential land use in those areas. The average residential price acts as an important indicator for measuring the value of residential land, with higher average prices correlating with increased land value. The distance to high-quality enterprises reflects the spatial relationship between the residential land and high-quality enterprises, with closer proximity leading to heightened economic vitality and improved land efficiency. Population agglomeration density, which reflects the spatial distribution characteristics of the population, is another key factor, areas with lower population density typically exhibit lower residential land use efficiency. In this study, Baidu heat maps are utilized to represent population agglomeration density [49], as illustrated in

Table 2. Correspondence between population concentration density and Alpha channel values.

Color	$\mathbf{Population} \mathbf{Concentration} \mathbf{Density}\mathbf{/}{\mathit{K}}_{\mathit{i}}$	$\mathbf{Alpha} \mathbf{Channel} \mathbf{Value}\mathbf{/}{\mathit{S}\mathit{A}}_{\mathit{i}}$
Blue	≤10	$60 < {SA}_i$< 132
Pale blue		$132 < {SA}_i$< 138
Cyan		$138 < {SA}_i$< 151
Green	>10–20	$151 < {SA}_i$< 163
Yellow	>20–40	$163 < {SA}_i$< 170
Orange	>40–60	$170 < {SA}_i$< 179
Red	>60	$179 < {SA}_i$< 194

Note: $K_i$ is the Population concentration density/(person/hm²) of the $i$th raster; ${SA}_i$is the Alpha channel value of the $i$th raster.

. (3) Living environment: The living environment of residential land is intrinsically linked to the physical and mental health of residents. A good living environment is essential for ensuring safety and well-being. Factors such as accessibility to commercial and service facilities, healthcare facilities, kindergarten and primary education facilities, and public transportation, as well as the overall quality of public life, play a significant role in determining residents’ satisfaction with their community. (4) Ecological environment: Parks and other open spaces provide urban residents with places for leisure and recreation. Reducing the distance between residential communities and parks has a positive impact, indicating that the closer the proximity to parks, the better the ecological environment around the residential community.

3.3. Identification Criteria for Inefficient Residential Landuse

The inefficient residential landuse in this paper is divided into two types: absolutely inefficient residential landuse and relatively inefficient residential landuse. The criteria for identifying absolutely inefficient residential landuse: If the residential land meets one of the indicators in the absolute evaluation indicators (Table 1), it is directly judged to be absolutely inefficient residential landuse. The criteria for identifying relatively inefficient residential landuse: The indicators were initially ranked, then standardized and weighted in order to determine the lowest level of residential land as relatively inefficient residential landuse.

3.3.1. Criteria for Delineating Indicator Levels

Seven indicators, including distance from high-quality enterprises, accessibility to commercial and service facilities, accessibility to healthcare facilities, accessibility to kindergarten and primary education facilities, accessibility to public transportation, quality of public life, and walking distance to parks, all involve the distance from residential land to corresponding facilities. The walking speed was set at 80 m/min with reference to relevant studies [50], combined with the concept of a “15-min living circle” [51], the analysis interruption values in the GIS 10.6 software service area were set to 320 m, 640 m, 960 m, and 1200 m. Extract the maximum value in each raster cell, and grade from low to high, and the levels 1–5 correspond to [0, 20] points, (20, 40] points, (40, 60] points, (60, 80] points, (80, 100] points.

3.3.2. Date Standardization Processing

The dimensions or data types of the evaluation indicators are quite different, and we need to standardize the raw data in order to be comparable.

$$Y_{ij}=\left\{\begin{array}{c}\begin{array}{c}{\displaystyle \frac{X_{ij}-{minX}_{ij}}{max{X}_{ij}-min{X}_{ij}}}\end{array}\\ {\displaystyle \frac{max{X}_{ij}-X_{ij}}{max{X}_{ij}-min{X}_{ij}}}\end{array}\right.$$

(1)

where $Y_{ij}$ denotes the standardized value of $X_{ij}$, the standardized value of the $j$th indicator of plot $j$; max$X_{ij}$ and min$X_{ij}$ denote the maximum and minimum values in the original data of the $j$th indicator, respectively.

3.3.3. Method for Indicator Weighting

There are many methods for determining the weights, referring to the relevant research [52], a combination of quantitative and qualitative methods are used. We use a hierarchical analysis method and entropy weight method to calculate the weights of each indicator separately, and then calculate the combined weights, finally obtaining the weight values of each indicator.

$$W_{\mathrm{c}\mathrm{o}\mathrm{m}i}={\displaystyle \frac{W_{AHPi ·}{W}_{EWi}}{\sum W_{AHPi ·}{W}_{EWi}}}$$

(2)

where $W_{AHPi}$ denotes the weight calculated by the hierarchical analysis method of the $i$th indicator, $W_{EWi}$ denotes the weight calculated by the entropy weight method of the $i$th indicator, and $W_{\mathrm{c}\mathrm{o}\mathrm{m}i}$ denotes the value of the comprehensive weight of the $i$th indicator.

3.3.4. Determination of Relatively Inefficient Residential Landuse

The determination of relatively inefficient residential landuse is mainly based on Equation (3) to calculate the grade of relatively inefficient residential landuse, and the grade is divided into levels 1–5 by using the natural breakpoint method, in which levels 1, 2, 3, 4 and 5 represent inefficient, average, medium, good and efficient, respectively. The residential land with the level 1 is determined as relatively inefficient residential landuse.

$$\mathrm{S}=\sum _{i=1}^{\mathrm{n}}{X}_i·W_i$$

(3)

where $\mathrm{S}$ is the grade of residential land use efficiency, $X_i$ is the score of the $i$th indicator in the relative index, and $W_i$ is the weight of the $i$th indicator in the relative index.

3.4. Research Methods of Spatial Distribution Characteristics

3.4.1. Standard Deviation Ellipse

Standard Deviation Ellipse (SDE) can accurately and effectively reveal the geocentric, discrete, and directional trends of the spatial distribution of geographic elements. The standard deviation ellipse parameters mainly include rotation angle, long-axis standard deviation and short-axis standard deviation, which can reflect the spatial distribution difference and diffusion trend of inefficient residential land in the study area from multiple perspectives such as centroid, shape and direction.

$$SD{E}_x=\sqrt{{\displaystyle \frac{\sum _{i=1}^n(x_i-\overline{x}{)}^2}{n}}}$$

(4)

$$SD{E}_y=\sqrt{{\displaystyle \frac{\sum _{i=1}^n(y_i-\overline{y}{)}^2}{n}}}$$

(5)

$$\mathrm{t}an\theta ={\displaystyle \frac{\left(\sum _{i=1}^n{\overline{x_i}}^2-\sum _{i=1}^n{\overline{y_i}}^2\right)+\sqrt{\left(\sum _{i=1}^n{\overline{x_i}}^2-\sum _{i=1}^n{\overline{y_i}}^2\right)+4{\left(\sum _{i=1}^n(\overline{x_i}-\overline{y_i}\right)}^2}}{2\sum _{i=1}^n\overline{x_i}\overline{y_i}}}$$

(6)

where $\theta$ is the azimuth angle, the angle formed by the y-axis and the long axis of the ellipse; ($x_i,y_i$) is the spatial location of the geographic element; $\overline{x}$ and $\overline{\mathrm{y}}$ denotes the coordinate deviation of each geographic element to the center of the ellipse; ($SD{E}_x,SD{E}_y$) is the weighted mean center.

3.4.2. Kernel Density Analysis

Kernel density (KD) analysis is a widely used spatial analysis method in GIS analysis. In this paper, the change in density of the research features is measured by the established distance attenuation function to explore the sample distribution and change characteristics in the spatial region. The result is a smooth surface with a large median value and a small peripheral value, and the grid value is the unit density, which will be 0 at the domain boundary [53]. The specific formula is as follows.

$$D={\displaystyle \frac{{3(1-{scale}^2)}^2}{{\pi r}^2}}$$

(7)

where $r$ is the radius of the search, $scale$ is the ratio of the distance from the center of the raster to the object with points and lines to the radius of the search.

3.5. Importance Ranking

Random Forest (RF) constructs multiple decision tree models by randomly sampling data and selecting features. For each decision tree construction, RF utilizes a bootstrap sampling method to generate different training sets by randomly sampling from the original dataset. The advantage of RF over traditional hypothesis-test models rests in its nonparametric nature, its capability to allow for effective analysis of data that is nonlinear, incorporating covariates and interactions, and providing a variable importance measure (VIM). Among them, one widely used method to evaluate the importance of features is the percentage increase of mean square error (%IncMSE). That evaluates the importance of features by randomly scrambling the value of each feature in the model and observing changes in the prediction accuracy of the model. In this paper, the relative importance of various factors in inefficient residential landuse is mainly reflected through the %IncMSE value. The higher the value, the more important this factor is for the formation of inefficient residential landuse.

4. Results

4.1. Identification of Inefficient Residential Landuse

4.1.1. Identification of Absolutely Inefficient Residential Landuse

Figure 3 and

Table 3. Statistics of inefficient residential landuse.

Region	Absolutely Inefficient Residential Landuse		Relatively Inefficient Residential Landuse		Total
	Number	Area (km2)	Number	Area (km2)	Number	Area (km2)
Daowai District	43	0.98	43	1.78	86	2.76
Xiangfang District	68	1.32	89	3.83	157	5.15
Nangang District	103	5.31	102	3.60	205	8.91
Daoli District	67	2.97	40	1.88	107	4.85
Songbei District	6	0.44	50	4.04	56	4.48
Total	287	11.02	324	15.13	611	26.15

illustrate the spatial distribution, quantity, and area characteristics of inefficient residential landuse. We observed that the absolutely inefficient residential landuse is primarily located in the city center. Moreover, relatively inefficient residential landuse is mainly distributed in a “perforated” pattern within the study area. Table 3 reveals that the total number of absolutely inefficient residential landuse parcels in the study region amounts to 287, covering an area of 11.02 km², which accounts for 42.14% of the total inefficient residential landuse area. From the perspective of each district, the absolutely inefficient residential landuse is primarily concentrated in Nangang District, with 103 parcels covering the largest area, accounting for 48.19% of the total area of absolutely inefficient residential landuse. Following Nangang District are Xiangfang District and Daoli District, which account for 11.98% and 26.95% of the total area of absolutely inefficient residential landuse, respectively. Although Xiangfang District has a great number of absolutely inefficient residential landuse than Daoli District, its total area is smaller by 1.65 km². Daowai District has a total of 43 parcels of absolutely inefficient residential landuse, accounting for 8.89% of the total area of absolutely inefficient residential landuse. Songbei District has the fewest number of absolutely inefficient residential landuse, with the smallest total area of absolutely inefficient residential landuse.

Table 4. Statistics on the number of factors in absolutely inefficient residential landuse.

Type	Number	Area (km2)	Type	Number	Area (km2)	Type	Number	Area (km2)
A1	39	0.97	A1 ∩ A3	16	0.23	A1 ∩ A2 ∩ A3	1	0.01
A2	0	0.00	A1 ∩ A4	1	0.01	A1 ∩ A2 ∩ B1	0	0.00
A3	93	5.85	A2 ∩ A3	11	0.15	A2 ∩ A3 ∩ B1	0	0.00
B1	113	3.64	A2 ∩ B1	0	0.00	A1 ∩ A3 ∩ B1	1	0.01
A1 ∩ A2	0	0.00	A3 ∩ B1	12	0.15	A1 ∩ A2 ∩ A3 ∩ B1	0	0.00

reveals the presence of many old residential communities and those with substandard green space ratio in the study area. Old residential communities are prevalent and extensive, accounting for 46.69% of the total number of absolutely inefficient residential landuse and 57.14% of the total area of absolutely inefficient residential landuse. Additionally, there exist 113 residential lands classified as absolutely inefficient residential landuse due to substandard green space ratio, covering an area of 3.64 km². Among all absolutely inefficient residential landuse, 42 residential lands (0.56 km²) were classified as absolutely inefficient residential landuse due to substandard performance on two or more indicators, indicating that some cases of absolutely inefficient residential landuse are formed due to multiple deficiencies and require comprehensive improvement from multiple perspectives.

4.1.2. Identification of Relatively Inefficient Residential Landuse

Table 3 reveals that the total number of relatively inefficient residential landuse in the study region amounts to 324, covering an area of 15.13 km², which accounts for 57.86% of the total residential land area. From the perspective of each district, relatively inefficient residential landuse is primarily concentrated in Nangang District, which contains 102 parcels, accounting for 23.79% of the total area of relatively inefficient residential landuse. Following Nangang District is Xiangfang District, with 89 parcels, accounting for 25.31% of the total area of relatively inefficient residential landuse. Moreover, compared to the number of absolutely inefficient residential landuse, Songbei District has a greater number of relatively inefficient residential landuse, accounting for 26.70% of the total relatively inefficient residential landuse area. Although Daowai District has more relatively inefficient residential landuse parcels than Daoli District, its total area is smaller by 0.1 km². Specifically, Daowai District has 43 parcels, accounting for 11.76% of the relatively inefficient residential landuse area.

Table 5. Statistics on the number of factors in relatively inefficient residential landuse.

Type	Level 1		Level 2		Level 3		Level 4		Level 5
	Number	Area (km2)	Number	Area (km2)	Number	Area (km2)	Number	Area (km2)	Number	Area (km2)
C1	75	1.82	43	1.12	62	1.99	63	3.32	81	6.88
D1	108	4.09	83	3.81	60	2.84	50	2.87	23	1.52
D2	244	11.43	25	1.32	24	0.83	22	1.10	9	0.45
D3	53	3.05	72	3.75	170	7.27	25	0.96	4	0.10
E1	50	2.78	19	1.08	64	3.76	84	3.66	107	3.85
E2	94	5.93	49	2.03	61	2.80	64	2.78	56	1.59
E3	253	12.44	27	1.04	30	1.05	9	0.31	5	0.29
E4	165	7.82	41	1.59	44	2.47	50	2.45	24	0.80
E5	32	0.42	23	1.29	16	0.67	54	3.10	199	9.65
F1	297	13.74	2	0.02	13	0.75	6	0.35	6	0.27

reveals that the quality of public life is generally excellent in the study region, as only 32 residential lands are classified as relatively inefficient residential landuse due to poor quality of public life, accounting for about 9.88% of the total number of relatively inefficient residential landuse. In terms of the economic environment, the population concentration density is relatively high, with only 53 residential lands classified as relatively inefficient residential landuse due to poor density of population concentration density. In comparison, 165 residential lands are classified as relatively inefficient residential landuse due to inconvenient public transportation. Most relatively inefficient residential landuse also face issues such as poor accessibility to kindergarten or primary education, and long distances from high-quality enterprises and parks.

4.2. Spatial Distribution Characteristics of Inefficient Residential Landuse

4.2.1. Spatial Location and Orientation Characteristics of Inefficient Residential Landuse

The ArcGIS 10.8 platform was used to convert the inefficient residential map spots in the study region into inefficient residential landuse points. The results of the elliptic standard deviation analysis of the inefficient residential landuse points in each district are shown in Figure 4. As can be seen from Figure 4 and

Table 6. Standard deviation ellipse parameters of inefficient residential landuse.

Region	Area (km2)	Center X	Center Y	XStdDist /m	YStdDist /m	Azimuth Angle/(°)
Daowai District	21.47	41,786,224.68	5,078,256.12	4404.06	1551.73	89.85
Xiangfang District	27.72	41,786,193.03	5,071,500.65	3660.87	2441.46	70.46
Nangang District	40.13	41,782,654.24	5,072,288.67	2143.50	5960.29	33.16
Daoli District	20.94	41,778,809.98	5,074,146.35	4366.25	1526.53	50.60
Songbei District	38.69	41,776,305.51	5,082,013.30	5833.06	2111.98	73.85
Ring Expressway Area	118.21	41,782,755.06	5,074,286.30	6447.23	5836.27	121.91

, the distribution characteristics of inefficient residential landuse in each district are inconsistent. In terms of the spatial distribution characteristics of inefficient residential landuse within the Ring Expressway Area, the concentration of elliptical center points is predominantly situated in the central urban region. The difference between the x-axis and y-axis measures 610.96, indicating a discernible “northwest-southeast” trend, representing the orientation of inefficient residential landuse distribution across the five districts, ranges from 30° to 90°. The sequence of these angles, in descending order, is as follows: Daowai District > Songbei District > Xiangfang District > Daoli District > Nangang District.

In terms of elliptical area, the standard deviation ellipse area of inefficient residential landuse points in Nangang District was the largest, which is 40.13 km², indicating that the distribution of inefficient residential landuse in Nangang District was relatively scattered. Conversely, the standard deviation ellipse area of inefficient residential landuse points in Daoli District is the smallest, which is 20.94 km², signifying a more compact distribution of inefficient residential landuse in this district. Analyzing the relationship between the long and short axes of the ellipses reveals that the disparity between the x-axis and y-axis in the standard deviation ellipse for Nangang District is the largest. This means the inefficient landuse there stretches out more in certain directions, showing a clear pattern of how it spreads. In contrast, Xiangfang District exhibits the smallest difference between the x-axis and y-axis, indicating minimal directionality and a more compact distribution pattern of inefficient residential landuse.

4.2.2. Aggregation Form Characteristics of Inefficient Residential Landuse

As illustrated in Figure 5, the inefficient residential landuse in the study region is predominantly concentrated in the central city, with sporadic distribution in all other areas, and the degree of aggregation is less pronounced outside the central city. When examining each district individually, the aggregation patterns exhibit slight variations. The density of nuclear centers varies significantly across districts. Xiangfang District displays the highest density of nuclear centers, with density values ranging from 0–18.94 per square meter, while Songbei District shows the lowest density of nuclear centers, with density values ranging from 0–3.14 per square meter. Further analysis reveals that Daowai District, Xiangfang District and Daoli District form two aggregation cores, whereas Nangang District and Songbei District each form one core. In Daowai District, the primary gathering centers are located near the intersections of “Taikoo Street” and “South Seventh Street” (DW1), and the intersections of “Hongtu Street” and “Nanzhi Lu” (DW2). In Xiangfang District, the main gathering centers are situated in the northwestern part, near the intersections of “Hexing Road” and “Wenchang Street” (XF1), and “CaiyiStreet” and “Gongbin Road” (XF2). The gathering centers of Nangang District are mainly located in the northern part of this district, the gathering core are mainly distributed along “Xuanhua Street” and “Wenchang Street” (NG1). Daoli District shows a “dual-core concentration” pattern, with cores near “Xinyang Road” and “Kang’an Road” (DL1), and “Xinyang Road” and “Anfa Street” (DL2). Songbei District mainly features a large central cluster with smaller scattered ones, with the main hub at “Songbei Avenue” and “Zhongyuan Avenue” (SB1).

4.3. Importance Ranking of Influencing Factors

As shown in Figure 6f, the top three contributing factors to inefficient residential landuse within the Ring Expressway Area are “old residential communities” (A3), green space ratio (B1) and distance to high-quality enterprises (D2). Specifically, there is significant difference in the contribution of factors affecting inefficient residential landuse between Songbei District and other districts. The main reasons for inefficient residential landuse vary across different districts. We can find from Figure 6e that the top three factors are accessibility to healthcare (E2), distance to high-quality enterprises (D2) and accessibility to kindergarten and primary education facilities (E3) in Songbei District. For Daoli District, we can find from Figure 6d that the primary contributing factors are green space ratio (B1), “old residential communities” (A3) and accessibility to healthcare facilities (E2). The top three contributing factors to inefficient residential landuse within the Xiangfang District are accessibility to healthcare facilities (E2), “old residential communities” (A3) and green space ratio (B1) in Figure 6b. The top three contributing factors to inefficient residential landuse within the Nangang District are green space ratio (B1), average residential price (D1) and accessibility to healthcare facilities (E2) in Figure 6c. The top three contributing factors to inefficient residential landuse within the Daowai District are “old residential communities” (A3), accessibility to public transportation (E4) and green space ratio (B1) in Figure 6a.

A comparative analysis shows inefficient residential landuse in most districts is primarily due to absolute factors, whereas in Songbei District. Specifically, in Daowai District, Xiangfang District, and Daoli District, the top contributing factors include “old residential communities” (A3) and green space ratio (B1). Nangang District’s inefficient landuse is mostly due to its green space (B1). In contrast, Songbei District’s inefficiency is driven mainly by factors related to the economic environment and living environment. It is also obvious that different parts of the living environment affect how inefficient residential landuse forms in various districts in different ways. Accessibility to healthcare facilities (E2) plays an important role in the inefficient residential landuse in Xiangfang District, Nangang District and Daoli District. This highlights how different aspects of the living environment impact residential land use differently across districts.

5. Discussion

5.1. Spatial Analysis of Inefficient Residential Landuse

In the study area, inefficient residential landuse is predominantly located in the city center. Nangang District and Xiangfang District exhibit a high number of inefficient residential landuse. Analyzed in terms of spatial characteristics, the main reason for this is due to multiple factors such as the history of urban development and topography that contribute to this phenomenon. From a historical perspective, Xiangfang District has a long history of development, primarily dominated by wood processing and raw material processing enterprises, with family buildings constructed near these factories. Considering the impact of topographical factors, the eastern part of Xiangfang District mainly consists of farmland and forest land, while the southern part experiences significant elevation changes. Conversely, the western part is relatively flat and more suitable for human habitation. Consequently, inefficient residential landuse is mainly concentrated in the city center, a pattern also observed in Nangang District. Among the relatively inefficient residential landuse, apart from Nangang District and Xiangfang District, Songbei District holds a significant proportion. As a newly developed area, Songbei District’s newly built communities meet national standards and residents’ requirements, but the construction of service facilities remains in its early stages. Additionally, the Songhua River creates a topographical barrier between Songbei District and the city center, limiting the reach of many city services to Songbei District. This barrier contributes to the relatively high amount of inefficient residential landuse in Songbei District. These factors culminate in a severe conflict between residents’ demands and the supply of housing. This mismatch affects social harmony and development and suppresses the local economy. Socially, the community’s lack of adequate infrastructure and public service facilities hinders the effective assurance of residents’ quality of life [54]. This situation can easily lead to residents’ dissatisfaction and social conflicts, thereby affecting social stability. Economically, the presence of inefficient residential land signifies a waste of land resources and low economic efficiency, impeding sustainable urban development and economic growth [55]. This misallocation of resources not only restricts the city’s development potential but also increases costs for the government and businesses [56].

5.2. Analysis of the Driving Factors behind Spatial Differences in Inefficient Residential Landuse

Although the relative importance of various factors affecting the efficiency of residential land use may fluctuate with different stages of urban development, certain similarities are notable [57,58]. Firstly, old neighborhoods in city centers play a key role in inefficient residential landuse. These findings align with previous studies that emphasize the impact of old neighborhoods on residents’ living experiences [59]. There are two main reasons for this situation. One reason is that these central urban areas were developed earlier, and many communities and infrastructures have aged. Over time, the number of old residential communities has increased annually. Another reason is that the city’s development history is reflected in the changes in residential land use. Harbin, being an old heavy industrial city, hosts many factory communities. As old factories shut down, the utilization efficiency of nearby family communities significantly decreases, turning them into old factory family communities. For example, the “Xingguang Family Community” in XF1 has an average housing price of CNY 5233/m². This community was built in 1977 and has poor basic infrastructure, such as water and electricity supply, making it a typical old neighborhood. Similarly, the “Meitian Family Building” and other factory family communities in XF2 were built in 1980. Many facilities are outdated and unusable, leading to these old neighborhoods being left idle. This characteristic is also observed in other industrial cities like Hegang City. Secondly, the amount of green space significantly impacts how efficiently residential land is used in city centers [60,61]. This is mainly due to the unresolved conflicts between early urban development and current residents’ living experiences. Due to the scarcity of land resources in central urban areas, early urban development aimed to accommodate more people, resulting in high development intensity and building density. This led to a lower green space ratio and less open space around residential lands. However, as societal expectations have evolved, residents now place greater importance on having green, open spaces within their living environments. For example, in Nangang District’s NG1, the “Shangfang Community” area focuses on commercial development. The community’s greenery levels are poor, failing to provide residents with fixed places for leisure and interaction. This ultimately leads to lower residential land use efficiency in that area.

In new development areas, both living and economic environments significantly impact the efficiency of residential land use, aligning with findings from previous research. It is widely acknowledged that proximity to the city center enhances access to social resources such as commercial services, healthcare facilities, and employment opportunities, thereby directly improving the efficiency of residential land use in surrounding areas [62,63]. This study highlights that inefficient residential land frequently appears in Songbei District, where living and economic environments are less favorable. In Songbei District, “Beian Xinghe community” is located near the northernmost end of the Ring Expressway, far from the city center. The lack of medical service facilities necessitates that residents travel longer distances to access healthcare services, often to the city center or other areas. This situation not only increases healthcare costs for local residents but also diminishes their willingness to reside in the area, thereby contributing to the classification of this residential community as inefficient residential land. Additionally, the availability of educational resources around a residential community can also inversely affect its usage efficiency. It is well established that families prefer communities with access to high-quality educational resources to ensure their children receive superior education. The lack of adequate primary education facilities in Songbei District severely impacts the efficiency of residential land use in the area. The focal point of inefficient residential land aggregation (SB1) includes “Jingyi Garden” which lacks primary and secondary schools. The nearest secondary school is 2.5 km away, posing challenges for residents in terms of their children’s education. This ultimately categorizes the residential land in this area as inefficient. This outcome can be attributed to several factors. On one hand, Songbei District is a newly developed area where infrastructure development, such as hospitals and schools, lags behind compared to the city center. This relative delay in public service facilities contributes to the inefficiency of residential land use. On the other hand, Songbei District is still in the early stages of economic development. Therefore, it may not attract as many high-quality enterprises as the central city, which has already established an agglomeration effect. New development zones typically require more time to cultivate a favorable business environment and attract top-tier businesses, thereby affecting employment opportunities and the overall business environment attractiveness in Songbei District.

5.3. Strategies for Redeveloping Inefficient Residential Landuse

Actively redeveloping inefficient residential landuse is crucial to bringing greater benefits and value to local residents and the entire city. For absolutely inefficient residential landuse, the following three recommendations are proposed: Firstly, to control land development intensity, implementing floor area ratio incentives and transfer policies based on the principle of “equivalent value transfer” is essential [64,65]. This balances developers’ economic interests with the long-term needs of urban planning. In balancing the relationship between government and the market, we can draw lessons from the diversified model paths in the Pearl River Delta region of China. Local efforts there aim to activate existing land resources by actively enhancing market value and encouraging multiple stakeholders to participate. Furthermore, considering the specific needs of different projects and the evolving market dynamics, we recommend adopting appropriate flexible adjustment strategies [66]. Secondly, there is a need for integrating scattered land parcels. To improve residents’ quality of life, the Shenzhen Special Economic Zone manages scattered residential land parcels through scientific management to enhance residential land use efficiency. Therefore, implementing strategies such as “integrating points and lines, leading small by large, and collaborative rectification” can transform scattered forms of residential plots like standalone buildings and dispersed neighborhoods, into multifunctional spaces [67]. This approach facilitates the reconfiguration of land functions. Finally, it is essential to enhance the quality of residential communities. The renewal and transformation of inefficient residential land should comprehensively consider factors such as community demographics, functional needs, and ecological environment. Specifically, for well-located, largescale, early-developed residential areas, it is feasible to reinforce building structures, repair bodies, and upgrade facilities without changing the buildings’ basic appearance. Thus enhancing safety and functionality, such as by installing elevators or introducing smart management systems, can improve residential land quality. Simultaneously, it is crucial to actively solicit residents’ opinions and requests during the construction process [68]. Drawing from Beijing’s typical neighborhoods, the Ecological Responsibility Planning Team was created to enhance effective communication between the government and residents. This team not only gathers residents’ feedback and identifies their core needs but also significantly reduces communication barriers between residents and government departments.

Furthermore, we propose the following three recommendations to improve relatively inefficient residential landuse. Firstly, enhance building quality. For communities characterized by early construction and substandard building integrity, comprehensive structural and functional renovations are mandatory for both residential and public edifices. Enhancing resident safety during transit, which includes installing slip-resistant pavements, adding safety handrails, and utilizing energy-efficient, environmentally friendly, and safe technologies throughout the renovation process. Secondly, stimulating the economic environment. This involves governmental initiatives to bolster economic infrastructure, cultivate a conducive market milieu for businesses, and foster robust market growth. Proactively attracting top-tier enterprises and beneficial commercial projects, coupled with the implementation of incentives like tax concessions, can invigorate economic development and attract investment. Lastly, improve the living and ecological environment around residential communities. Renovations in urban areas should really tap into how people live their daily lives, including their usual behaviors and the surrounding environment. The goal is to make the area more welcoming and lively. For instance, improvements could include upgrading fitness and cultural facilities to support the concept of the “15-min living circle”, which ensures residents have access to essential services within a short walk or bike ride from their homes. Adding green touches to the streets, creating small “pocket parks”, and expanding recreational spaces can all help form a connected green ecological network. Moreover, combining smart ecological landscape design concepts with local cultural can create green spaces that are not only cosmetically appealing but also environmentally friendly [69].

5.4. Advanced Methods for Analyzing Inefficient Residential Landuse

Traditional identification of inefficient land use mainly relies on field surveys or remote sensing imageries. However, while remote sensing imageries can accurately identify land use types, they cannot show human activities or gauge residents’ satisfaction with their living conditions. On the other hand, while field surveys can address this issue effectively, they come with drawbacks like high operational demands and limited spatial coverage. Using big data and machine learning has changed how we identify and analyze inefficient residential landuse. These technologies improve the accuracy of determining which factors are most important. With the advancement of big data, innovative inputs such as Point of Interest (POI) data, real-time dynamic data of Baidu Heat Maps and mobile phone signals have become a key focus in modern research on urban studies. In this study, Baidu Heat Maps are effectively integrated with population density data to laterally assess the economic vitality surrounding neighborhoods, enriching the evaluation indicators for inefficient residential landuse. In analyzing driving factors, while some traditional methods like Ordinary Least Squares (OLS) and logistic regression models are often used, they often overfit with complex data and nonlinear relationships, and have various limitations [35,70]. In contrast, the Random Forest (RF) algorithm stands out as a superior variable importance estimator, noted for its robustness, accuracy, and simplicity in comparison to conventional models. The RF algorithm’s broad applicability is derived from its proficiency in analyzing a vast array of complex nonlinear relationships. Additionally, distinguished from other prevalent machine learning methods, RF is adept at handling classification tasks, processing data features, accommodating missing values, and filtering redundant features. Through using Variable Importance Measures (VIMs), the RF algorithm produces reliable assessments of the contributions of driving factors. This method helps to understand which factors are most influential and how they influence each other in various scenarios. Given these advantages, we used the RF algorithm in this study to assess the relative importance of various determinants related to inefficient residential landuse.

5.5. Adaptability and Considerations of the RF Algorithm in Urban Land Analysis

With advancements in technology and machine learning, the use of the RF algorithm in urban land studies is expected to experience significant growth. Its ability to handle diverse data types and applications underscores its considerable adaptability in this field. Firstly, RF can process diverse data types, including both numerical and categorical data. In the analysis of inefficient residential landuse, for instance, the data sources include population density, economic activity indicators, and environmental quality metrics, which are heterogeneous in nature. RF’s adaptability allows it to integrate and analyze these varied data sources effectively, providing meaningful insights. Moreover, RF’s versatility extends to different fields and applications. Beyond inefficient residential landuse analysis, RF can be applied to urban land expansion [37], ecological environment monitoring [71], urban traffic analysis [72], and other areas. This broad applicability highlights RF’s capability to address complex problems across various domains. Despite these advantages, several caveats must be considered when applying the RF model. One key issue is the model’s computational complexity. RF requires the training of multiple decision trees, which involves extensive data sampling and feature selection for each tree. This process can result in high computational demands and extended training times [73]. Therefore, when dealing with large-scale datasets, it is essential to allocate sufficient computational resources to support the training and validation processes effectively. Another important consideration is the tuning of hyperparameters. The performance of the RF model is significantly influenced by hyperparameters such as the number of trees (n_estimators), the maximum depth of each tree (max_depth), and the number of features considered for each split (max_features). Selecting optimal hyperparameters is crucial for achieving the best model performance and computational efficiency. Typically, this involves using cross-validation techniques to systematically adjust and evaluate the hyperparameters, ensuring that the model achieves optimal settings for various scenarios [74,75]. In summary, while the RF model offers substantial adaptability and advantages in analyzing inefficient residential landuse, careful attention must be given to its computational demands and hyperparameter tuning to fully leverage its potential.

5.6. Limitations and Research Outlook

The formation of inefficient residential landuse is an exceedingly complex process that often exhibits striking similarities and patterns, as confirmed by numerous studies. Despite its contributions, this study also presents several limitations that offer avenues for future research. Firstly, the scope of the study was confined to residential land within the inner ring of Harbin delimited by the Ring Expressway, without extending to the entire city of Harbin. This was mainly because the city center of Harbin represents the central economic development hub, and the efficiency of residential land within the city can provide insights into quality of life for urban residents. However, with the ongoing trends of urban–rural integration, residents’ lives are gradually expanding beyond the urban core. Therefore, future research should encompass a comprehensive study of the entire city or urban agglomerations to capture a more holistic understanding of residential land use efficiency. Secondly, this study established an evaluation index system based on four aspects: building attributes, economic environment, living environment, and ecological environment. However, the identification of inefficient residential landuse also involves a complex interaction between residents and the government. Neighborhoods listed by the government as “old residential communities” may not necessarily be perceived as old or inefficient by residents themselves, and vice versa.

Therefore, evaluating inefficient residential landuse needs to consider the perspectives of different stakeholders, and future research should supplement relevant indicators. Moreover, while the application of big data in urban studies promises enhanced accuracy and real-time data access, this study was limited by the availability of specific datasets. For instance, data on parking lot usage, nighttime taxi terminal records, and other travel-related information were not utilized due to the absence of comprehensive travel data sets in downtown Harbin. However, as data availability improves, future studies could leverage these smart data platforms to enhance the depth and breadth of urban research. Finally, in terms of redevelopment of inefficient residential landuse, this study primarily proposes targeted refurbishment suggestions based on current residential land issues identified within residential land. It lacked a broader macroscopic analysis that could guide holistic the redevelopment strategies. Future redevelopment of inefficient residential landuse should adopt an “adaptive totalization” approach. In terms of policy support, technical specifications, mechanism building, and land assessment, it is imperative to engage in effective information sharing, planning and cohesion, and stakeholder coordination. The formation of inefficient residential landuse involves an intricate interplay involving economic, social, and natural factors. Academic inquiry into these dynamics spans a diverse array of disciplines such as urban and rural planning, geography, and land resource management. Addressing the redevelopment of urban inefficient residential landuse necessitates a multidisciplinary approach that integrates the expertise of urban planners, economists, and environmental scientists. This team effort is essential to broaden the scope and enhance the efficacy of redevelopment strategies. The RF algorithm also presents broad potential for analyzing the driving factors behind urban land use changes. In the future, it will be valuable to compare RF with other machine learning models, such as Support Vector Machines (SVM) and Gradient Boosting Trees (GBT), to assess their performance and suitability across different datasets and application scenarios. This comparative analysis will further validate RF’s strengths and refine its applications. Additionally, integrating time series analysis to examine the spatiotemporal variations in driving factors will enhance the capability of RF to monitor and predict dynamic changes in inefficient residential landuse. This approach promises to significantly improve the practical relevance and forward-looking potential of the research, offering deeper insights into the evolving patterns of land use inefficiency. In the end, this combined approach aims to use land more efficiently and improve urban environments overall.

6. Conclusions

In this study, we identified the inefficient residential landuse in the Ring Expressway Area of Harbin City using the “two-step identification method”, and analyzed its spatial distribution characteristics. The RF model was employed to determine the contribution of various factors to the formation of inefficient residential land. Based on these analyses, we came up with specific redevelopment strategies for different types of inefficient residential landuse. The principal conclusions of the study are as follows:

(1)

In the study area, inefficient residential landuse is mainly concentrated in the city center. Influenced by factors such as topography and urban historical development, residential land in Nangang District and Xiangfang District is predominantly located in the city center, leading to a higher prevalence of inefficient residential landuse in these areas. Additionally, Songbei District is notably affected by underdeveloped service facilities and geographical barriers, such as the Songhua River, which restrict the extension of central city services to this district. These lead to the higher proportion of relatively inefficient residential landuse in Songbei District.

(2)

The causes behind inefficient land use vary widely across different regions. In areas other than Songbei District, primary contributors to inefficiency include established residential communities and insufficient green space ratios. These issues often stem from outdated industrial locations and archaic housing design principles. In contrast, in Songbei District, inefficiencies in residential land primarily arise from deficiencies in essential services like healthcare and education. This pattern is typical in the early development phases of many new urban areas, where infrastructure lags behind residential growth.

(3)

Taking proactive redevelopment measures is crucial for enhancing land utilization efficiency and improving residents’ living experience. To improve absolutely inefficient residential landuse, comprehensive transformations should focus on controlling land development intensity, integrating scattered plots, enhancing community quality, and improving ecological environments. Conversely, for relatively inefficient residential landuse, the focus should be on upgrading building quality, enhancing economic, living, and ecological environments around neighborhoods.