A CPSO-BPNN-Based Analysis of Factors Influencing the Mental Health of Urban Youth ^†

Abstract

The fast-paced lifestyle, high-pressure work environment, crowded traffic, and polluted air of urban environments often have a negative impact on urban youth’s mental health.Understanding the factors in urban environments that influence the mental health of young people and the differences among groups can help improve the adaptability and mental health of urban youth. Based on the 2024 report on the health status of urban youth in China, this paper first analyzes this through a combination of multiple linear regression and automated machine learning methods. The key influencing factors of different living styles and environments on the mental health of urban youth and the priority of influencing factors are evaluated. The results are obtained by using the chaos particle swarm optimization-based back propagation neural network (CPSO-BPNN) model. Then, the heterogeneity of the different types of urban youth groups is analyzed. Finally, the conclusions and recommendations of this article are presented. This study provides theoretical support and a scientific decision-making reference for improving the adaptability and health of urban youth.

1. Introduction

The fast-paced lifestyle and highly competitive pressures of modern cities have made mental health an increasingly important topic. The changes in urban environment have had a profound impact on urban youth’s mental health, which is not only reflected in their working and living conditions, but also in their demands for social relationships, physical and mental health, and self-development [1]. Urban planners, policy makers, and individuals themselves should all pay attention to it. Efforts should be made to improve the urban environment and provide urban youth with healthier and more harmonious living spaces. In recent years, research on the mental health of urban youth has also received widespread attention.

From the perspective of urban environments, prevention and early intervention in youth mental health was studied in [2]. The impact of climate change on mental health was investigated in [3], which reveals that the effects of climate change can be direct or indirect, short-term or long-term. Urban adolescent mental health requires attention from decision makers as well as advocates who seek to establish sustainable cities. The opportunities to increase the prominence of urban adolescent mental health in global health were examined in [4]. The association between baseline substance use and mental health and non-partner violence trajectories among youth was studied in [5]. Utilizing structural equation modeling, the independent and combined associations among parent and adolescent variables hypothesized to be associated with the youth’s engagement in mental health treatment was explored in [6]. A novel conceptual framework to bridge public health, planning, and neurourbanism was proposed in [7] to analyze how urban environments affect young people’s mental health. The prevalence of violence exposure and associated mental health consequences among urban and non-urban youth was examined in [8]. The relationships between urban environments and psychiatric symptoms was studied in [9]. Shaping the aspects of urban life that influence youth mental health could have an enormous impact on adolescent well-being and adult trajectories, as stated in [10]. However, the above studies did not complete quantitative analyses.

Urban environments are becoming increasingly important determinants of health. Using a qualitative approach, the relationships between specific urban designs and adolescent mental health indicators was investigated in [11]. Mental health and physical health are closely related [12]. The prevalence of violence exposure and associated mental health consequences among urban and non-urban youth was investigated in [8]. Using on-site ecological momentary assessment surveys and adjusted linear mixed models, associations between adolescent mental health indicators and multiple pedestrian design and architecture concepts was explored in [13].

In a systematic review, using deep learning techniques, the influence of outdoor environments on human perceptions and emotions, with an emphasis on mental health outcomes, was investigated in [14]. Applying geographic ecological momentary assessment and mental models to enhance pro-environmental behaviours and mental health through nature contact for urban youth was proposed in [15]. The impact of meaning in life on loneliness, thriving, and well-being was investigated [16]. A qualitative exploration of the built environment as a key mechanism of safety and social cohesion for youth in high-violence communities was studied in [17].

Note that there are many existing studies that separately analyze the impact of living habits or living environments on the mental health of urban youth, and few that systematically consider both factors related to living under the same theoretical framework. The mental health of urban youth is a process in which various factors interact and regulate each other, and traditional linear operation models cannot clearly explain its internal relationships and functions. Using statistical methods may lead to deviations between the results and the actual situation, thereby imposing limitations on induction and application. Neural networks have more advanced learning mechanisms and can adapt to more complex nonlinear problems.

Currently, there is a lack of objective methods for mental health evaluation, and most of the available methods are limited to subjective face-to-face discussions. To this end, an objective method for mental health evaluation using a combination of convolutional neural networks and long short-term memory algorithms was introduced in [18]. A deep learning-based mental health monitoring scheme for college students was designed in [19], where a model used the most efficient convolutional neural network to classify the mental health status. In [20], a facial emotion recognition method for college student’s mental health diagnosis was proposed.

These observations inspire our current study. In this paper, the influencing factors on the mental health of urban youth are analyzed using a chaos particle swarm optimization-based back propagation neural network (CPSO-BPNN). Furthermore, the heterogeneity of these influencing factors among different urban youth populations is explored. The main contributions of this paper are summarized as follows: (i) A multiple linear regression model based on ordinary least squares (OLS) to evaluate the priority of the impact of living habits and the living environment on the mental health of urban youth. (ii) A CPSO-BPNN with automated machine learning method is used to further reduce the analysis dimensions and remove the insignificant factors. (iii) The descriptive statistical results from the observed samples are first analyzed. By using CPSO-based BPNN, some insignificant factors are removed and the key influencing factors and the priority are evaluated. (iv) The conclusions and recommendations are presented to provide theoretical support and a scientific decision-making reference for improving the adaptability and health level of urban youth.

The architectural diagram of the article is shown in Figure 1.

2. Data and Model Analysis

2.1. Data Source

The data used in this article are sourced from the 2023 report on the health status of urban youth in China and the tracking survey data on factors influencing the health of urban youth. This survey is jointly conducted by the National Center for Mental Health Research and several social science academies of universities. It is the first multidimensional study in China tracking the psychological health and social behavior of urban youth. It adopts a random sampling method, covering the 25–30-year-old youth group in first tier and new first tier cities across the country (including 15 cities such as Beijing, Shanghai, Guangzhou, and Shenzhen). The effective sample size reached 12,350, and the group composition included typical urban youth subgroups such as local youth with registered residence registration (38.5%), migrant youth (30.5%), and college graduates (17.5%). It has a good representativeness and scientific basis. The data collection was completed through mixed online and offline modes, covering six core modules including mental health assessment, residential status, and socio-economic characteristics.

2.2. Variables and Operationalization

1. Explained variable: The explained variable is the psychological health level of urban youth. Current research often uses indicators such as stress index, social satisfaction, and emotional stability to assess the mental health status. Based on relevant research on the youth population and combined with survey data on the residential characteristics and behavioral patterns of urban youth, this article selects a multidimensional psychological assessment system for measurement. Positive psychology focuses on evaluating stress resistance, independent living efficiency, and social initiative, with specific indicators including “adaptability to job mobility”, “satisfaction with independent space management”, and “ability to maintain stable social relationships”. Negative psychology focuses on perceived economic pressure, living environment anxiety, and social avoidance tendencies, involving dimensions such as “housing/consumer loan stress tolerance”, “privacy anxiety in shared living environments”, and “online social dependence”. The evaluation is quantified using a percentage system, with the medium risk interval defined as 60–75 points. The average score of the research sample is 65 points, among which structural factors such as residential space independence, commuting pressure index, and debt type have a significant impact on the distribution of scores. The scoring system integrates environmental intervention variables such as community environmental quality (mean score of 3.5/5) and psychological counseling pathway selection (50.5% dependent on self-regulation). The higher the final score, the better the level of mental health.

2. Explanatory variables: The core explanatory variable variables of this article are residential style and living environment. Based on the survey results for the question “who are you currently living with?” in the questionnaire, the living arrangements are divided into four types: living alone, sharing (with nonfamily), living with parents, and other types. The residential environment system covers three dimensions: social network support, housing environment, and community environment. The evaluation of social network support adopts an improved scale, which redivides the sources of support into four categories by detecting the social relationships that respondents can obtain substantial help from (such as discussing personal matters, solving difficulties, etc.): only relatives, only friends, coexistence of relatives and friends, and no support. The housing environment includes dual indicators of residential type and living space: residential types are mainly shared apartments, commercial housing, and long-term rental apartments; residential space is measured by whether there is a separate bedroom, with 74% of respondents having independent space. The community environment adopts a service facility quantity evaluation system, with a quantitative score of 3.5 points (on a scale of 1–5), among which green spaces/parks and sports venues have a significant effect on improving mental health.

3. Controlled variables: The controlled variables cover eight dimensions: gender, age, marital status, years of education, physical health status, income, transportation mode, and psychological counseling solutions. The research focuses on urban youth aged 25–30, distinguishing between local and foreign categories. The specific measurement methods for each variable are as follows: gender is classified using binary classification (male/female), and marital status is divided into two categories: married and unmarried. A dual measurement system is implemented for educational background: the continuous variable of years of education (mean 14.1 years, corresponding to undergraduate education level) and the categorical variable of educational level. The data show that the high education group has a single living rate of 24.3%, significantly higher than the 16.8% of undergraduate and below groups. The income situation is characterized by the debt structure and subdivided into three categories: consumer loans, housing loans, and education loans. The self-rated health scale is used to classify physical health into three levels: good, average, and poor. Behavioral types can be classified into three categories based on the degree of addiction: dual addiction (simultaneous smoking and drinking), single smoking, and excessive drinking. Transportation modes are associated with psychological stress assessment and are divided into four types of travel groups: walking/cycling, subway/bus, private car/ride hailing, and shared bicycles/electric bicycles. The psychological counseling solution adopts multidimensional classification, including self-regulation, confiding with friends and family, online anonymous social networking, professional psychological counseling, digital mental health services, and six coping modes of ignoring/avoiding problems. This system effectively covers the heterogeneous characteristics of urban youth groups.

2.3. Model Configuration

The model used in this article is a multiple linear regression model with automated machine learning, where the multiple linear regression model uses Stata17.0 operation, and the automated machine learning method is implemented through Matlab R2023a.

1. Multiple linear regression model: This article first establishes a multiple linear regression model and uses the OLS to analyze the impact of living habits and living environment on the mental health of urban youth. This model can provide convenience for the subsequent analysis of BP neural networks by reducing the analysis dimensions (i.e., removing insignificant factors from the control variables) to improve the computational speed of BP neural networks. The model is described as follows:

$$\begin{array}{ccc}\hfill J_{mh}^i& =& a_{12}+a_{13}{X}_{1i}+a_{14}{X}_{2i}+\cdots \hfill \\ & +& a_{19}{X}_{7i}+\sum _{j=0}^6{b}_{ij}{Z}_{ij}+{\epsilon }_i\hfill \end{array}$$

(1)

where $J_{mh}^i$ is the psychological health level of the ith urban youth; $X_{si},s=1,2,\cdots ,7$ are the explanatory variables such as residential style, living environment, social network support, etc. $Z_{ij},j=1,2,\cdots ,6$ are the controlled variables; $a_{12}$ is the intercept term; $a_{13},a_{14},\cdots ,a_{19}$ and $b_{ij},j=1,2,\cdots ,6$ are the regression coefficients of each variable; and ${\epsilon }_i$ is the random error term.

2. BP neural network model: On the basis of multiple linear regression, this article adopts the most classic BPNN (back propagation neural network) algorithm (shown in Figure 2) to conduct further research on the model. The BPNN enhances the prediction ability of the network by introducing hidden layers, and continuously adjusts the weights and biases of the network through back propagation to minimize the sum of squared errors of the network. The learning mechanism of the BPNN enables it to adapt to more complex nonlinear problems.

In the BPNN, $X=\{X_{1i},X_{2i},\cdots ,Z_{ij}\}$ represents the input vector; $v_1,v_2,\cdots ,v_m$ are the weights for neurons; Y is the output variable, denoting the psychological health level.To solve the nonlinear problems, the activation function uses the Sigmoid function. Note that the BPNN uses the mean square error as the loss function during training and converges using gradient descent. Therefore, the BPNN is highly sensitive to initialization weights and biases, making them prone to becoming stuck in local optimal.

In order to improve the performance of the BPNN, the meta-heuristic algorithm is used to optimize the initialization weights and biases of the BPNN in this paper. On the other hand, chaos particle swarm optimization (CPSO) algorithm, as a representative of meta-heuristic algorithms, is easy to implement and encode. The advantages of having fewer control parameters and being able to flexibly mix with other optimization algorithms have meant that it has always been regarded by researchers as an efficient, accurate, and simple algorithm, which has been applied to parameter optimization and solving numerical problems. This article also uses the CPSO algorithm to optimize the initialization weights and biases of the BPNN.

3. The CPSO algorithm and mean impact value algorithm: The CPSO algorithm uses N particles to form a particle swarm and iteratively finds the optimal solution in D-dimensional space. Each particle i in the t-th generation has a velocity $V_i^t=[v_{i1}^t,v_{i2}^t,\cdots ,v_{iD}^t]$ and a position $X_i^t=[x_{i1}^t,x_{i2}^t,\cdots ,x_{iD}^t]$, where t represents the current number of iterations, D is the dimension of the problem function, and i is a positive integer less than or equal to N. In addition, each particle i will save its historical best position $P_i=[p_{i1},p_{i2},\cdots ,p_{iD}]$. The historical optimal position in the population is represented as $B=[b_1,b_2,\cdots ,b_D]$. The velocity of the i-th particle in generation $t+1$ is updated using the following formula:

$$v_{id}^{t+1}=w{v}_{id}^t+c_1{r}_{1d}^t(P_{id}-x_{id}^t)+c_2{r}_{2d}^t(b_d-x_{id}^t)$$

(2)

where w is an inertia weight, and the coefficients $c_1$ and $c_2$ are constants usually set to 2 or adaptively changed according to the evolutionary state. $r_{1d}^t$ and $r_{2d}^t$ are two randomly generated values within the range of [0, 1]. According to the velocity of the particle, the position of the i-th particle in the $t+1$-st generation is updated using the following formula:

$$x_{id}^{t+1}=x_{id}^t+v_{id}^{t+1}+\delta C_i$$

(3)

where $\delta$ is the disturbance intensity and $C_i$ is the disturbance vector generated for chaotic sequences.

The mean impact value (MIV) algorithm is a method used to evaluate the importance of independent variables in neural networks and is considered one of the best algorithms for assessing the correlation between input and output variables. Sorting variables based on the absolute value of MIV can determine the degree of influence of input variables on network output variables. The sign of the MIV value represents its relative direction, and the relative importance of influence is represented by the absolute value of the MIV.

4. The automated machine learning method and CPSO-BPNN: The CPSO-BPNN is used to analyze the factors influencing the mental health of urban youth. A flowchart of this method is shown in Figure 3.

We first traverse the first and second layers of the BPNN and select the optimal BPNN architecture based on the prediction accuracy returned from the test set. Secondly, we construct a BPNN population based on the optimal network architecture and optimize the initialization parameters of the BPNN using the PSO algorithm. Finally, we obtain the optimal neural network architecture that includes the best initialization weights and biases, and use the MIV algorithm to determine the degree of influence of the independent variables.

3. Analysis and Discussion

3.1. Descriptive Statistical Analysis

The descriptive statistical results of the observed samples are shown in

Table 1. The descriptive statistical results.

Variable	Mean Value/ Proportion (%)	Variable	Mean Value/ Proportion (%)
Mental Health Level (Average)	66	Controlled Variables
Explanatory Variables		Average Age	27.5
Residential Style		Sex
Living Alone	21.14	Male	47.46
Sharing	44.68	Female	52.54
Living With Parents	32.82	Marital Status
Other	1.36	No Spouse	73.68
Residential Type		With Spouse	26.32
Shared House	47.28	Years Of Education (Average)	14.1
Self-Owned House	23.65	Graduate Degree Or Above	64.28
Long-Term Rental Apartment	17.60	Bachelor Degree Or Below	35.72
Affordable House	6.15	Liabilities Type
Unit Dormitory	5.32	Consumer Loans	51.28
Residential Space		House Loans	33.24
Independent Space	74.32	Education Loans	11.74
Non-Independent Space	25.68	Other	3.74
Community Environment (Average)	3.50	Health Behavior
Social Lifestyle		Smoking	20.36
Online	71.88	Drinking Alcohol	26.83
Offline	28.12	Both	50.19
Psychological Counseling		None	2.62
Self-Regulation	43.83	Self-Rated Health
Confiding	18.12	Good	52.83
Professional Counseling	10.55	General	34.67
Ignore Or Avoid	27.50	Poor	12.50

. The mean value of the mental health of urban youth is 65 points. This implies that, if the mental health level is on a percentage scale, and if passing is 60 points, which is in the medium-risk range, this indicates that the overall mental health status of the sample is still acceptable but there is some pressure.

In terms of the explanatory variables, shared living (with nonfamily members) has the highest proportion among the living arrangements, followed by living with parents and living alone. More than 60% of people living alone pursue independent space as the main reason for this. Among the residential types representing the living environment, shared housing (private house) is the main type, followed by self-owned commercial housing and long-term rental apartments, with a relatively low proportion of affordable housing and unit dormitories. In terms of the residential space, over 70% of the sample have independent spaces, but 26% still lack privacy protection. The average score for the community environment is 3.5 points (on a 1–5 point scale), indicating that the living environment is at a moderate level, and there is still room for improvement in public facilities and services.

The social mode shows significant online characteristics, with 70% relying on online social interaction and only 30% relying on offline interaction. The main approach to psychological counseling is self-regulation, followed by confiding with family and friends, and professional psychological counseling, indicating that social support networks still play a certain role in psychological adjustment. It is worth noting that 27.50% of the samples chose to ignore or avoid psychological issues, and we need to be alert to the long-term health and psychological risks that may arise from this case.

The demographic characteristics show that the sample has a balanced gender ratio, with an average age of 27.5 years, and the proportion of unmarried individuals is as high as 70%. The average length of education is 14.1 years (about up to a university degree level), with a significantly higher proportion of people with higher education (graduate degree or above) living alone compared to those with bachelor’s degree or below. The main types of liabilities are consumer loans (including credit cards), followed by housing loans and education loans.

In terms of health behavior, the proportion of individuals with depression/anxiety caused by smoking (daily) and excessive alcohol consumption is close to half, and the psychological problems of the dual behavior (coexistence of tobacco and alcohol) group are the most prominent, accounting for more than half of the sample population. More than half of the self-rated health respondents believe that their condition is good, but 12.50% rate it poorly. Among other influencing factors, there is a significant difference in the psychological stress index of transportation modes: subways/buses have the highest stress, while walking/cycling has the lowest. Public facilities have a significant effect on improving mental health, with green spaces/parks and gyms having the most significant impact. The population classification shows that local youth with registered residence registration and migrant workers account for the majority, followed by college graduates and youth with flexible employment.

3.2. The Impact of Residential Style and Living Environment

Firstly, a multiple linear regression analysis is conducted to remove insignificant control variables and reduce the analysis dimensions of the neural network. Then, the neural network is used to analyze all of the samples. On the basis of comparing the accuracy of two models to demonstrate the accuracy of the neural networks, a heterogeneity analysis is conducted.

1. Analysis results of CPSO-BPNN: The training parameters of the BPNN and hyperparameters of the PSO algorithm are set as follows: the number of hidden layers is 2, and the number of hidden layer nodes is 15. The maximum amount of training time $M=100$, and learning rate $L_r=0.10$. The minimum error of the training objective $e={10}^{-5},$ momentum factor $c=0.01.$ The minimum performance gradient $g={10}^{-6}.$ The population size of CPSO is $N=40$, the maximum number of iteration is $T=200$, and the inertial weight $w=0.90.$ The disturbance intensity $\delta =0.30$ and the disturbance vectors $C_i=0.25$. The analysis results of the CPSO-BPNN are shown in

Table 2. The CPSO-BPNN-based analysis results.

Variable	Coefficient	Variable	Coefficient
Explanatory Variables		Controlled Variables
Residential Style		Age	−0.015
Living Alone	0.195	Sex (Reference Group: Male)
Sharing	0.412	Female	0.329
Living With Parents	0.036	Marital Status (Reference Group: No Spouse)
Other	0.140	With Spouse	0.528
Residential Type		Category (Reference Group: Local)
Shared House	0.438	Foreign	0.194
Self-Owned House	1.135	Education Level	0.165
Long-Term Rental Apartment	0.362	Self-Rated Health (Reference Group: Good)
Affordable House	0.478	General	−2.207
Unit Dormitory	1.021	Poor	−3.854
Residential Space		Loans (Reference Group: Other)
Independent Space	1.320	House Loans	−3.84
Non-Independent Space	−0.875	Education Loans	−2.56
Community Environment	1.530	Smoking Frequency (Reference Group: None)
Social Lifestyle		Always Smoking	−3.07
Online	0.884	Sometimes Smoking	0.383
Offline	0.712	Drinking Alcohol Frequency (Reference Group: None)
Psychological Counseling		Always Drinking Alcohol	−0.473
Self-Regulation	0.233	Sometimes Drinking Alcohol	−0.296
Confiding	−0.136	Income Status (Reference Group: Enough)
Professional Counseling	0.355	Not Enough	−1.32
Ignore Or Avoid	−0.135

.

4. Conclusions and Recommendations

4.1. Conclusions

Through a CPSO-BPNN analysis, different coefficients reflect the importance of different influencing factors on the mental health of urban youth. The specific conclusions are as follows.

1. Among the factors related to residence, young people who choose to share apartments, rent long-term apartments, or own commercial housing and have access to independent space and a high-quality community environment have relatively higher levels of mental health. The proportion of “Sharing” is 44.68, while the coefficient is 0.412. So, the shared rental model can effectively alleviate loneliness through high-frequency social interactions and cost of living sharing mechanisms, and the possibility of daily communication can partially offset the risk of social isolation. However, we need to be vigilant about the negative effects that may arise from blurred privacy boundaries and social pressure. From the data from the 2023 report on the health status of urban youth in China and the tracking survey data, we can see that although living alone can meet the demand for independent space, the combined effect of social isolation and economic pressure significantly exacerbates the risk of depression, especially for those who passively choose to live alone (such as those experiencing financial difficulties). Some young people living alone tend to adopt negative coping strategies due to a lack of social support networks. It is worth noting that high self-efficacy youth can build an alternative social support system that combines freedom and belonging through online community participation and offline interest activities.

Shared housing (with nonfamily members) can alleviate loneliness through frequent social interactions and sharing the cost of living, and the daily communication that occurs can partially offset the risk of social isolation. However, the blurring of privacy boundaries and social pressure may weaken the positive effects. Although living alone satisfies the need for independent space, the combination of social isolation and economic pressure significantly exacerbates the risk of depression, especially for passive choice groups such as economically disadvantaged individuals. Some young people living alone adopt negative coping strategies due to a lack of support networks. It is worth noting that high self-efficacy youth can construct an alternative social model of ”freedom and belonging coexisting” through online communities and offline activities.

Self-owned commercial housing enhances mental health through both economic stability and community belonging, but its high housing price pressure may in turn trigger long-term debt anxiety. The flexibility of long-term rental apartments meets the needs of occupational mobility, and shared spaces promote weak social relationships. However, through the questionnaire investigation, it was found that problems such as poor sound insulation and lack of privacy have led to some residents exhibiting depressive tendencies, highlighting the paradox of the “resource deprivation cycle”. The priority ranking of influencing factors, from high to low, is as follows: privacy issues in long-term rental apartments (negative), negative coping with living alone (negative), debt anxiety in self-owned commercial housing (negative), and social pressure in shared rental housing (negative). Youth groups are often influenced by factors such as occupational mobility, economic pressure, and intergenerational relationships, and their housing choices often face a trade-off between “social needs and privacy protection” and “economic costs and sense of belonging”. Previous studies have shown that flexibility and accessibility to community resources are crucial for mental health, but excessive reliance on a single living mode (such as living alone or renting apartments for a long time) can lead to psychological risks.

2. The impact of other housing-related factors on the mental health of local youth and migrant workers in cities shows a high degree of consistency, except for the mode of residence and type of housing.

There are significant group differences in the impact of the living environment on the mental health of urban youth, but local youth and migrant workers show high consistency in non-residential factors. At the level of living patterns, local youth mainly live alone and share apartments with non-relatives: those who actively live alone pursue spatial autonomy or adapt to occupational mobility, but high rental pressure and social isolation have prompted some groups to adopt a “problem neglect” strategy. It is worth pointing out that although shared living alleviates loneliness through cost sharing and weak social relationships, the blurred physical boundaries lead to a tendency towards depression. On the other hand, among the migrant worker population, the three-generation cohabitation model significantly reduces the risk of depression by 32%. The intergenerational economic mutual assistance and emotional connection effectively alleviate mobility anxiety, but this model only accounts for 7% among local youth, who are more inclined towards the core family structure of their partners.

The group differentiation characteristics of residential types are also significant. Local youth rely on the stability of property rights in commercial housing and the community resource network to build a sense of belonging, while long-term rental apartments adapt to occupational mobility needs through shared spaces. The migrant worker population tends to gather in affordable rental housing (with poor sound insulation and a per capita area of less than 5 square meters, leading to a cycle of resource deprivation) or unit dormitories (where crowded spaces and poor hygiene conditions directly trigger depression), both of which constitute systemic risk sources.

The breaking mechanism of social support networks further strengthens group differences. Local youth rely on a strong offline network to establish a stress buffer zone, and 63% of people living alone regularly use community counseling stations to alleviate their emotions. The migrant worker group relies on online anonymous social networking to achieve shallow emotional connections, and their depression factor score is 1.4 times higher than that of the local group. Additionally, passive solitary individuals are more likely to fall into negative coping strategies due to the support network vacuum. This rupture reflects the structural imbalance of the urban public service system’s response to the needs of different groups.

3. The housing-related factors that contribute to gender differences are living habits and sources of social network support.

There are significant gender differences in the impact of living environment on young people, mainly reflected in two dimensions: living situation and social support network.In terms of living patterns, urban female youth are more suitable for equal sharing of shared or self-owned housing, while men are more able to gain development momentum in independent living (living alone or self-owned housing) and core family models (living with spouses and children). This difference is further confirmed in the social support system: men’s offline social interactions exhibit high-frequency but low-quality characteristics, such as the use of alcohol as a link in long-term rental apartments, which can temporarily relieve stress but is accompanied by health hazards such as alcoholism. Women tend to rely more on online platforms for social interaction, but excessive virtualization can lead to a lack of emotional support, especially in shared living environments, where physical boundaries are blurred and privacy anxiety may be exacerbated.

We note that the gender sensitivity analysis shows structural differences in the priority ranking of residential elements. The impact of the lack of independent space on women’s mental health is significantly higher than that of men, while the traditional large family model has a higher negative impact on men’s mental health and shows a significant negative correlation. This difference reflects a deep-seated contradiction in the social structure—women face systemic difficulties in housing rights, relationships, and emotional support pathways, while men need to focus on avoiding the spatial constraints and social externalization risks of traditional families (such as substance abuse in offline social interactions).

4.2. Recommendations

1. Optimize the design of residential policies. It is recommended to promote a diversified residential community model by configuring modular shared spaces (such as public living rooms and shared office areas) in long-term rental apartments and affordable housing, and promoting social interaction among young people through a “private semi-open public” layered space design. The government can incorporate the allocation of shared spaces into housing construction standards and link community activity with rental subsidies to stimulate participation. At the same time, a “privacy-first” co-rental housing certification system should be established, clarifying basic standards such as independent bathrooms and sound insulation facilities. The housing management department can use an intelligent evaluation system to classify and manage the quality of housing resources, and dynamically optimize the reward and punishment mechanism based on complaint data, to ensure the living experience and mental health of young people living together from an institutional perspective.

2. Hierarchical matching of residential resources. In response to the differentiated needs of human capital and occupational characteristics among the youth population, it is recommended to establish a hierarchical allocation system for residential resources.

For highly educated groups, we should promote the transformation of knowledge capital into residential models. Academic exchange spaces and skill sharing facilities should be set up in young professionals’ apartment complexes, and resource utilization efficiency should be optimized through intelligent management.Local governments can link cultural consumption subsidies with housing policies, forming a linkage mechanism between academic achievements and housing cost reductions. For low-education and flexible employment groups, we should strengthen basic living service guarantees. The layout of 24 h convenience stores and affordable fitness facilities in residential areas should be prioritized, and a community service point redemption system should be established. Participation in public services (such as garbage classification supervision and shared facility maintenance) should be quantified as rental deductions, and digital technology should be used to achieve the transparent management of service records.

This system adopts a dual track path of knowledge empowerment and basic service supplementation to enhance the accuracy and fairness of residential resource allocation, and promote the effective satisfaction of the housing needs of different groups.

3. Build a gender-friendly support system. Based on the gender differences among young people, it is recommended to establish gender-specific intervention mechanisms: targeting male youth, focus on promoting healthy social replacement programs. Set up mental health service stations around industrial parks, equipped with group sports facilities and anonymous communication services; incorporate regular sports activities into the corporate social responsibility evaluation system; and provide policy incentives to communities that have achieved significant results in improving tobacco and alcohol dependence. We should strengthen safety and psychological support for young women by promoting a certification system for safe apartments, requiring housing units to be equipped with intelligent security systems, and setting up nighttime psychological counseling stations in areas where single women are concentrated. Social activity trajectories should be analyzed through intelligent technology and safe travel routes should be planned for high-frequency participants.

This system can effectively reduce the rate of unhealthy habits in males and the safety anxiety index in females, achieving an appropriate response to gender-differentiated needs.

4. Improve the community service ecosystem. Building a collaborative service system of “pressure-buffering capital optimization” is suggested. (a) The community psychological service network should be optimized: A 15 min accessible psychological stress relief space could be set up (including green spaces, book bars, fitness cabins, etc.), wherein an intelligent reservation management system is implemented, and personalized services based on usage frequency can be utilized (such as nighttime psychological counseling). (b) Innovative debt psychological linkage mechanism: A linkage evaluation system can be established between debt pressure and psychological state, allowing young people to select low-interest housing loans, providing financial management and psychological adjustment training, and dynamically adjusting repayment plans through physiological data monitoring. (c) Building a service value transformation platform: blockchain technology can be used to establish a community service point system, transforming behaviors such as psychological counseling and skill sharing into credit assets that can be exchanged for residential benefits, and supporting rent or loan interest deductions.

Practice has shown that this system, through three-dimensional interventions of spatial empowerment, debt restructuring, and service incentives, can effectively enhance the psychological resilience of young people and reduce systemic psychological risks at the community level.