Optimization of Urban Renewal Planning Schemes and Community Vitality Enhancement Strategies Based on Deep Learning Algorithms and Smart City Evaluation

In today’s society, the development of cities is changing rapidly, and the importance of urban planning is becoming more and more prominent. Among them, urban renewal and community vitality enhancement have become key issues, which are of great significance in improving the quality of life of residents and promoting sustainable urban development [1-4]. Urban renewal is not a simple demolition of the old and construction of the new, but a comprehensive and systematic process. It involves the improvement and optimization of the physical space, socio-economic and human environment of the old city [5-6]. In the past, in the process of urban renewal, some cities often focus only on the transformation of physical space, large-scale demolition of old buildings, the construction of high-rise buildings, but neglected the protection of the original social network and cultural characteristics of the community, resulting in a sense of identity and sense of belonging to the community is reduced. Nowadays, we need to pay more attention to the concept of people-centeredness, give full consideration to the needs and interests of residents, and realize urban renewal while improving community vitality [7-10].

The improvement of community vitality cannot be separated from the improvement of community functions. A vibrant community should have diversified functions to meet the needs of residents in life, work, study, entertainment and other aspects [11-13]. Community governance is also the key to enhancing community vitality. It is necessary to establish a sound community governance mechanism, encourage residents to participate in the management of community affairs and decision-making, and give full play to the main role of residents. At the same time, the construction of community service teams should be strengthened to improve the quality and level of community services [14-17].

Literature [18] mentions the development of urban renewal and its characteristics in China, distinguishes between wholesale redevelopment and micro-renewal, and analyzes the impacts of different urban renewal modes on NA, and finds that micro-renewal is hardly able to destroy the NA of the residents, and that NA is mainly affected by spatial reconfiguration. Literature [19] describes a framework for assessing community sustainability to support urban regeneration decisions in high-density cities. It also describes the application of this framework in the decision-making process of urban renewal projects in a case study, which is expected to support urban renewal decision-making. Literature [20] examines the transformative potential of urban renewal, particularly community revitalization. Taking a community as an example, and based on the dimensions of public space and ecological environment, the renewal process of the community was examined, emphasizing the importance of community ownership and participation. The results of the study are of relevance to urban planning and other stakeholders. Literature [21] examined the lived experience of urban renewal and place attachment in Zhuanghe, and adopted a case study approach to address the corresponding key research questions. The findings pointed out that the residents showed ambivalence in the face of urban renewal, and proposed the method of in-situ relocation in order to improve the living conditions of the residents. Literature [22] highlights the importance of focusing on slum solutions, proposes solutions for urban renewal programs and revitalization models, and reveals the many challenges of this approach, which requires the participation and efforts of all parties, including citizens of the community, landlords, developers, and the government, and emphasizes the important role of better social well-being and social justice in this process. Literature [23] provides a detailed characterization of the urban development pattern based on Harbin as the subject of the study, and based on depicting the local changes in the influencing factors, it lays the foundation and objective references for urban planning and provides insights and frameworks for the analysis of regeneration in other cities. Literature [24] discusses the influencing factors, methods, advantages and limitations of public space vitality in the context of urban regeneration based on urban renewal, public space and vitality, and proposes corresponding strategies. Literature [25] takes a residential neighborhood as a research object and integrates a variety of datasets such as pol. A comprehensive living environment quality evaluation index system was established by analyzing the concept of urban renewal. The results verify the effectiveness of the living environment quality evaluation method based on multi-source data, which has important research and practical value.

The article firstly proposes an evaluation system of urban renewal planning scheme covering three dimensions of society, economy and environment by combining ERG theory, and analyzes in detail the significance of evaluation indexes of three dimensions of society, economy and environment. Subsequently, a radial basis neural network model (AHP-RBF neural network model) based on hierarchical analysis is proposed. After constructing the smart city evaluation index system, the neural network structure is determined and trained, and the network performance is evaluated using the root mean square error and other indicators until the network performance reaches the preset standard. Then we take 2 cities, Yumen and Otago, as examples to validate the method, conduct applied research on the model, and excavate the existing problems. Finally, strategies for enhancing community vitality are proposed at three levels: physical space, social space, and cultural space.

2

Smart city evaluation method based on deep learning algorithm

2.1

Evaluation system of urban renewal planning program based on ERG theory

The evaluation system for urban renewal planning programs based on ERG theory is shown in Figure 1. This evaluation system covers three dimensions: social, economic and environmental, and each dimension contains specific evaluation indicators and evaluation methods to ensure that urban renewal projects promote social well-being and environmental sustainability while transforming physical space [26-27].

2.1.1

Evaluation of social dimensions

The degree of public participation is a core indicator for assessing whether urban renewal projects can truly reflect the wishes and needs of community residents. In order to improve the scientific nature of project evaluation, the authorities concerned can collect residents’ opinions by means of questionnaires, record and analyze the collection of questionnaires, and conduct qualitative and quantitative analyses of public needs.

Social justice evaluation focuses on assessing whether urban renewal projects can meet the needs of various social groups equally, especially the needs of some marginalized and disadvantaged groups. By conducting social impact assessment and equity analysis, the departments concerned can identify and assess the possible social impacts of a particular urban renewal project, and ensure that different groups are served equitably in terms of their basic needs of life, such as housing, education and medical care. Relevant departments should collect relevant data, such as housing affordability, rationality of education resource allocation and accessibility of medical facilities, and combine them with interviews and community research to gain an in-depth understanding of the actual demand for and satisfaction with these public services by different groups.

The evaluation of cultural preservation and promotion focuses on the impact of urban renewal projects on local cultural heritage and support for cultural diversity. The evaluation includes monitoring and analyzing the conservation status of cultural landmarks, and assessing whether urban renewal projects cause damage to cultural heritage or contribute to the preservation and transmission of cultural heritage. At the same time, the authorities concerned are also required to assess the extent to which urban renewal projects support regional cultural activities such as festivals, art exhibitions, and cultural performances.

2.1.2

Evaluation of the economic dimension

In evaluating economic growth, the relevant departments should focus on the direct contribution of urban renewal projects to local economic activities, such as the creation of employment opportunities and the improvement of residents’ income levels. In this process, relevant departments can use economic impact analysis tools, such as input-output modeling, to predict the long-term impact of urban renewal projects on the local economy after implementation. Input-output modeling can reveal the circulation and multiplier effects of construction expenditures on the local economy, and thus predict the huge potential for job creation and income increase. For example, the relevant authorities can carefully count and analyze the employment opportunities created directly or indirectly during the construction and operation phases of an urban renewal project, as well as the positive impact of these employment opportunities on the total income level of the local community.

In order to accurately measure the return on investment of an urban renewal project, it is important to use a variety of evaluation indicators. These indicators should cover not only the direct return on investment of the project, but also the net contribution of the project to the commercial vitality of the area, in order to promote the sustainability and diversity of the economy.

In assessing business vitality, the authorities should pay attention to the diversity and activity of business activities in the area where the urban renewal project is implemented, and assess the role of urban renewal project implementation in enhancing the economic vitality and attractiveness of the area. By monitoring the commercial leasing rate and the rate of new business start-ups, as well as conducting consumer satisfaction surveys, the authorities can comprehensively assess the degree of improvement in the business environment.

2.1.3

Evaluation of environmental dimensions

In the context of urban renewal projects, an environmental dimension assessment can clarify the far-reaching impact of urban renewal projects on the ecology of the city and the quality of life of its residents. By comprehensively evaluating the improvement of environmental quality, the integration of green infrastructure and the effective utilization of resources, the relevant authorities can ensure that urban renewal projects can effectively promote environmentally sustainable development while enhancing the vitality of the regional economy.

Evaluation of environmental quality improvement is an important indicator for assessing the effectiveness of urban renewal projects. Through close monitoring of key environmental indicators such as air quality index, water quality reports and noise levels, the relevant authorities will be able to have a comprehensive understanding of the impact of project implementation on the surrounding environment.

In assessing the integration of green infrastructure, the authorities should focus on the extent to which green infrastructure has been incorporated into urban renewal projects, such as green space coverage, biodiversity index and stormwater management efficiency. An increase in the area of green space not only improves the microclimate of the city and enhances its aesthetics, but also helps to improve the quality of life and mental health of the residents.

In assessing the efficiency of resource utilization, the authorities should pay attention to the performance of urban renewal projects in terms of energy and resource utilization. Through in-depth analysis of energy consumption data and resource utilization efficiency, the authorities can objectively evaluate the actual effectiveness of urban renewal projects in promoting environmental sustainability. Overall, by analyzing the performance of urban renewal projects in the three dimensions of society, economy and environment, the relevant departments can scientifically evaluate the feasibility and reasonableness of urban renewal projects, and clarify the role of urban renewal projects in maintaining social equity, promoting urban economic growth and protecting the ecological environment.

2.2

AHP-RBF neural network model construction

AHP derives its indicator weights through the scoring of judges or experts, which leads to a strong artificial subjectivity of the derived weights, and too many indicators will seriously affect the accuracy of the comparative judgment matrix, so that it can not pass the consistency test. The RBF neural network method has a better self-learning, self-adaptive ability, but in the network training process, it is only through the learning of the training samples to establish the network model, and it can not make the human intervention, the calculation process is not controllable, so that the results obtained lack a certain degree of subjectivity. Based on the advantages and disadvantages of the two evaluation methods, this paper adopts AHP and RBF neural network to jointly construct the evaluation model, the key of which lies in the fact that AHP allows the RBF neural network to generate new training sample data [28-29]. As the evaluation results of AHP are rated by experts, it makes the RBF neural network also have the experience and knowledge of experts. When evaluating the study area, the RBF neural network is able to simulate the expert’s thinking based on the input samples, so that the computer combines with the human brain to make judgments, which improves the accuracy of the evaluation. The computational schematic diagram of the AHP-RBF neural network model is shown in Fig. 2, which shows that the indicator system is established firstly. Then the AHP is used to determine the weight of each indicator and rank their importance, and then the RBF neural network tool is used to establish the AHP-RBF neural network evaluation model.

2.3

Smart city evaluation model based on AHP-RBF

2.3.1

Hierarchical Analysis

Hierarchical analysis is a decision analysis method for solving multi-level and multi-criteria problems, and its specific process can be described as follows: 1)

Establish a hierarchical model.

2)

Construct the comparative judgment matrix B, and compare the scoring according to the SAATY 1~9 scale method to get the value of matrix B, as shown in equation (1): (1) $B = {\begin{array}{l} b_{i j} & i \neq j \\ 1 & i = j \end{array}$

3)

Calculate the eigenvector ξ and the maximum eigenvalue λ_max of the judgment matrix B, and then normalize the eigenvectors to obtain the ranking weights of each evaluation index in the same level with respect to the importance of an evaluation index in the previous level.

4)

Consistency test. If the relative consistency ratio coefficient CR of judgment matrix B is less than 0.1, then it passes the consistency test, and CR is as small as possible. On the contrary, it does not pass the consistency test, then return to re-select the two-two comparison, the construction of qualified judgment matrix B, CR calculated as formula (2): (2) ${\begin{array}{l} C R = C I / R I \\ C I = (λ_{\max} - n) / (n - 1) \end{array}$

5)

Calculate the weight of each indicator W_i based on BW = λ_maxW.

2.3.2

RBF Neural Network Models

1)

Artificial Neural Network Model

The most basic unit of artificial neural network is neuron. Neurons are mainly based on the simulation of the human brain and work on the principle of human brain thinking. Neurons are the basis for information processing and transmission in the human brain, and they mainly utilize axons, synapses and free cell bodies as the main media to form the neuron system. Free cell bodies play a key role in the operation of the whole system, while axons are the most important pathway for information transmission. Information can be transmitted from the beginning of the axon to the end, which then releases the dendrites of the neuron and then transmits and receives information.

Artificial neurons are the simplest information processing unit, and the artificial neural network model is shown in Figure 3.

Artificial neurons are mainly composed of inputs, processing functions and outputs, where the inputs can be more than one such that the artificial neuron exists: (3) $X = \sum_{j = 1}^{n} w_{i j} x_{j} - θ_{i}$ (4) $y_{i} = f (x)$

x_j denotes the information transmitted from the other neuron to the present neuron: θ denotes the threshold value. w_ij denotes the magnitude of the connection strength between two different neurons, this parameter is mainly expressed using the weights. f(x) denotes the activation function in the expression of this neuron. The activation function mainly expresses how the input in the neuron is processed to get the output. Since the information between neurons has a nonlinear relationship, the activation function also has a nonlinear characteristic. In order to perform the calculation, the effect of x₀ on the neuron’s information transfer can be ignored, so that the neuron expression is: (5) $X = \sum_{j = 1}^{n} w_{i j} x_{j}$

2)

RBF neural network model

RBF neural network is also known as radial basis neural network, the network is mainly composed of neurons in the structure, and these neurons can realize the local adjustment function between them. In the whole RBF neural network is equivalent to a forward neural network and a single layer of hidden layer neural network, and the hidden layer neural network in the output can provide radial basis function for the whole RBF neural network. The input to each radial basis function for the RBF neural network is the distance between the input vector and the center position. The RBF neural network model is shown in Figure 4.

The first layer in the RBF neural network is the input layer of the network model, which is mainly for the input of data information, and in practical applications there are several feature vectors that need to be entered in the input layer. The implicit layer of the network model is where the activation function is calculated using a radial basis function. The third layer is the output layer of the network model, in which the main thing is to output the result. In RBF neural network, the data between the input layer and the hidden layer has nonlinear characteristics, and the data from the hidden layer to the output layer has linear characteristics. This also means that the weight value of the indicator between the input layer and the hidden layer is 1 constant value and the weight value between the hidden layer to the output layer is dynamically changing.

Suppose that there are n_i inputs to the RBF neural network, c radial basis functions, and n₀ network outputs. Thus, in the RBF neural network input vector there is x ∈ X ∈ R^n_i, while the prototype of the output vector is v_j ∈ R^n_i. Suppose that in the RBF neural network there are v_j ∈ R^n_i, 1 ≤ j ≤ c, and the output for the radial basis function is: (6) $h_{j} = h_{j} (‖ x - v_{j} ‖), 1 \leq j \leq c$

In the above formulation h_j(·) is a basis function. ║·║ is an input space, which is characterized by normality, and in the general condition the space is represented using the Euclidean paradigm. Since the function expression can be decomposed, the radial basis function is often applied in RBF as a Gaussian function.

Assuming that σ is the width of the radial basis function in the RBF neural network, and the number of radial basis functions in the function is c, the result of the i th output unit in the neural network is: (7) ${y^{'}}_{i} = w_{i}^{T} \cdot h = \sum_{j = 0}^{c} w_{j i} h_{j}, 1 \leq j \leq n_{0}$

In the above formula there exists w_i = [w_i0, w_i1,…,w_ic]^T.

Since in RBF neural network every input vector exists in relation to x ∈ R^n_i, so that in this neural network then exists ${y^{'}}_{i} = W \cdot h \in R^{n_{0}}$ . Suppose $y^{'} = {[{y^{'}}_{1}, {y^{'}}_{2}, \dots, {y^{'}}_{n_{0}}]}^{T}$ , W is the weight matrix of the RBF neural network, at this time the expression of this matrix is: (8) $W = {[w_{1}, w_{2}, \dots, w_{n_{0}}]}^{T}$

At this point a complete training set mapping from vector x_k ∈ Rⁿ to another vector x_k ∈ R^n_i is formed for the RBF neural network. And there exist (x_k,y_k), 1 ≤ k ≤ M in this training set.

3)

Construction of RBF neural network model

Neural network model is a commonly used estimation method in current engineering project management, and in this thesis, RBF neural network model is chosen to carry out post-evaluation of the construction of a highway project in the research process.

Before the post-evaluation, the management problem in the construction process of the highway project should be converted into a mathematical problem, assuming that the factors affecting the effectiveness of engineering project management are m and m ≥ 3. The cost data to be obtained in the estimation process are n and n ≥ 1, and the mapping from m-dimensional space to n-dimensional space should be accomplished before the evaluation. m is the input data, n is the output data, and the Euclidean space of m is R^m in the RBF neural network model. There exists a bounded subset A, then there exists such a bounded subset mapping to the n-dimensional Euclidean space Rⁿ, which is expressed by Eq. F: A ∈ R^m → Rⁿ, through the training and learning of the network model can be obtained by the model optimization process of the approximate mapping G, and G approximates F. The output expression of the radial basis function established in this paper is: (9) $h_{j} = h_{j} (‖ x - v_{j} ‖), 1 \leq j \leq c$

In the above formula h_j(·) is the base number. ║·║ is an input space which is characterized by normality and in general condition this space is represented using Euclidean paradigm. Assuming that σ is the width of the radial basis function in the RBF neural network, and the number of radial basis functions in the function is c, then the output of the ith output unit in the RBF neural network is: (10) ${y^{'}}_{i} = w_{i}^{T} \cdot h = \sum_{j = 0}^{c} w_{j i} h_{j}, 1 \leq j \leq n_{0}$

In the above formula there exists w_i = [w_i0, w_i1,…,w_ic]^T.

Since in RBF neural network every input vector exists in relation A ∈ R^m → Rⁿ, so that in this neural network then exists ${y^{'}}_{i} = W \cdot h \in R^{n_{0}}$ . Suppose y = [y₁,y₂,…,y_n₀]^T, w is the weight matrix of the RBF neural network, at this time the expression of this matrix is: (11) $W = {[w_{1}, w_{2}, \dots, w_{n_{0}}]}^{T}$

4)

Training and testing of RBF neural network models

(1)

Selection of expansion speed of RBF neural network

The newrb function [net,tr] = newrb (P, T, goal, spread, MN, DF) in MATLAB is a commonly used function in the simulation process, and spread indicates the expansion speed of the radial basis function, so the merit of its assignment has a very large impact on the evaluation results of the RBF neural network model. At present, there is no unified data reference for spread value in the RBF model at home and abroad, and it can only be determined through multiple debugging. Spread assignment is greatly influenced by the total sample data. If the number of selected samples is large, Spread assigns too small a value, which can lead to inaccurate model evaluation results. Conversely, the accuracy of the model evaluation results is higher when the amount of data in the selected samples is smaller and the Spread takes a larger value. In the process of post-evaluation of this highway project, the error between the estimated data and the actual data in the sample database is calibrated as the main basis for determining the Spread data. The process of determining the Spread data is shown in Appendix 3, and the value of the Spread in this paper is taken to be 1 through comprehensive analysis.

(2)

Training of RBF neural network models

For the RBF neural network model, the number of hidden nodes, the basis function for the hidden layer, expansion constants, and weight correction of the output data should be solved during the application process. The specific process can be divided into two stages, one of which is to determine the hidden layer node data using the K-means clustering algorithm, which belongs to the self-learning stage of the model without any supervision. The output layer’s supervised learning stage involves using the least squares method. The number of hidden layer nodes in the RBF neural network model and the amount of weight on the output result can be easily calculated through learning.

3

Experiments and results

3.1

Calculation of weights

The judgment matrix is normalized by columns, $w_{i j} = a_{i j} / \sum_{i = 1}^{n} a_{i j}$ , and each row of the normalized matrix is summed to obtain a column vector, and the resulting column vector is normalized to obtain a weight vector. The weights of smart city construction effect evaluation indicators are shown in Table 1. From the table, it can be seen that the social dimension has the highest indicator weight ratio of 0.422 in the smart city construction effect.

Table 1.

Intelligent urban construction effect evaluation index weight

Criterion layer	Weighting	Index layer	Weighting	Composite weight
x₁	0.422	x₁₁	0.523	0.155
		x₁₂	0.266	0.085
		x₁₃	0.211	0.182
x₂	0.362	x₂₁	0.362	0.101
		x₂₂	0.352	0.205
		x₂₃	0.286	0.056
x₃	0.216	x₃₁	0.302	0.063
		x₃₂	0.281	0.122
		x₃₃	0.417	0.031

3.2

RBF Neural Network Training and Tuning

3.2.1

Learning Sample Establishment

The RANDBETWEEN function is utilized to randomly form 20 groups of data in each level range in the determined 5-level evaluation standard range, and a total of 100 groups of random data are formed, corresponding to the network expectation value of 0.5, 0.9, 1, 1.4, 0.8, corresponding to the level of smart city renewal as too low, too low, normal, too high, too high. The randomly formed data and the corresponding network expectations together form 100 groups of random samples and are divided into the training set and test set according to the ratio of 8:2, 80 groups of sample data are randomly selected from the 100 groups of data to form the training set, and the remaining 20 groups of data to form the test set, and some of the sample data are shown in Table 2. After the training and test sets are divided, the data are normalized in MATLAB.

Table 2.

Partial sample size

x₁₁	x₁₂	x₁₃	x₂₁	x₂₂	x₂₃	x₃₁	x₃₂	x₃₃	y
0.2599	0.0444	0.1111	0.3838	0.0776	0.0045	0.9291	0.1555	0.0129	0.5
0.2824	0.0621	0.1544	0.7301	0.1431	0.00809	3.0395	0.2161	0.0153	0.9
0.3127	0.0798	0.1625	0.7442	0.1749	0.01763	3.1797	0.246	0.0147	1
0.3202	0.1572	0.2127	0.7716	0.1905	0.02515	3.3968	0.2754	0.0169	1.4
0.739	0.9236	0.8279	0.9185	0.3469	0.50903	8.6752	0.6648	0.1722	0.8
0.1579	0.0435	0.1305	0.0873	0.009	0.00464	0.2858	0.078	0.0061	0.9
0.2814	0.0456	0.1457	0.7232	0.1384	0.01016	3.0297	0.2135	0.0074	1
0.2909	0.1089	0.1601	0.7557	0.1832	0.01679	3.0861	0.2554	0.0142	1.3
0.3243	0.1567	0.1968	0.7654	0.1899	0.02635	3.4137	0.2687	0.02	1.5

3.2.2

Neural network structure determination and parameter tuning

The values of the parameter SPREAD are tuned, and the neural network is trained 20 times at different expansion speeds, and the maximum and minimum values are removed and averaged to obtain the RMSE, R², and MAPE values for the training and test sets, and the model training effects at different values of SPREAD are shown in Table 3. As shown in the table, when the SPREAD value is 1, 50, 100, the neural network shows overfitting state, the training set shows good model fitting effect, but the test set shows very poor model fitting effect, at this time, the test set fitting effect with the increase of the value of constantly become better. Continue to increase the value of SPREAD, when the value of SPREAD is 800, 1300, 1500, the training set and the test set show that the model fits well. When the value of SPREAD continues to increase, the MAPE of the training set and the test set gradually increases, and the fitting effect becomes worse. Therefore, the model fits better when the value of SPREAD is around 1300, and in this paper, we take the value of SPREAD as 1500 as the better expansion speed.

Table 3.

The model training effect of different spread values

Parameter value	1	50	100	300	800	1000	1300	1500	2000
Training set RMSE	5.555	0.024	0.07	0.001	0.034	0.006	0.057	0.0707	0.0029
Training set R²	1	0.991	1.017	1.004	0.973	0.965	0.987	0.9599	0.9596
Training set MAPE	2.27%	0.16%	1.34%	2.38%	1.97%	4.56%	5.62%	6.29%	3.9%
Test set RMSE	10.301	1.878	0.62	0.139	0.053	0.116	0.083	0.058	0.1124
Test set R²	0.277	0.26	0.342	0.794	0.949	0.984	0.915	0.986	0.9181
Test set MAPE	240.63%	50.33%	18.87%	6.07%	3.12%	5.18%	4.89%	4.96%	5.38%

When SPREAD takes the value of 1500, the actual and predicted values of the RBF neural network regression training set and test set are compared with the predicted values as well as the error plots are shown in Figures 5 to 8. The goodness of fit of both the training set and the test set is above 0.9 and close to 1, the average absolute percentage error is around 6%, the root mean square error is around 0.06, and all the judging indexes are within the acceptable range, the model fitting effect of both the training set and the test set are better, and the performance of the neural network meets the requirements, and it is capable of evaluating the level of the smart city.

3.2.3

Application and analysis based on smart city evaluation models

1)

Study area

In this paper, 2 cities, Yumen and Otago, are selected as the research objects. Yumen is located in the northwestern part of Gansu Province, with a total area of 13,500 km² and a total population of 180,000 at present, and its average winter temperature can reach 6.9. The annual temperature difference can be up to about 30°C. Yumen has a high ecological and environmental pressure, suffers from sandstorms and dust storms all year round, has a small population and slow economic development, and its main pillar industries are traditional industries such as agriculture and animal husbandry. Otago is located in the southern part of New Zealand’s South Island, with an area of about 32,000 km², and a population of about 2,323,300 in June 2018, making it the country’s third-largest local region. Weather conditions within the Otago region vary considerably, with cool and wet winters in the extreme south, especially in the hills and plains south of Otago. In contrast, summers tend to be very warm and dry. Otago has a uniquely mixed economy, with vineyards and wineries booming in recent years in the central Otago region.

2)

Assessment of Urban Renewal Planning

The data used in this paper is mainly from local government reports, public data from the Bureau of Statistics, literature, and yearbooks. The input and output indicators when applying RBF neural network are the selected 9 variables respectively.

Through the pre-experiment the number of neurons in the hidden layer is set to 10, R² ≥ 0.9, the maximum number of training times is 1000, and the learning accuracy is 0.05, at which time the results are acceptable. When the number of neurons is greater than 10, the test results are poorer, although the RBF is better trained. The small size of the evaluation sample may be the main reason for this anomaly. So in this paper, the residuals remain quite low when the number of neurons is 10 and their average residuals are less than 0.05, which means that the model performs well. In order to evaluate the performance of RBFNN, the reserved partial truth values are compared with the output of the algorithm. The RBFNN output is shown in Fig. 9. (Figures a~d show the Yumen prediction results, Otago prediction results, Otago residual plots, and Yumen residual plots, respectively.) The results show that the output values are very close to the true values, which indicates that the neural network has been trained well. The final comprehensive score of each index is shown in Table 4. From the table, the comprehensive average score of smart city in Yumen City is 0.0403, and the comprehensive average score of Otago City is 0.0435, and the overall level of smart city development in Otago City is higher than that of Yumen City.

Table 4.

Comprehensive grading of each index

Serial number	Evaluation index	Jade door	Otago
1	x₁₁	0.0545	0.0258
2	x₁₂	0.0313	0.0377
3	x₁₃	0.0436	0.0543
4	x₂₁	0.0328	0.0632
5	x₂₂	0.0499	0.0798
6	x₂₃	0.0369	0.0249
7	x₃₁	0.0194	0.0123
8	x₃₂	0.0365	0.0307
9	x₃₃	0.058	0.0632

MATLAB software was used to obtain ci, and min-max was selected in the toolbox to map all the evaluation indicators of Yumen and Otago into the interval [0, 1], and ci was obtained by outputting the results from the software. The development level of Yumen and Otago can be obtained by calculating RD. The RD values of Yumen and Otago are shown in Table 5. This shows that the growth plans of the two cities have been successful to some extent, but the level of smart growth in Yumen is lower than in Otago.

Table 5.

The RD value of jade gate and Otago

City	RD
Jade gate	0.04366
Otago	0.04628

The development structure of the 2 cities is shown in Figure 10. In Yumen, the social benefit with the highest contribution rate is 62.2%, followed by economic benefit (26.31%) and ecological benefit (12.66%), which is consistent with the actual situation in Yumen. In the past development model, Yumen has been in a state of resource depletion, where natural resources such as coal, oil and natural gas have been consumed in an uncontrolled manner in exchange for short-term economic benefits and a temporary improvement in people’s quality of life. While blindly pursuing the maximization of economic benefits, Yumen’s natural environment has suffered great trauma, with environmental damage such as desertification and land salinization abounding. Therefore, the ecological efficiency of Yumen’s current smart urban planning is inevitably a shortcoming that restricts the city’s development, which is also an uncoordinated development pattern. In Otago’s urban development structure, the contributions of eco-efficiency and economic efficiency are roughly equal, which indicates the focus on eco-environmental protection in its smart growth process, and the emphasis on eco-efficiency as an important consideration in Otago’s development planning. While pursuing economic growth and improved living standards, it has continued to strengthen environmental governance and protection, cracking down on substandard chemical plants and manufacturing industries, controlling sources of environmental pollution, protecting existing wetlands and grasslands, and focusing on preserving nature’s pristine purification capacity.

4

Research on community space renewal strategies based on vitality enhancement

4.1

Environmental Vitality Enhancement Strategies

4.1.1

Increased spatial vitality of the “point” system

Dotted space usually refers to the inter-house public space of the community, which is the most common type of space in the community and has a greater impact on its vitality. Dotted public space has independence and privacy, belongs to the semi-open public space type, and the main user group is the residents around the building. The results of the study found that residents mainly use the public space between houses for social and leisure activities such as communication, shade, and resting. To enhance the vitality of point-like public space, the practicality, living and humanization of inter-house public space can be improved through three aspects, namely, creating a residency space, planting edible shared landscapes, and optimizing the spatial interface, so as to improve the spatial atmosphere.

4.1.2

Increased spatial vitality of the “linear” system

Linear space usually refers to the internal transportation system of the community, i.e., the public space of the road, which is the main medium for connecting the lives of residents, and not only assumes the function of the transportation network space, but also provides a public place for residents to communicate with each other. Neighborhood interactions and walking greetings all occur in the “linear” space. Therefore, it is important to create a good walking space by improving the accessibility, safety and interest of the linear space, and to enhance the willingness of the residents to walk, which is important for the communication between neighbors.

4.1.3

Increased spatial vitality of the “surface area” system

Planning to integrate spatial resources to create different regional functions. Planning to integrate spatial resources to create different regional functions. Add public furniture to enhance community vitality.

4.2

Strategies to enhance social vitality

The key to creating social vitality lies in the agglomeration effect of people and the occurrence of crowd activities, which create interaction between people and space. The social relationships established in rural communities are based on blood and geographic ties, and residents rely on longstanding relationships and familiar living environments to create good social vitality. However, the construction of farmers’ resettlement communities and the rise in heterogeneity of interaction objects have changed the original living environment and broken the original network of interpersonal relationships, leading to a subsequent decline in the frequency of interaction. In this paper, four aspects of improving the social vitality of public space in resettlement communities are to change the behavioral patterns of the crowd, to create a composite public space, to improve the public infrastructure, and to build a community network platform.

4.3

Strategies to enhance the vitality of the cultural economy

Community culture, as an important soft power, is conducive to promoting the formation of community perceptions, enhancing the sense of belonging to the community, and promoting community governance. However, the single form of cultural construction in the community and the low participation in public cultural activities are the main reasons leading to the problems of poor recognizability of the community, weak public spirit, and indifference to neighborhood relations. Community culture construction refers to a kind of group culture construction in the public space of the community, with residents as the main body, colorful cultural activities as the main content, and a sense of cultural identity as the core. Therefore, this paper mainly focuses on four aspects of community material culture, spiritual culture, behavioral culture and institutional culture to effectively enhance the vitality of community culture.

5

Conclusion

This paper is constructing a smart city evaluation system based on the combination of hierarchical analysis method and RBF neural network, aiming to provide guidance for the development of urban renewal planning programs and community vitality enhancement strategies through this system. The main findings of this paper in conducting the research are as follows:

The comprehensive evaluation value of Yumen city is 0.0545, 0.0313, 0.0436, 0.0328, 0.0499, 0.0369, 0.0194, 0.0365 and 0.058. The City of Otago has a combined assessed value of 0.0258, 0.0377, 0.0543, 0.0632, 0.0798, 0.0249, 0.0123, 0.0307, and 0.0632. This leads to the overall low level of urban renewal in Yumen City.

This paper proposes strategies for enhancing community vitality from three aspects: physical, social, and cultural space. In terms of physical space, inter-house public space, road public space, and central square space are targeted for renewal, in order to create a community public space environment with a strong vitality atmosphere. In terms of social space, the reconstruction of social network and the expansion of residents’ interaction channels are promoted by stimulating the behavioral patterns of the crowd, creating a composite public space, optimizing public infrastructure, and building a community network platform. In terms of cultural space, community cultural vitality is effectively enhanced from four perspectives: community material culture, spiritual culture, behavioral culture, and institutional culture, in order to promote the benign development of public space in farmers’ resettlement communities.

Through the research of this paper, we aim to evaluate and judge the level of smart urban renewal, help government departments control the current situation of urban renewal planning and development, provide reasonable strategies for community vitality enhancement, and promote the healthy and sustainable development of the city.

Lingua:: Inglese

Frequenza di pubblicazione:: 1 volte all'anno
Argomenti della rivista:: Scienze biologiche, Scienze della vita, altro, Matematica, Matematica applicata, Matematica generale, Fisica, Fisica, altro

Feed RSS della rivista

Optimization of Urban Renewal Planning Schemes and Community Vitality Enhancement Strategies Based on Deep Learning Algorithms and Smart City Evaluation

Xuan Han

Baishu Xia

Pubblicato online: 21 mar 2025

Ricevuto: 21 ott 2024

Accettato: 15 feb 2025

DOI: https://doi.org/10.2478/amns-2025-0598

Parole chiave<kwd>Erg theory</kwd>, <kwd>Hierarchical analysis</kwd>, <kwd>Neural network</kwd>, <kwd>Urban renewal</kwd>

© 2025 Xuan Han et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Parole chiave
<kwd>Erg theory</kwd>, <kwd>Hierarchical analysis</kwd>, <kwd>Neural network</kwd>, <kwd>Urban renewal</kwd>