Machine learning enhanced city comprehensive vitality evaluation method and system

Abstract

The invention discloses a machine learning enhanced urban comprehensive vitality evaluation method and system, and the method comprises the following steps: 1, constructing a basic database, 2, constructing a comprehensive vitality evaluation index system, 3, constructing an evaluation model and comparing multiple models, 4, analyzing the spatial distribution characteristics of vitality, and 5, evaluating the comprehensive vitality. Step 5, proposal of vitality promotion oriented space update; according to the method, multi-source geographic big data is utilized, a more comprehensive comprehensive vitality evaluation index system is provided, urban comprehensive vitality evaluation is enhanced by applying a machine learning combination model, a comprehensive vitality index is optimized and improved, vitality distribution characteristics are clarified through local space self-correlation analysis, and the comprehensive vitality evaluation efficiency is improved. The problems that in the prior art, evaluation indexes are not comprehensive, an evaluation method needs to be improved, evaluation accuracy needs to be enhanced, and a city internal activity distribution structure needs to be deeply excavated can be solved, and scientific basis and reference can be provided for updating planning under activity promotion guidance.

Description

Technical Field

[0001] The present invention belongs to the technical field of urban comprehensive vitality evaluation and analysis, and specifically relates to a machine learning enhanced urban comprehensive vitality evaluation method and system. Background Art

[0002] Urban vitality is the explicit manifestation of urban prosperity and development, and is the comprehensive externalization of the economic, cultural and social core. Stock renewal has become a new important means of development for many cities around the world, and is a key strategy to re-stimulate urban vitality; adjusting the urban functional structure, improving traffic conditions, and improving the living environment by optimizing the built environment, and then reshaping urban vitality, is an important goal of renewal planning and development. Therefore, scientific evaluation of urban vitality is of great significance to the sustainable development of urban renewal.

[0003] Although many scholars have used Jacobs' theoretical framework to construct an indicator system for vitality evaluation and have carried out evaluation practices in different cities, there are still problems such as incomplete evaluation indicators, evaluation methods need to be improved, evaluation accuracy needs to be enhanced, and the internal vitality distribution structure of the city needs to be further explored. Therefore, constructing a more comprehensive urban vitality comprehensive evaluation indicator system and using machine learning enhanced evaluation methods will help obtain more accurate and detailed evaluation results, which will be particularly important for urban decision makers and planners to formulate and update relevant strategies. Summary of the invention

[0004] The purpose of the present invention is to provide a machine learning enhanced urban comprehensive vitality evaluation method and system, which solves the problem of how to establish a more comprehensive and adaptive urban comprehensive vitality evaluation index system based on Jacobs' urban vitality theory, and then use machine learning algorithms to improve the accuracy of urban vitality evaluation.

[0005] To achieve the above object, the present invention provides the following technical solution: a machine learning enhanced urban comprehensive vitality evaluation method, comprising the following steps:

[0006] Step 1: Basic database construction: Use 1km square grid to build analysis units and integrate multi-source geographic big data;

[0007] Step 2: Construction of comprehensive vitality evaluation index system: Based on the primary factors affecting vitality proposed by Jacobs and combined with accessibility characteristics, a comprehensive vitality evaluation index system covering 5 dimensions, 11 first-level indicators and 22 specific second-level indicators is constructed;

[0008] Step 3: Evaluation model construction and multi-model comparison: Use game theory to determine the subjective and objective combination weights, use the ideal solution sorting method to calculate the preliminary comprehensive vitality index, and combine the analysis performance of various machine learning models such as particle swarm, extreme learning machine, and multi-kernel function to obtain the optimal comprehensive vitality index;

[0009] Step 4: Analysis of spatial distribution characteristics of vitality: Use the local spatial autocorrelation method to analyze the optimized comprehensive vitality index to reveal the spatial distribution and structural characteristics of vitality;

[0010] Step 5: Spatial renewal suggestions guided by vitality promotion: Based on the distribution characteristics of urban vitality, urban space renewal suggestions guided by vitality promotion are put forward.

[0011] As a preferred technical solution of the present invention, the multi-source geographic big data includes data on land use, road data, points of interest, buildings, etc.

[0012] As a preferred technical solution of the present invention, the construction of the basic database also includes aggregating data such as population, building age and housing prices, Baidu heat map, etc. into a grid.

[0013] As a preferred technical solution of the present invention, the built environment related data includes buildings, points of interest, green spaces, road networks and subway stations.

[0014] As a preferred technical solution of the present invention, in the construction of the comprehensive vitality evaluation index system, the index composition is refined with reference to existing literature evidence, and accessibility-related characteristic indicators are also incorporated into the evaluation system.

[0015] As a preferred technical solution of the present invention, in the construction of the evaluation model and the comparison of multiple models, a game theory method is used to combine subjective and objective weights, and the combined weight is obtained by selecting the minimum deviation between the subjective weight and the objective weight.

[0016] As a preferred technical solution of the present invention, in the evaluation model construction and multi-model comparison, a swarm intelligence optimization algorithm is introduced to optimize the network structure of the extreme learning machine.

[0017] As a preferred technical solution of the present invention, the swarm intelligence optimization algorithm includes simulated annealing, genetic algorithm, differential evolution, artificial bee colony and particle swarm optimization.

[0018] As a preferred technical solution of the present invention, in the evaluation model construction and multi-model comparison, a multi-kernel function based on a polynomial kernel function and a radial basis function kernel function is constructed to improve the prediction accuracy of the machine learning model.

[0019] The present invention also discloses a machine learning enhanced urban comprehensive vitality evaluation system, comprising

[0020] Basic database construction module: responsible for the collation and preprocessing of basic data, including the use of 1km square grids to construct analysis units and integrate multi-source geographic big data;

[0021] Comprehensive vitality evaluation index system construction module: Based on the primary factors affecting vitality proposed by Jacobs and combined with accessibility characteristics, a comprehensive vitality evaluation index system covering 5 dimensions, 11 first-level indicators and 22 specific second-level indicators is constructed;

[0022] Evaluation model construction and multi-model comparison module: game theory method is used to determine the subjective and objective combination weights, the ideal solution sorting method is used to calculate the preliminary comprehensive vitality index, and the preliminary comprehensive vitality index is optimized and predicted by combining machine learning models such as particle swarm, extreme learning machine, and multi-kernel function. The analysis performance of multiple machine learning combination models is compared to obtain the optimal comprehensive vitality index;

[0023] Spatial distribution characteristics analysis module of vitality: Use local spatial correlation index to analyze the optimized composite vitality index to reveal the spatial distribution and structural characteristics of vitality;

[0024] Optimization suggestion generation module: Based on the distribution characteristics and spatial autocorrelation of urban vitality, optimization suggestions for urban space renewal are put forward.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] Based on Jacobs' urban vitality theory and using multivariate geographic big data, this paper proposes a more comprehensive comprehensive vitality evaluation index system, which improves the richness of measurement indicators. On the basis of the preliminary evaluation based on the subjective and objective combined weights, it further uses multiple machine learning combination models to optimize and improve the comprehensive vitality index, and obtains the optimal comprehensive vitality index by comparing the model performance, which significantly improves the reliability and accuracy of the evaluation. Based on the optimized comprehensive vitality index, local spatial correlation analysis is carried out, which is conducive to summarizing the structural characteristics of the vitality distribution within the city, and can provide a basis and reference for urban space renewal under the guidance of vitality promotion. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a technical flow chart of the present invention;

[0028] Figure 2 It is a spatial distribution characteristic diagram of the preliminary comprehensive vitality index of the present invention;

[0029] Figure 3 It is a comparative analysis diagram of the preliminary and optimized comprehensive vitality indexes of the present invention;

[0030] Figure 4This is a spatial distribution characteristic diagram of the optimal comprehensive vitality index of the present invention. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] See also Figure 1 The present invention provides a machine learning enhanced urban comprehensive vitality evaluation method, comprising the following steps:

[0033] Step 1: Construction of basic database: A 1km square grid was used for analysis, with a total of 498 analysis sample units, integrating multi-source geographic big data, and collecting open data on land use, road data, points of interest (POI) and buildings as the basis for the study; multi-source data were collected to conduct empirical analysis; the first data set was 100mX100m grid population data; the second data set was about the age of buildings and housing prices, which came from a housing transaction platform that used Geocoding to locate the space and calculated the average year of the building and the average price of second-hand houses in the unit grid; the third data set was Baidu heat map; it was based on location service functions, and the remaining data sets were related to the built environment, including buildings, points of interest, green spaces, road networks and subway stations; all basic data were aggregated into grids to carry out indicator measurement, evaluation and analysis, and the basic processing of related data was completed in ArcGIS10.8;

[0034] Step 2: Construction of comprehensive vitality evaluation index system: Based on the four primary factors proposed by Jacobs, namely mixed use, density, short blocks and historical buildings, taking into account that location characteristics and whether public transportation is convenient have an important impact on the travel behavior of local residents and tourists, the secondary factor proposed by Jacobs (i.e. accessibility) is also included in the evaluation system; based on this, the index composition is refined with reference to existing literature evidence, and finally a comprehensive vitality evaluation index system covering 5 dimensions, 11 primary indicators and 22 specific secondary indicators is constructed (Table 1);

[0035] Table 1 Urban comprehensive vitality evaluation index system

[0036]

[0037] Step 3: Construction of evaluation model and comparison of multiple models: In order to achieve the optimal solution, the idea of ​​game theory is adopted, and subjective weight (AHP) and objective weight (CRITIC) are regarded as two participants in the game; by minimizing the sum of the deviations between the final weight and the subjective weight and the objective weight, the combination coefficient is determined to form a combination weight with complementary advantages in subjective and objective aspects; TOPSIS is a multi-attribute decision analysis method designed to help decision makers make choices among multiple alternatives; the method is based on the following principles: the optimal solution should be as close to the ideal solution as possible and as far away from the negative ideal solution as possible; the ideal solution consists of the best value of each criterion, while the negative ideal solution is the opposite, and then the ideal solution ranking method (TOPSIS) is used to calculate the vitality score, that is, the preliminary comprehensive vitality index (CVI);

[0038] Empowerment Method

[0039] 1) AHP

[0040] AHP is a multi-criteria decision-making method, which includes the construction of a hierarchical model of three parts: target layer, criterion layer and solution layer. In the present invention, AHP is only used to obtain the weight of each indicator, so only the target layer and criterion layer are used. Ten experts from relevant fields of universities and scientific research institutions were invited to score the importance of the indicators, and the questionnaire scoring sheet was sent by email, and 10 valid scoring sheets were received. The YAAHP12.11 software was used to calculate the weight of each indicator based on the expert survey data, and all consistency ratio results were less than 0.1, and the judgment results were consistent.

[0041] 2)CRITIC

[0042] The CRITIC method can reveal the characteristics of the distribution of differences between indicators and calculate the weights according to the contrast intensity and conflict degree between indicators. In this method, the contrast intensity is represented by the standard deviation. The larger the standard deviation, the more obvious the difference and the more significant the impact on the comprehensive vitality of the city. The conflict between indicators is measured by the correlation coefficient. If there is a strong positive correlation between two indicators, it means that their conflict is small, so their weight will be lower.

[0043] 3) Combination weighting method based on game theory

[0044] When using AHP and CRITIC weighting methods, they are easily affected by subjective bias and discrete extreme values; the traditional combined weighting method introduces custom coefficients when integrating subjective and objective weights, which may amplify or reduce the influence of a certain weighting method on the result in the combined weighting method; in order to achieve the optimal solution, the present invention adopts the idea of ​​game theory, takes subjective and objective weights as the two parties of the game, minimizes the difference between the final weight and the sum of subjective and objective weights, determines the combined coefficient, and forms a combined weight that complements the subjective and objective advantages;

[0045] TOPSIS method

[0046] TOPSIS is a multi-attribute decision analysis method used to help decision makers choose among multiple options; the method is based on the following principles: the optimal solution should be closest to the ideal solution and farthest from the negative ideal solution; the positive ideal solution consists of the optimal value of each criterion, and vice versa;

[0047] The preliminary CVI is a benchmark for training and testing within the framework of the extreme learning machine (ELM). During the training process, swarm intelligence optimization algorithms such as simulated annealing (SA), genetic algorithm (GA), differential evolution (DE), artificial bee colony (ABC), and particle swarm optimization (PSO) were introduced to optimize the network structure of the ELM. Considering the characteristics of global functions, i.e., the polynomial kernel function exhibits strong generalization performance but weak learning ability, and the characteristics of local functions, i.e., the radial basis function (RBF) exhibits strong learning ability but weak generalization performance, a multi-kernel function is constructed based on the Mercer property of the kernel function. The multi-kernel function is a weighted combination of the polynomial kernel function and the RBF kernel function. These models are then evaluated by correlation coefficients and model fitting to achieve a high-precision urban vitality evaluation. Ten multi-kernel extreme learning or optimized multi-kernel extreme learning algorithms are used to optimize the prediction of the vitality index calculated by AHP-CRITIC-TOPSIS, and the model performance is compared. Subsequently, the results of the best-performing model are used as the optimized CVI.

[0048] Figure 2 The figure shows the fit between the predicted value and the actual value and the prediction model parameters. It can be seen that the model with MK has a significantly better performance than other models. 2 The PSO–MK–ELM model has a higher R value than the PSO–MK–ELM model, which has a higher R value than the PSO–MK–ELM model. 2 The highest value (0.977) and the lowest MSE (0.0001) indicate that the model performs best and the reliability of the analysis results is relatively the best; therefore, the present invention conducts spatial distribution characteristic analysis based on the comprehensive vitality index after PSO-MK-ELM optimization.

[0049] Table 2 shows the data changes of the CVI index before and after optimization. It can be seen that after the optimization of the machine learning algorithm, the minimum value, 1 / 4 value, average value and maximum value of CVI have slightly increased by 0.005, 0.001, 0.0002 and 0.002 respectively; the 3 / 4 value has decreased by 0.003; at the same time, the standard deviation has also decreased from 0.076 to 0.075, indicating that the optimized CVI is more normally distributed.

[0050] Table 2 Descriptive statistical analysis of the preliminary comprehensive vitality index and the optimized comprehensive vitality index

[0051]

[0052]

[0053] Machine Learning Algorithms (Swarm Intelligence-MK-ELM)

[0054] 1) ELM

[0055] ELM is a single-layer feedforward neural network (SLFNs) learning algorithm based on Artificial Neural Networks. When determining model parameters, ELM can randomly initialize the connection weights (w) and neuron thresholds (b) between the input layer and the hidden layer to obtain a unique optimal solution. ELM is widely used in many fields for its superior self-learning ability, intelligent and efficient performance, and local fixed search advantages.

[0056] The basic topological structure of the ELM model consists of an input layer, a hidden layer, and an output layer, where n, H, and m represent the number of nodes in the input layer, the hidden layer, and the output layer, respectively; and g(x) is the neuron activation function in the hidden layer;

[0057] 2) MK-ELM

[0058] When the feature mapping function g(x) is unknown, the kernel function is introduced into ELM, which can form a new kernel-based kernel extreme learning machine method and define the kernel matrix;

[0059] 3)Swarm Intelligence(SA / GA / DE / ABC / PSO)

[0060] 1) SA simulated annealing optimization algorithm

[0061] Simulated Annealing (SA) is a random optimization algorithm based on Monte Carlo iterative solution. By using the SA algorithm to optimize the input weights and thresholds of hidden layer neurons, the classification accuracy and generalization ability of the model can be improved. The mean square error of the classification results is selected as the optimized fitness function.

[0062] 2) GA particle swarm optimization algorithm

[0063] Genetic algorithm (GA) is a computer model that simulates natural evolution and is usually used to find the best solution. It first encodes the parameters to be optimized to form a series population. Then, according to the fitness function, individuals are screened through selection, crossover and mutation. Finally, the individual with the best fitness is selected. The weight w and threshold b of ELM can be adjusted to improve the generalization performance.

[0064] 3) DE particle swarm optimization algorithm

[0065] Differential Evolution (DE) is a new optimization algorithm based on swarm intelligence theory. It achieves the purpose of search through cooperation and competition between species. Compared with other optimization algorithms, DE has the advantages of less parameter adjustability, generation information sharing, and strong global search convergence ability. It can efficiently solve optimization problems in complex environments.

[0066] 4) ABC Particle Swarm Optimization Algorithm

[0067] The artificial bee colony algorithm (ABC) is used to find the most appropriate number of hidden neurons in ELM. The algorithm simulates the intelligent foraging behavior of bees swinging their arms. It is a very simple three-dimensional population based on a random optimization algorithm. Each nectar source corresponds to a feasible solution, and the fitness of the nectar source corresponds to the fitness of the feasible solution. The algorithm introduces three types of bee colonies: employed bees, onlooker bees, and scout bees to find nectar sources. The ABC algorithm only uses fitness as the basis in the optimization process and does not require additional information. It also has no requirements on the type of optimization function. It has the characteristics of few control parameters, high search accuracy, and strong robustness.

[0068] 5) Particle Swarm Optimization Algorithm

[0069] The PSO (particle swarm optimization algorithm, PSO) algorithm belongs to the category of swarm intelligence technology. Particles move in multidimensional space at an adaptive speed. In PSO, particles remember the best position in the past and the best position the particle has ever reached. This property helps particles search multidimensional space faster. It can better perform global exploration, while smaller values ​​lead to fine search in the solution space.

[0070] Step 4: Analysis of spatial distribution characteristics of vitality: The optimized CVI of the optimal model was selected to further conduct local spatial association (LISA) analysis to reveal the spatial distribution and aggregation patterns;

[0071] The spatial structure of the vitality of the main urban area is obvious, forming a main vitality center in the middle of the research area, with the secondary vitality node surrounded by two circles outside it, and the vitality expanding outward along the cross axis ( Figure 3 );

[0072] Figure 4 The spatial autocorrelation of the optimized vitality index in the local area is shown. According to the analysis results of LISA, the HH agglomeration area is mainly located in the central and south-central parts of the study area, forming a large and continuous area (108.478km2, accounting for 23.710%). At the same time, there are three HH grid units in the northern part of the study area, and the corresponding areas are distributed with important external transportation facilities (high-speed rail stations), government services and life service facilities, thus showing a high vitality aggregation. In addition, there are four contiguous LL agglomeration areas outside the study area, with a total area of ​​about 140.401km2 (accounting for 30.688%); these areas are relatively underdeveloped, relatively biased, and lack relevant built environment elements to create vitality.

[0073] LISA

[0074] LISA (Local Spatial Autocorrelation) is a statistical method used to identify local autocorrelation features in spatial data; its principle is to reveal the spatial aggregation and anomalies of data by analyzing the similarities between spatial units.

[0075] Step 5: Spatial renewal recommendations oriented towards promoting vitality: Based on the distribution characteristics of urban vitality and the local spatial autocorrelation, the structural characteristics of urban vitality are clarified, areas with insufficient vitality are identified, and urban space renewal strategies are provided to promote vitality enhancement.

[0076] Relevant countermeasures include: 1) Cities with similar scale and construction characteristics to Xi'an can focus their renewal and optimization on improving the supporting living and tourism services and improving basic transportation conditions; 2) Relying on rail transit stations to carry out stock renewal outside the urban core area, promote the emergence of secondary vitality areas, and develop a multi-polar vitality structure; 3) Increasing road network density and improving the degree of land use mix through renewal, and regulating community rents to attract more people to live, may play a positive role in enhancing the vitality of urban space.

[0077] Although the embodiments of the present invention have been shown and described, as detailed above, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A machine learning enhanced urban comprehensive vitality evaluation method, characterized by: The steps include: Step 1: Basic database construction: Use 1km square grid to build analysis units and integrate multi-source geographic big data; Step 2: Construction of comprehensive vitality evaluation index system: Based on the primary factors affecting vitality proposed by Jacobs and combined with accessibility characteristics, a comprehensive vitality evaluation index system covering 5 dimensions, 11 first-level indicators and 22 specific second-level indicators is constructed; Step 3: Evaluation model construction and multi-model comparison: Use game theory to determine the subjective and objective combination weights, use the ideal solution sorting method to calculate the preliminary comprehensive vitality index, and combine particle swarm, extreme learning machine, multi-kernel function and other machine learning models to optimize the prediction of the preliminary comprehensive vitality index, compare the analysis performance of various machine learning combination models, and obtain the optimal comprehensive vitality index; Step 4: Analysis of spatial distribution characteristics of vitality: Use the local spatial autocorrelation method to analyze the optimized comprehensive vitality index to reveal the spatial distribution and structural characteristics of vitality; Step 5: Spatial renewal suggestions guided by vitality promotion: Based on the distribution characteristics of urban vitality, urban space renewal suggestions guided by vitality promotion are put forward.

2. The method for evaluating comprehensive urban vitality enhanced by machine learning according to claim 1, characterized in that: The multi-source geographic big data includes land use, urban roads, points of interest and building data.

3. The method for evaluating comprehensive urban vitality enhanced by machine learning according to claim 1, characterized in that: The construction of the basic database also includes aggregating population, building age, housing prices and Baidu heat map data into a grid.

4. The method for evaluating comprehensive urban vitality enhanced by machine learning according to claim 3 is characterized by: The built environment related data includes buildings, points of interest, green spaces, road networks and subway stations.

5. The method for evaluating comprehensive urban vitality enhanced by machine learning according to claim 1, characterized in that: In constructing the comprehensive vitality evaluation index system, the index composition is refined with reference to existing literature evidence, and accessibility-related characteristic indicators are also incorporated into the evaluation system.

6. The method for evaluating comprehensive urban vitality enhanced by machine learning according to claim 1, characterized in that: In the construction of the evaluation model and the comparison of multiple models, the game theory method is used to combine the subjective and objective weights, and the combined weight is obtained by selecting the minimum deviation between the subjective weight and the objective weight.

7. The method for evaluating comprehensive urban vitality enhanced by machine learning according to claim 1, characterized in that: In the evaluation model construction and multi-model comparison, a swarm intelligence optimization algorithm is introduced to optimize the network structure of the extreme learning machine.

8. The method for evaluating comprehensive urban vitality enhanced by machine learning according to claim 7, characterized in that: The swarm intelligence optimization algorithms include simulated annealing, genetic algorithm, differential evolution, artificial bee colony and particle swarm optimization.

9. The method for evaluating comprehensive urban vitality enhanced by machine learning according to claim 1, characterized in that: In the evaluation model construction and multi-model comparison, a multi-kernel function based on a polynomial kernel function and a radial basis function kernel function is constructed to improve the prediction accuracy of the machine learning model.

10. A machine learning-enhanced urban comprehensive vitality evaluation system, characterized by: include Basic database construction module: responsible for the collation and preprocessing of basic data, including the use of 1km square grids to construct analysis units and integrate multi-source geographic big data; Comprehensive vitality evaluation index system construction module: Based on the primary factors affecting vitality proposed by Jacobs and combined with accessibility characteristics, a comprehensive vitality evaluation index system covering 5 dimensions, 11 first-level indicators and 22 specific second-level indicators is constructed; Evaluation model construction and multi-model comparison module: game theory method is used to determine the subjective and objective combination weights, the ideal solution sorting method is used to calculate the preliminary comprehensive vitality index, and the preliminary comprehensive vitality index is optimized and predicted by combining particle swarm, extreme learning machine, multi-kernel function and other machine learning models. The analysis performance of various machine learning combination models is compared to obtain the optimal comprehensive vitality index; Spatial distribution characteristics analysis module of vitality: the optimized comprehensive vitality index is analyzed using the local spatial autocorrelation method to reveal the spatial distribution and structural characteristics of vitality; Spatial renewal proposal model guided by vitality promotion: Based on the distribution characteristics of urban vitality, urban spatial renewal proposals guided by vitality promotion are put forward.

Images