Cross-regional city management method based on big data driving

Abstract

The invention provides a cross-regional city management method based on big data driving, and relates to the technical field of cross-regional city management. The cross-regional city management method comprises the following steps: collecting and preprocessing multi-source data of a target city; dividing the target city into N regions, and extracting time features, spatial features and space-time interaction features of the preprocessed data from the N regions; and constructing a prediction model based on the graph convolutional neural network, optimizing model hyper-parameters by using a decision tree algorithm, and outputting a prediction development trend and prediction occurred problems of the target city. According to the method, the target city is divided into N regions, the development condition and spatio-temporal evolution of each region of the city can be deeply analyzed, relevance and difference between the regions can be found, reference is provided for cross-region cooperation, and therefore driving factors and restricting factors of city development can be better recognized. And meanwhile, a graph convolutional neural network model is applied to mine a complex association rule between urban areas, so that the accuracy of a prediction result is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of cross-regional urban management, and in particular to a cross-regional urban management method driven by big data. Background Art

[0002] In today's society, with rapid urbanization and the continuous expansion of urban scale, urban management faces many challenges, such as increasingly severe traffic congestion, environmental pollution, and public safety. The traditional management model based on manual intervention has been unable to cope with the increasingly complex urban operation conditions.

[0003] Cross-regional urban management driven by big data is becoming an emerging solution. Using big data technology, we can achieve all-round perception and accurate analysis of all areas of the city, providing scientific decision-making support for city managers. However, in actual application, predictive analysis models often rely too much on static features and are difficult to fully reflect the complex dynamic evolution of the city. Summary of the invention

[0004] The present invention provides a cross-regional urban management method driven by big data to solve the defect that the prediction and analysis models in the prior art often rely too much on static features and are difficult to fully reflect the complex dynamic evolution laws of the city.

[0005] The present invention provides a cross-regional urban management method driven by big data, including: S1. Establish a data interface to collect multi-source data from the target city in real time; S2, preprocessing the collected multi-source data to obtain preprocessed data; S3, divide the target city into N regions, extract the temporal and spatial features of the N regions from the preprocessed data, and combine the temporal and spatial features to construct the spatiotemporal interaction features; S4. Use the prediction model based on graph convolutional neural network to output the prediction results of the target city, including: S41. Build a prediction model based on graph convolutional neural network and use decision tree algorithm to optimize the hyperparameters of the prediction model. S42, using cross-validation method to train the prediction model, and evaluating the prediction model by accuracy, recall and F1 value; S43, inputting the time features, space features and time-space interaction features into the trained prediction model, and outputting the predicted development trend and prediction problems of the target city; S5. Establish an emergency command center to determine whether the predicted development trend and predicted problems exceed the preset warning threshold. If so, the emergency command center will activate the emergency response mechanism; S6. Establish a continuous improvement mechanism, regularly evaluate the effectiveness of the management of target cities, analyze existing problems in urban management, and make dynamic improvements.

[0006] According to the cross-regional urban management method driven by big data provided by the present invention, in step S1, the multi-source data includes land use planning data, infrastructure construction planning data, traffic flow data, road congestion information, air quality data, water pollution monitoring data, industrial structure data, enterprise development related data, population size, age structure and population flow data of the target city.

[0007] According to the cross-regional urban management method driven by big data provided by the present invention, in step S2, the preprocessing of multi-source data includes: removing duplicate data and invalid data in the multi-source data; filling missing values ​​in the multi-source data with the mean filling method; and identifying and deleting abnormal data in the multi-source data using a box plot.

[0008] According to the cross-regional urban management method based on big data drive provided by the present invention, in step S3, the method for extracting time features includes: Fourier transform is used to extract seasonal characteristics in preprocessed data. The expression formula of Fourier transform is:

[0009] In the formula, represents time series data, , and represents the Fourier coefficient, T is the period; The autocorrelation function is used to analyze the periodic characteristics of the preprocessed data. The expression formula of the autocorrelation function is:

[0010] In the formula, k represents the lag period, is the mean of the data; Use the linear regression equation to fit the trend characteristics of the preprocessed data. The linear regression equation formula is:

[0011] In the formula, s is the intercept term and r is the slope term.

[0012] According to the cross-regional urban management method based on big data drive provided by the present invention, in step S3, the spatial feature extraction method includes: The geographically weighted regression model is used to represent the neighborhood characteristics between each region in N regions and its adjacent regions. The expression formula of the geographically weighted regression model is:

[0013] In the formula, is the target variable, is the explanatory variable, and is the local regression coefficient, is the random error term; The PageRank algorithm is used to calculate the centrality characteristics of each region in N regions. The formula of the PageRank algorithm is:

[0014] Where PR(i) is the PageRank value of region i, d is the damping coefficient, and L(j) is the out-degree of region j.

[0015] According to the cross-regional urban management method based on big data drive provided by the present invention, in step S41, the step of constructing a prediction model based on a graph convolutional neural network includes: S411, each of the N regions is represented as a node, and the attribute relationship is represented as an edge, so as to construct a graph structure of the target city; S412, construct multiple convolutional layers, each convolutional layer updates the feature representation of the node by aggregating the information of the node neighbors, and the graph convolution operation of the convolutional layer is expressed as:

[0016] In the formula, is the feature matrix of the next layer, It is The node feature matrix of the layer, It is The learnable parameter matrix of the layer, is the activation function, =A+I, where I is the identity matrix, A is the adjacency matrix, yes The degree matrix of S413. Construct a fully connected layer as the output layer of the prediction model. The fully connected layer is connected to the last convolutional layer to output the predicted value. The predicted value Y is expressed as:

[0017] In the formula, is the output feature matrix of the last convolutional layer, is the learnable parameter matrix of the output layer, and the softmax function is used to convert the output value into a probability distribution.

[0018] According to the cross-regional urban management method based on big data drive provided by the present invention, in step S41, the step of using the decision tree algorithm to optimize the hyperparameters of the prediction model includes: a. Use the convolutional layer size, number of layers, learning rate and regularization parameter as hyperparameters of the prediction model, and set the value range for each hyperparameter; b. Use random sampling to generate multiple hyperparameter combinations from the range of hyperparameter values. Each hyperparameter combination is used as a sample, the value of the hyperparameter is used as a feature, and the performance of the prediction model corresponding to the hyperparameter is used as a label. c. Build a relationship model between the hyperparameters of a decision tree and the performance of the prediction model through sample data, and split the decision tree using information gain. The formula for information gain is:

[0019] In the formula, M is the data set, a is the feature, is the information entropy of the data set M, are all possible values ​​of feature a, It is a subset of feature a whose value is v; The calculation formula of information entropy is:

[0020] In the formula, K is the number of categories, is the proportion of samples of the kth category in the data set; d. According to the decision tree, obtain the performance of the prediction model corresponding to the next set of hyperparameter combinations; e. Repeat steps b to d until the preset number of iterations is reached, and output the hyperparameter combination corresponding to the optimal performance of the prediction model as the hyperparameter combination of the prediction model.

[0021] According to the cross-regional urban management method based on big data drive provided by the present invention, in step S42, the step of using the cross-validation method to train the prediction model includes: S421, dividing the preprocessed data into K subsets of equal size; S422, select one of the subsets as a validation set and the remaining K-1 subsets as training sets; S423, perform K training and validation, and take the average of the K validation results as the final performance of the prediction model.

[0022] According to the cross-regional urban management method based on big data drive provided by the present invention, in step S5, the emergency response mechanism includes: Personnel deployment: quickly organize staff from relevant departments to participate in emergency work; Material resource allocation: allocation of emergency supplies, equipment and transportation; Financial support: Ensure that the funds needed for emergency work are in place in a timely manner.

[0023] According to the cross-regional urban management method driven by big data provided by the present invention, in step S6, the effect evaluation includes the comparison between the actual development of the target city and the predicted results and the execution effect of the emergency response mechanism; the dynamic improvement includes improving the quality issues of multi-source data by strengthening data collection and preprocessing, and improving the problems of the prediction model by adjusting the model structure and continuing to optimize the hyperparameters; and improving the problems of the emergency response mechanism by improving the process, strengthening training and drills.

[0024] The cross-regional urban management method based on big data drive provided by the present invention realizes comprehensive perception and integration of key data in various fields of the city by establishing a multi-source data interface. By dividing the target city into N regions, it is possible to deeply analyze the development status of each region of the city and its spatiotemporal evolution, discover the correlation and differences between regions, and provide a reference for cross-regional cooperation, so as to better identify the driving factors and constraints of urban development. By applying the graph convolutional neural network model, the complex correlation rules between urban areas are excavated to improve the accuracy of the prediction results. The prediction and analysis results are seamlessly connected with the emergency response mechanism to realize intelligent decision-making support for urban management. Provide comprehensive, accurate and intelligent management support for cities, realize the overall improvement of urban governance level, and inject lasting impetus into the sustainable development of cities. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0026] Figure 1 It is a flow chart of a cross-regional urban management method driven by big data provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0028] Combine the following Figure 1 The present invention describes a cross-regional urban management method driven by big data.

[0029] Figure 1It is a flow chart of a cross-regional urban management method driven by big data provided in an embodiment of the present invention.

[0030] like Figure 1 As shown, the cross-regional urban management method based on big data driving provided by an embodiment of the present invention includes: S1. Establish a data interface to collect multi-source data from the target city in real time.

[0031] Multi-source data include the target city’s land use planning data, infrastructure construction planning data, traffic flow data, road congestion information, air quality data, water pollution monitoring data, industrial structure data, enterprise development-related data, population size, age structure and population mobility data.

[0032] In this embodiment, by collecting the above data, the operating status of the target city can be fully displayed, providing a basis for subsequent analysis and decision-making, which is of great significance for promoting cross-regional urban management, improving scientific decision-making and management efficiency, and achieving sustainable development of cities.

[0033] S2. Clean the collected multi-source data, process missing values ​​and outliers, and obtain preprocessed data.

[0034] The multi-source data processing methods include: removing duplicate data and invalid data from multi-source data. Using the mean filling method to fill missing values. Using box plots to identify and delete abnormal data.

[0035] In this embodiment, the collected multi-source data is preprocessed by the above processing method, which can eliminate redundancy and noise in the data, improve the accuracy and reliability of the data, make up for the missing data, fully reflect the operating status of various systems in the city, and ensure that the analysis results are not interfered by abnormal values. After preprocessing the collected multi-source data, the actual situation of various fields in the target city can be more accurately portrayed, providing high-quality data support for subsequent urban analysis and decision-making.

[0036] S3. Divide the target city into N regions, extract the temporal and spatial features of the N regions from the preprocessed data, and construct spatiotemporal interaction features by combining the temporal and spatial features.

[0037] Temporal features include seasonal features, periodic features and trend features. Seasonal features use Fourier transform to extract seasonal periodic patterns in preprocessed data. By analyzing the size of Fourier coefficients, the main periodic components in preprocessed data are identified to extract seasonal features. The Fourier series expression formula is:

[0038] In the formula, represents time series data, , and represents the Fourier coefficient and T is the period.

[0039] The periodicity feature is to use the autocorrelation function to analyze the periodicity of the preprocessed data. The expression formula of the autocorrelation function is:

[0040] In the formula, k represents the lag period, is the mean of the data.

[0041] The trend feature is to use the linear regression equation to fit the trend line of the preprocessed data. The linear regression equation formula is:

[0042] In the formula, s is the intercept term and r is the slope term.

[0043] The spatial characteristics include the neighborhood characteristics and centrality characteristics of N regions. The neighborhood characteristics use the geographically weighted regression model to represent the difference characteristics between each region in the N regions and its adjacent regions. The representation formula of the geographically weighted regression model is:

[0044]

[0045] In the formula, is the target variable, is the explanatory variable, and is the local regression coefficient, is a random error term.

[0046] The centrality feature uses the PageRank algorithm to calculate the importance of each region in N regions. The formula of the PageRank algorithm is:

[0047] Where PR(i) is the PageRank value of region i, d is the damping coefficient, and L(j) is the out-degree of region j.

[0048] In this embodiment, by dividing the target city into N regions, it is possible to deeply analyze the development status of each region of the city and its spatiotemporal evolution, discover the correlation and differences between regions, provide a reference for cross-regional cooperation, and better identify the driving factors and constraints of urban development.

[0049] S4. Build a prediction model, input time features, space features and time-space interaction features, and output the prediction results of the target city, including: S41. Build a prediction model based on graph convolutional neural network and use decision tree algorithm to optimize the hyperparameters of the prediction model.

[0050] The steps to build a prediction model based on graph convolutional neural network include: S411. Represent each of the N regions as a node, represent the attribute relationship as an edge, and construct a graph structure of the target city.

[0051] S412. Construct multiple convolutional layers. Each convolutional layer updates the feature representation of the node by aggregating the information of the node's neighbors. The operation of one layer of graph convolution is expressed as:

[0052] In the formula, is the feature matrix of the next layer, It is The node feature matrix of the layer, It is The learnable parameter matrix of the layer, is the activation function, =A+I, where I is the identity matrix, A is the adjacency matrix, yes The degree matrix of .

[0053] S413. Construct a fully connected layer as the output layer of the prediction model. The fully connected layer is connected to the last convolutional layer and outputs the predicted value Y. The predicted value is expressed as:

[0054] In the formula, is the output feature matrix of the last convolutional layer, is the learnable parameter matrix of the output layer, and the softmax function is used to convert the output value into a probability distribution.

[0055] The steps to optimize the hyperparameters of a predictive model using the decision tree algorithm include: a. Use the convolutional layer size, number of layers, learning rate, and regularization parameter as hyperparameters of the prediction model, and set a value range for each hyperparameter.

[0056] b. Use random sampling to generate multiple hyperparameter combinations from the range of hyperparameter values. Each hyperparameter combination is used as a sample, the value of the hyperparameter is used as a feature, and the performance of the prediction model corresponding to the hyperparameter is used as a label.

[0057] c. Build a relationship model between the hyperparameters of a decision tree and the performance of the prediction model through sample data, and split the decision tree using information gain. The formula for information gain is:

[0058] In the formula, M is the data set, a is the feature, is the information entropy of the data set M, are all possible values ​​of feature a, is a subset of feature a whose value is v.

[0059] The calculation formula of information entropy is:

[0060] In the formula, K is the number of categories, is the proportion of samples of the kth class in the data set.

[0061] d. Based on the decision tree, obtain the performance of the prediction model corresponding to the next set of hyperparameter combinations.

[0062] e. Repeat steps b to d until the preset number of iterations is reached, and output the hyperparameter combination corresponding to the optimal performance of the prediction model as the hyperparameter combination of the prediction model.

[0063] S42. The prediction model was trained using the cross-validation method and evaluated using accuracy, recall, and F1 value.

[0064] The steps to train a prediction model using cross-validation include: S421. Divide the preprocessed data into K subsets of equal size.

[0065] S422. Select one of the subsets as a validation set and the remaining K-1 subsets as training sets.

[0066] S423, perform K training and validation, and take the average of the K validation results as the final performance of the prediction model.

[0067] The formula for accuracy is:

[0068] The formula for recall is:

[0069] The formula for the F1 value is:

[0070] in,

[0071] Among them, TP represents the number of samples correctly detected as positive examples, TN represents the number of samples correctly detected as negative examples, FP represents the number of samples falsely detected as positive examples, and FN represents the number of samples falsely detected as negative examples.

[0072] S43. Input the time features, space features and space-time interaction features into the trained prediction model, and output the predicted development trend and predicted problems of the target city.

[0073] In this embodiment, by constructing a graph convolutional neural network model, the complex spatiotemporal correlation characteristics between urban areas can be effectively captured. By optimizing hyperparameters through the decision tree algorithm, the model's ability to learn and generalize urban development laws can be improved. The cross-validation method is used for model training and evaluation to ensure the robustness and credibility of the prediction results.

[0074] At the same time, it utilizes complex spatiotemporal correlation characteristics to improve the accuracy and reliability of predictions, ensures the robustness of prediction results through model optimization and cross-validation, outputs urban development trends and potential problems, and provides intelligent support for decision-making.

[0075] S5. Set the warning threshold and establish an emergency command center. When the output of the prediction model exceeds the warning threshold, the emergency command center activates the emergency response mechanism.

[0076] The emergency response mechanism includes: personnel deployment, that is, quickly organizing the staff of relevant departments to participate in emergency work. Material resource deployment, that is, deploying emergency materials, equipment and transportation. Financial support, that is, ensuring that the funds needed for emergency work are in place in a timely manner.

[0077] For example, based on historical traffic data analysis, determine the warning thresholds for the average speed or congestion index of urban roads, such as when the average speed is less than 20km / h or the congestion index exceeds 0.8. Establish an emergency command center to coordinate emergency responses. Equip professional data analysts and decision managers to monitor urban traffic conditions in real time and analyze forecast results. When the traffic conditions output by the forecast model exceed the warning threshold, the emergency command center immediately activates the emergency response mechanism.

[0078] Quickly mobilize on-duty personnel to increase road control. Deploy traffic command and dispatch personnel to promptly direct and relieve on-site congestion. Mobilize volunteers to participate in traffic diversion and guide vehicles to pass in an orderly manner.

[0079] Call on emergency material reserves, such as variable information signs, traffic cones, warning lights, etc. Provide emergency evacuation vehicles to divert stranded vehicles. Call on emergency rescue vehicles to ensure rapid processing of the accident scene.

[0080] Urgently allocate funds needed for emergency response to ensure smooth progress of all work. Coordinate and allocate budget funds to ensure the entire emergency response process. Actively seek financial support to provide guarantees for long-term emergency preparedness.

[0081] In this embodiment, by setting early warning thresholds, major risks and problems in urban development can be discovered and responded to in a timely manner, thereby improving the efficiency and coordination capabilities of urban emergency management, maintaining urban safety and stability, and protecting the lives and property of the people.

[0082] S6. Establish a continuous improvement mechanism, regularly evaluate the management effectiveness of target cities, analyze existing problems, and make dynamic improvements.

[0083] The effect evaluation includes the comparison between the actual development of the target city and the forecast results and the implementation effect of the emergency response mechanism. Dynamic improvement includes improving the quality of multi-source data by strengthening data collection and preprocessing, and improving the forecast model by adjusting the model structure and continuing to optimize hyperparameters. For the emergency response mechanism, it is improved by improving the process, strengthening training and drills.

[0084] By establishing a continuous improvement mechanism, we can achieve closed-loop optimization of urban management, continuously improve the management level, promptly discover and correct problems and deficiencies in management, and ensure that urban management plans meet actual needs and meet the requirements of sustainable development.

[0085] In summary, this embodiment provides a cross-regional urban management method driven by big data, which realizes the comprehensive perception and integration of key data in various fields of the city by establishing a multi-source data interface. Dividing the target city into N regions can deeply analyze the development status of each region of the city and its spatiotemporal evolution, discover the correlation and differences between regions, and provide a reference for cross-regional cooperation, so as to better identify the driving factors and constraints of urban development. Apply the graph convolutional neural network model to explore the complex correlation patterns between urban areas and improve the accuracy of the prediction results. Seamlessly connect the prediction and analysis results with the emergency response mechanism to realize intelligent decision-making support for urban management. Provide comprehensive, accurate and intelligent management support for cities, realize the overall improvement of urban governance level, and inject lasting impetus into the sustainable development of cities.

[0086] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A cross-regional urban management method driven by big data, characterized in that the steps include: S1. Establish a data interface to collect multi-source data from the target city in real time; S2. Preprocessing the collected multi-source data to obtain preprocessed data; S3, dividing the target city into N regions, extracting the time features and spatial features of the N regions from the preprocessed data, and combining the time features and the spatial features to construct a spatiotemporal interaction feature; S4. Using a prediction model based on a graph convolutional neural network to output the prediction result of the target city, including: S41, constructing a prediction model based on a graph convolutional neural network, and using a decision tree algorithm to optimize the hyperparameters of the prediction model; S42, training the prediction model using a cross-validation method, and evaluating the prediction model using accuracy, recall, and F1 value; S43, inputting the time feature, the space feature and the time-space interaction feature into the trained prediction model, and outputting the predicted development trend and prediction problem of the target city; S5. Establish an emergency command center to determine whether the predicted development trend and the predicted problem exceed a preset warning threshold, and if so, the emergency command center initiates an emergency response mechanism; S6. Establish a continuous improvement mechanism, regularly evaluate the effectiveness of the management of the target cities, analyze existing problems in urban management, and make dynamic improvements.

2. The cross-regional urban management method based on big data drive according to claim 1 is characterized in that: In step S1, the multi-source data includes the target city's land use planning data, infrastructure construction planning data, traffic flow data, road congestion information, air quality data, water pollution monitoring data, industrial structure data, enterprise development related data, population size, age structure and population mobility data.

3. The cross-regional urban management method based on big data drive according to claim 1 is characterized in that: In step S2, the preprocessing of the multi-source data includes: removing duplicate data and invalid data in the multi-source data; filling missing values ​​in the multi-source data using a mean filling method; and identifying and deleting abnormal data in the multi-source data using a box plot.

4. The cross-regional urban management method based on big data drive according to claim 1 is characterized in that: In step S3, the method for extracting the time feature includes: Fourier transform is used to extract seasonal characteristics in the preprocessed data. The expression formula of Fourier transform is: In the formula, represents time series data, , and represents the Fourier coefficient, T is the period; The autocorrelation function is used to analyze the periodic characteristics of the preprocessed data. The expression formula of the autocorrelation function is: In the formula, k represents the lag period, is the mean of the data; A linear regression equation is used to fit the trend characteristics of the preprocessed data. The linear regression equation formula is: In the formula, s is the intercept term and r is the slope term.

5. The cross-regional urban management method based on big data drive according to claim 1 is characterized in that: In step S3, the spatial feature extraction method includes: A geographically weighted regression model is used to represent the neighborhood characteristics between each region in the N regions and its adjacent regions. The representation formula of the geographically weighted regression model is: In the formula, is the target variable, is the explanatory variable, and is the local regression coefficient, is the random error term; The PageRank algorithm is used to calculate the centrality characteristics of each region in the N regions. The formula of the PageRank algorithm is: Where PR(i) is the PageRank value of region i, d is the damping coefficient, and L(j) is the out-degree of region j.

6. The cross-regional urban management method based on big data drive according to claim 1 is characterized in that: In step S41, the step of constructing a prediction model based on a graph convolutional neural network includes: S411, representing each of the N regions as a node, representing the attribute relationship as an edge, and constructing a graph structure of the target city; S412, construct multiple convolutional layers, each convolutional layer updates the feature representation of the node by aggregating the information of the node neighbors, and the graph convolution operation of the convolutional layer is expressed as: In the formula, is the feature matrix of the next layer, It is The node feature matrix of the layer, It is The learnable parameter matrix of the layer, is the activation function, =A+I, where I is the identity matrix, A is the adjacency matrix, yes The degree matrix of S413: construct a fully connected layer as the output layer of the prediction model, the fully connected layer is connected to the last convolutional layer, and outputs a prediction value, the prediction value Y is expressed as: In the formula, is the output feature matrix of the last convolutional layer, is the learnable parameter matrix of the output layer, and the softmax function is used to convert the output value into a probability distribution.

7. The cross-regional urban management method based on big data drive according to claim 1 is characterized in that: In step S41, the step of optimizing the hyperparameters of the prediction model using a decision tree algorithm includes: a. Using the convolutional layer size, number of layers, learning rate and regularization parameter as hyperparameters of the prediction model, and setting a value range for each hyperparameter; b. Generate multiple hyperparameter combinations from the value range of the hyperparameter using a random sampling method, wherein each hyperparameter combination is used as a sample, the value of the hyperparameter is used as a feature, and the performance of the prediction model corresponding to the hyperparameter is used as a label; c. Constructing a relationship model between the hyperparameters based on a decision tree and the performance of the prediction model through the sample data, and splitting the decision tree using information gain, wherein the formula of the information gain is expressed as: In the formula, M is the data set, a is the feature, is the information entropy of the data set M, are all possible values ​​of feature a, It is a subset of feature a whose value is v; The calculation formula of information entropy is: In the formula, K is the number of categories, is the proportion of samples of the kth category in the data set; d. According to the decision tree, obtaining the performance of the prediction model corresponding to the next set of hyperparameter combinations; e. Repeat steps b to d until a preset number of iterations is reached, and output the hyperparameter combination corresponding to the optimal performance of the prediction model as the hyperparameter combination of the prediction model.

8. The cross-regional urban management method based on big data drive according to claim 1 is characterized in that: In step S42, the step of training the prediction model using the cross-validation method includes: S421, dividing the preprocessed data into K subsets of equal size; S422, select one of the subsets as a validation set and the remaining K-1 subsets as training sets; S423, performing K training and validation, and taking the average of the K validation results as the final performance of the prediction model.

9. The cross-regional urban management method based on big data drive according to claim 1 is characterized in that: In step S5, the emergency response mechanism includes: Personnel deployment: quickly organize staff from relevant departments to participate in emergency work; Material resource allocation: allocation of emergency supplies, equipment and transportation; Financial support: Ensure that the funds needed for emergency work are in place in a timely manner.

10. The cross-regional urban management method based on big data drive according to claim 1 is characterized in that: In step S6, the effect evaluation includes the comparison between the actual development situation of the target city and the predicted results and the implementation effect of the emergency response mechanism; the dynamic improvement includes improving the quality issues of the multi-source data by strengthening data collection and preprocessing, and improving the problems of the prediction model by adjusting the model structure and continuing to optimize the hyperparameters; and improving the problems of the emergency response mechanism by improving the process, strengthening training and drills.

Images