A deep learning-based urban governance dynamic analysis system and method

Abstract

The application discloses a kind of urban governance dynamic analysis system and method based on deep learning, it is related to governance dynamic analysis technical field, for solving in complex event frequently occurs or multi-source event bursts in situation, resource allocation scheme lacks the problem of accuracy, by collecting the event report and resource scheduling data of urban governance platform, resource scheduling response rate and event processing completion rate are calculated, and weighted to form comprehensive efficiency value, to judge current governance efficiency state.Efficiency is lower than threshold value, call pending event and governance resource, divided into governance information unit and collected its operating state;Extract historical scheduling resource quantity and non-repeated reporting rate, construct time series feature vector, use recurrent neural network to extract resource potential characteristics, combined with current resource to generate local strategy coefficient, strategy matching and conflict judgment are carried out, when conflict, based on position matching optimization configuration, realize the dynamic evaluation and tuning of governance efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of governance dynamic analysis, more particularly, the present application relates to a city governance dynamic analysis system and method based on deep learning. BACKGROUND

[0002] With the continuous expansion of the city scale and the continuous improvement of the information governance level, the city management platform gradually accesses a large amount of event reporting and resource scheduling data, which is used to support event response and public resource allocation in the city operation process. The city governance system is widely deployed in traffic, environmental sanitation, municipal, emergency and other governance scenarios, and through the interface of multi-source sensing devices, government information systems and mass interaction ports, real-time sensing and problem discovery of the city operation state are realized.

[0003] The prior art has the following disadvantages:

[0004] At present, the evaluation of city governance efficiency mainly depends on fixed threshold or static rules, which does not fully consider the dynamic change characteristics of the resource capacity of each governance unit, resulting in lack of accuracy of resource allocation scheme in the case of complex event frequent occurrence or multi-source event concentrated outbreak, therefore, a city governance dynamic analysis system and method based on deep learning are proposed.

[0005] The above information disclosed in the background section is only used to strengthen the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a city governance dynamic analysis system and method based on deep learning, which solves the problems proposed in the above background technology by using a recurrent neural network model to model the historical governance data in time sequence and extract resource potential characteristics.

[0007] To achieve the above-mentioned purpose, the present application provides the following technical scheme, a city governance dynamic analysis method based on deep learning, comprising the following steps:

[0008] Step S1: detecting the event processing state to evaluate the city governance efficiency, determining whether to retrieve the event data to be processed and the city governance information according to the city governance efficiency, dividing the city governance information into different governance information units, and collecting the current running state data of each governance information unit;

[0009] Step S2: obtaining the task execution data of each governance information unit, constructing a recurrent neural network model, obtaining the resource potential characteristics of each governance information unit, and generating a local strategy coefficient in combination with the current running state data;

[0010] Step S3: combine the to-be-processed event data with the local policy coefficient of all governance information units to perform policy analysis, and determine whether there is an action policy conflict based on the policy analysis result of each governance information unit;

[0011] Step S4: after detecting the action policy conflict, match the event location of the to-be-processed event with the geographical location of the governance information unit, generate an optimal configuration result, and reevaluate the urban governance efficiency.

[0012] In a preferred embodiment, in step S1, the ratio of the number of events successfully scheduled resources in a preset unit time window to the total number of reported events is calculated as the resource scheduling response rate;

[0013] The ratio of the number of events with a status identifier of completed in a preset unit time to the total number of reported events is calculated as the event processing completion rate;

[0014] The resource scheduling response rate and the event processing completion rate are weighted and fused to form a comprehensive efficiency value;

[0015] If the comprehensive efficiency value is less than a preset efficiency evaluation threshold, it is determined that the current urban governance efficiency is low, otherwise it is determined that the current urban governance efficiency is normal.

[0016] In a preferred embodiment, in step S1, when it is determined that the current urban governance efficiency is low, the event number, event type and event processing progress of the to-be-processed event are retrieved to construct a two-dimensional data matrix as the to-be-processed event data;

[0017] Collect the structured data of all governance resources to obtain urban governance information;

[0018] According to the business type of the urban governance information, it is divided into different governance information units;

[0019] For each governance information unit, the number of remaining resources currently available for task allocation is collected to obtain the amount of schedulable resources.

[0020] In a preferred embodiment, in step S2, the total amount of various resources allocated by the governance information unit in a preset historical time interval is counted to obtain the amount of scheduled resources;

[0021] The ratio of the number of unique valid events processed by the governance information unit in the historical time interval to the total number of events received is counted to obtain the non-repeated reporting rate;

[0022] The scheduled resource time series and the non-repeated reporting time series are constructed and combined into a time series feature vector.

[0023] In a preferred implementation, in step S2, the time series feature vector is input into a recurrent neural network model to obtain a hidden state through recursive calculation;

[0024] The hidden state is mapped to a resource potential feature;

[0025] The resource potential feature is fused with the schedulable resource amount to generate a local policy coefficient.

[0026] In a preferred implementation, in step S3, the set of disposable task types in each governance information unit is matched with the event type of each to-be-processed event in the to-be-processed event data;

[0027] It is determined whether the set of disposable task types of each governance information unit contains the event type;

[0028] If the event type is contained in the set of disposable task types of the governance information unit, the scheduling feasibility value of the to-be-processed event corresponding to the governance information unit is set to 1;

[0029] Otherwise, the scheduling feasibility value is set to 0.

[0030] In a preferred implementation, in step S3, based on the scheduling feasibility value, the processing progress of the to-be-processed event is fused with the local policy coefficient of the governance information unit to obtain a policy adaptation score of the governance information unit for each to-be-processed event;

[0031] The expression of the policy adaptation score is: wherein A is the scheduling feasibility value of the governance information unit, is the processing progress of each to-be-processed event, is the local policy coefficient of the governance information unit, and is the policy adaptation score.

[0032] Among the to-be-processed events, the governance information unit with the highest policy adaptation score is taken as the first governance information unit corresponding to the to-be-processed event;

[0033] If the number of times that the governance information unit appears as the first governance information unit is greater than a preset number threshold, it is determined that there is an action policy conflict, and the corresponding governance information unit is marked.

[0034] In a preferred implementation, in step S4, after detecting that there is an action policy conflict, the event position coordinates of the to-be-processed event and the governance information unit are obtained;

[0035] The position distance of the to-be-processed event corresponding to the marked governance information is calculated based on the position coordinates;

[0036] In the label governance information unit, the position distance between all associated pending events is calculated respectively;

[0037] The pending events associated with the label governance information unit are sorted in ascending order according to the position distance to construct a sorted pending set.

[0038] In a preferred embodiment, in step S4, a preset number of pending events based on a threshold number are selected from small to large in the pending set and assigned to the label governance information unit in step S4;

[0039] The strategy adaptation score of each strategy adaptation score in the pending set that has not been assigned is normalized with the corresponding position distance, the processed strategy adaptation score and the position distance are subtracted to obtain the logistic regression parameter, and the comprehensive assignment score is obtained based on the logistic regression parameter through the Sigmoid function;

[0040] The maximum value of the comprehensive assignment score of the unassigned pending event corresponds to the first processing information unit, and is assigned to the corresponding processing information unit;

[0041] After all the pending events are assigned, the city governance efficiency is re-evaluated.

[0042] A city governance dynamic analysis system based on deep learning includes an efficiency evaluation module, a potential modeling module, a strategy analysis module and a conflict decision module, and the functions of each module are as follows:

[0043] Efficiency evaluation module: detect the reporting event processing state to evaluate the city governance efficiency, determine whether to call the pending event data and city governance information according to the city governance efficiency, divide the city governance information into different governance information units, and collect the current running state data of each governance information unit;

[0044] Potential modeling module: obtain the task execution data of each governance information unit, construct a recurrent neural network model, obtain the resource potential characteristics of each governance information unit, and generate a local strategy coefficient in combination with the current running state data;

[0045] Strategy analysis module: combine the pending event data with the local strategy coefficients of all governance information units for strategy analysis, and determine whether there is an action strategy conflict based on the strategy analysis results of each governance information unit;

[0046] Conflict decision module: after detecting the action strategy conflict, match the event position of the pending event with the geographical position of the governance information unit, generate an optimal configuration result, and re-evaluate the city governance efficiency.

[0047] The technical effects and advantages of the present application are as follows:

[0048] The application judges the current urban governance efficiency state by collecting the event reporting and resource scheduling situation of the urban governance platform, calculating the resource scheduling response rate and event processing completion rate, and weighting and fusing to form a comprehensive efficiency value, when the governance efficiency is lower than the preset threshold, the to-be-processed event and governance resource data are called, the governance resources are divided into different governance information units, the current running state is obtained, two key indicators of scheduling resource quantity and non-repeated reporting rate are extracted from historical governance data, and a feature vector is constructed in the form of time series, a recurrent neural network model is trained to extract resource potential characteristics of each governance information unit, the potential characteristics are fused with the current schedulable resource quantity to generate a local strategy coefficient reflecting the resource allocation capability, based on the local strategy coefficient, strategy matching analysis of the to-be-processed event and governance resources is performed, whether there is a strategy conflict is judged, and in the case of conflict, position matching optimization resource allocation is performed, finally the re-evaluation and dynamic optimization of urban governance efficiency are realized, by fusing historical data mining and real-time resource scheduling state, the resource allocation potential of each governance unit can be accurately described, and the scientificity of urban governance response and the rationality of resource matching are improved. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 The method flowchart of the application is a dynamic analysis method of urban governance based on deep learning.

[0050] Figure 2 The module schematic diagram of the application is a dynamic analysis system of urban governance based on deep learning. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.

[0052] The application judges the current urban governance efficiency state by collecting the event reporting and resource scheduling situation of the urban governance platform, calculating the resource scheduling response rate and the event processing completion rate, and weighting and fusing to form a comprehensive efficiency value. When the governance efficiency is lower than the preset threshold, the to-be-processed event and governance resource data are called, the governance resources are divided into different governance information units, the current running state thereof is obtained, two key indicators of scheduling resource quantity and non-repeated reporting rate are extracted from historical governance data, a feature vector is constructed in the form of time series, a recurrent neural network model is trained to extract the resource potential characteristics of each governance information unit, a local strategy coefficient reflecting the resource allocation capability is generated by fusing the potential characteristics and the current schedulable resource quantity, and the strategy matching analysis of the to-be-processed event and the governance resource is performed based on the local strategy coefficient to judge whether there is a strategy conflict, and the position matching optimization resource allocation is performed in the conflict case, so as to finally realize the reevaluation and dynamic optimization of the urban governance efficiency.

[0053] Embodiment 1, a deep learning-based urban governance dynamic analysis method, as shown in Figure 1 The method comprises the following steps:

[0054] Step S1: detecting the reporting event processing state to evaluate the urban governance efficiency, judging whether to call the to-be-processed event data and the urban governance information according to the urban governance efficiency, dividing the urban governance information into different governance information units, and collecting the current running state data of each governance information unit;

[0055] Step S2: obtaining the task execution data of each governance information unit, constructing a recurrent neural network model, obtaining the resource potential characteristics of each governance information unit, and generating a local strategy coefficient in combination with the current running state data;

[0056] Step S3: performing strategy analysis on the to-be-processed event data in combination with the local strategy coefficients of all governance information units, and determining whether there is an action strategy conflict based on the strategy analysis results of each governance information unit;

[0057] Step S4: after detecting the action strategy conflict, matching the event position of the to-be-processed event and the geographical position of the governance information unit, generating an optimal configuration result, and reevaluating the urban governance efficiency.

[0058] The specific implementation is as follows:

[0059] In step S1, the urban governance platform is accessed, the resource scheduling response rate and the event processing completion rate are calculated, and the reporting event processing state is detected, wherein the resource scheduling response rate is defined as the ratio of the number of successfully scheduled resources to the total number of reported events in a unit time window, and the event processing completion rate is defined as the ratio of the number of events with a state identifier of completed to the total number of reported events in a unit time window;

[0060] It should be noted that the urban governance platform refers to an information system platform for urban management affairs, which has functions of event reporting, data recording, state synchronization, task dispatching, result feedback for the public and management departments, and can realize digital recording and management of urban governance events; the unit time window is a preset custom sliding time window, which is configured based on the response cycle and event density of urban governance, and will not be described here.

[0061] The resource scheduling response rate and the event processing completion rate are weighted and fused to form a comprehensive efficiency value for evaluating the efficiency of urban governance, which is calculated as follows:

[0062]

[0063] Among them, is the comprehensive efficiency value, is the resource scheduling response rate, is the event processing completion rate, and is a preset weight coefficient, satisfying .

[0064] The comprehensive efficiency value is compared with the preset efficiency evaluation threshold , if , it is determined that the current urban governance efficiency is low, otherwise it is determined that the current urban governance efficiency is normal, when it is determined that the current urban governance efficiency is low, the to-be-processed event data and urban governance information are retrieved.

[0065] The events in the urban governance platform that have not been completed are structured and expressed to form to-be-processed event data for subsequent strategy analysis. The above to-be-processed event data is a two-dimensional data matrix composed of event number, event type and event processing progress retrieved for events that have not been completed.

[0066] The event number refers to a unique identifier assigned by the urban governance platform to each reported event, denoted as , which is used for indexing, searching, tracking and association processing of events.

[0067] The event type refers to the classification attribute of the event in the urban governance scene, which is used to identify the business nature and attribution field of the event. Each type of event corresponds to a unique type identifier, denoted as , where i represents the event index.

[0068] The event processing progress refers to the processing state of the event, which is expressed as a percentage to represent the overall processing completion degree of the event, denoted as , which is used to measure the current processing state of the event.

[0069] ​It should be noted that the percentage of event processing progress setting rules are set by the city governance platform according to the event type and governance process stage. Different event types can correspond to different processing stage division and progress weight configuration. For example, the event processing progress is set to 25% after completing task dispatching, and the event processing progress is set to 100% after completing disposal feedback record uploading.

[0070] All event data that have not been completed processing are collected from the city governance platform, and the corresponding event number, event type and event processing progress are extracted. They are constructed into a structured two-dimensional data matrix according to the event, which is represented as:

[0071] ;

[0072] Among them, is a two-dimensional data matrix, is the number of the ith event; is the event type; is the event processing progress. The two-dimensional data matrix satisfies the condition , where n is the total number of events to be processed, and each row represents the processing data record of an event that has not been completed.

[0073] The city governance information, i.e. the structured data of all governance resources registered in the city governance platform, is called. According to the business type of the city governance information, it is divided into different governance information units, and each governance information unit corresponds to an independent functional department.

[0074] For each governance information unit, its current running state data is collected, which is the schedulable resource amount of the governance information unit, i.e. the remaining resource quantity of the governance information unit that can be used for task allocation at present.

[0075] In step S2, task execution data formed by each governance information unit in the process of processing historical governance events is collected from the city governance platform as training samples for building a deep learning model, so as to realize modeling and analysis of resource potential characteristics of each governance information unit.

[0076] The task execution data includes scheduling resource amount and non-repeated reporting rate. The scheduling resource amount is obtained by counting the scheduling records of various resources of each governance information unit. The scheduling resource amount represents the total amount of various resources allocated by each governance information unit within a preset historical time interval. The non-repeated reporting rate is used to reflect the effectiveness of the governance information unit in processing events, which is defined as the ratio of the number of unique effective events processed by the governance information unit to the total number of events received by the governance information unit within the historical time interval.

[0077] The scheduling resource quantity and the non-repeated reporting rate are continuously collected in time series form within a preset historical time interval, and a scheduling resource time series and a non-repeated reporting time series are constructed, respectively.

[0078] A time series feature vector is constructed based on the scheduling resource time series and the non-repeated reporting time series, as a training sample for constructing a deep learning model, and the deep learning model adopts a recurrent neural network model.

[0079] The input layer of the recurrent neural network model receives the time series feature vector, updates the hidden state through recursive calculation, and the hidden state calculation process adopts the following formula:

[0080] ;

[0081] wherein, is the hidden state at the tth sampling time, is the time series feature vector, , are an input weight matrix and a hidden state weight matrix, respectively, and initial values are generated by random initialization, is a bias vector, is a nonlinear activation function, the hyperbolic tangent function is used as the activation function in the embodiment, j is an index number of the governance information unit, and t is an index number of the sampling time in the time series.

[0082] After recursive processing of the sampling time t = 1 to t = T, the output hidden state is obtained, and the resource potential feature is mapped through the output layer weight matrix and the bias term, and the calculation formula is:

[0083] ;

[0084] wherein, is the resource potential feature, is the hidden state at the Tth sampling time, is the output layer weight matrix, is the output layer bias term, represents a Sigmoid activation function, which limits the output to the interval [0, 1], and T is the last sampling time in the time series.

[0085] The resource potential feature is fused and calculated with the schedulable resource quantity to generate a local strategy coefficient, in combination with the current running state data of each governance information unit, i.e., the schedulable resource quantity, and the fusion is performed in a weighted linear manner, and the formula is represented as:

[0086] ;

[0087] wherein, is a local policy coefficient, is a normalized value of the schedulable resource quantity, and a maximum-minimum normalization method is adopted for normalization, so as to ensure that the numerical range is unified to , is a weight coefficient, and the value range is [0, 1], which is used to adjust the influence proportion of the resource potential feature and the current state on the local policy coefficient.

[0088] It should be noted that the recurrent neural network model is a kind of artificial neural network suitable for processing sequence data, and the basic feature is that there is a time recursive connection relationship between states in the network structure; the hidden state refers to the encoding result of the recurrent neural network model on the historical input information at each time, and its role is to carry the memory information of the previous time to participate in the feature calculation at the current time; the activation function is a nonlinear mapping function in the neural network, which acts on the intermediate result after linear transformation, and is used to enhance the nonlinear expression ability of the model, which will not be described here.

[0089] In step S3, the set of dispoable task types in each governance information unit is called through the governance resource capability database;

[0090] The event type of each to-be-processed event in the to-be-processed event data is matched with the set of dispoable task types of each governance information unit, and it is judged whether the set of dispoable task types of each governance information unit contains the event type;

[0091] If the set of dispoable task types of the governance information unit contains the event type of the corresponding event, the scheduling feasibility value of the to-be-processed event corresponding to the governance information unit is set to 1, otherwise, the scheduling feasibility value is set to 0;

[0092] Based on the scheduling feasibility value, the processing progress of the to-be-processed event and the local policy coefficient of the governance information unit are fused and analyzed to obtain the strategy adaptation score of the governance information unit to each to-be-processed event, and the expression of the strategy adaptation score is: wherein, A is the scheduling feasibility value of the governance information unit, is the processing progress of each to-be-processed event, is the local policy coefficient of the governance information unit, is the strategy adaptation score;

[0093] Among the to-be-processed events, the governance information unit with the highest strategy adaptation score is taken as the first governance information unit corresponding to the to-be-processed event;

[0094] The number of times that each governance information unit appears as the first governance information unit is counted, and if the number of times that the governance information unit appears as the first governance information unit is greater than a preset number threshold, it is judged that there is an action strategy conflict, and the corresponding governance information unit is marked.

[0095] By constructing the policy adaptation score and determining the action policy conflict, the scheduling relationship between the to-be-processed event and the governance information unit can be accurately matched and dynamically balanced, the overload or concentration of governance resources in a local unit can be avoided, the rationality of task allocation and the resource utilization efficiency of the urban governance scheduling system can be effectively improved, and a decision basis for subsequent conflict mitigation and optimal allocation is provided.

[0096] It should be noted that the governance resource capability database records the task type support capability information of each governance information unit in the city, including task type identifier, processing history frequency, etc., for supporting task adaptation and capability matching analysis; the preset number threshold is preset by a professional, the maximum number of event assignments that the governance information unit can undertake, to prevent the governance information unit from undertaking too many tasks in a short time due to a high policy adaptation score.

[0097] In step S4, after detecting the existence of the action policy conflict, the event location coordinates of the to-be-processed event and the governance information unit are obtained through the geographic information system;

[0098] The distance of the marked governance information corresponding to the to-be-processed event is calculated: wherein, and are the coordinates of the marked governance information, and are the coordinates of the to-be-processed event, and d is the location distance;

[0099] In the marked governance information unit, the location distances between all associated to-be-processed events are calculated respectively;

[0100] The to-be-processed events associated with the marked governance information unit are sorted in ascending order according to the location distance, and a sorted to-be-processed set is constructed;

[0101] In the to-be-processed set, the to-be-processed events based on the preset number threshold are selected from small to large, and are allocated to the marked governance information unit;

[0102] The to-be-processed events in the to-be-processed set that are not allocated are calculated based on the location distance and the policy adaptation score to obtain a comprehensive allocation score. The policy adaptation scores in the allocated to-be-processed events and the corresponding location distances are normalized, the processed policy adaptation scores and the location distances are subtracted to obtain a logistic regression parameter, and the comprehensive allocation score is obtained based on the logistic regression parameter through a Sigmoid function: wherein, z is the logistic regression parameter, is the comprehensive allocation score;

[0103] The maximum value of the comprehensive allocation score of the unassigned to-be-processed event is taken as the first processing information unit, and is allocated to the corresponding processing information unit;

[0104] After all to-be-processed events are allocated, the city governance efficiency is re-evaluated to ensure that the subsequent scheduling dynamically adapts to the changes in resource state.

[0105] After detecting the action strategy conflict, the comprehensive optimization configuration is performed based on the spatial distance, which effectively alleviates the scheduling conflict problem caused by uneven or excessive concentration of governance resources, realizes the dynamic balanced allocation of city governance resources and the collaborative response of multi-event tasks, and further improves the scheduling stability and city governance efficiency of the system.

[0106] It should be noted that the geographic information system is a computer system for collecting, storing, managing and displaying geographic spatial data, and the geographic location information of to-be-processed events and governance information units is uniformly collected and managed; the Sigmoid function is a nonlinear function in a machine learning model, which is used to map any real number value to the interval [0, 1], and in this example, the comprehensive allocation score is mapped to the interval [0, 1] for sorting.

[0107] Embodiment 2, a city governance dynamic analysis system based on deep learning, as shown in Figure 2 The city governance dynamic analysis method based on deep learning, as shown in

[0108] Efficiency evaluation module: detect the reported event processing state to evaluate the city governance efficiency, determine whether to retrieve to-be-processed event data and city governance information according to the city governance efficiency, divide the city governance information into different governance information units, and collect the current running state data of each governance information unit;

[0109] Potential modeling module: obtain the task execution data of each governance information unit, construct a recurrent neural network model, obtain the resource potential characteristics of each governance information unit, and generate a local strategy coefficient in combination with the current running state data;

[0110] Strategy analysis module: combine the to-be-processed event data with the local strategy coefficients of all governance information units for strategy analysis, and determine whether there is an action strategy conflict based on the strategy analysis results of each governance information unit;

[0111] Conflict decision module: after detecting the action strategy conflict, match the event location of the to-be-processed event with the geographic location of the governance information unit, generate an optimal configuration result, and re-evaluate the city governance efficiency.

[0112] The above-described embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented by software, the above-described embodiments can be implemented in whole or in part in the form of a computer program product.

[0113] Those skilled in the art can realize that the modules and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and the constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0114] In addition, each functional module in each embodiment of the present application can be integrated in one processing module, or each module can exist physically alone, or two or more modules can be integrated in one module.

[0115] The above is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0116] Finally, the above is merely preferred embodiments of the present application, and is not intended to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A deep learning-based dynamic analysis method for urban governance, characterized in that: Includes the following steps: Step S1: Detect the status of reported event processing to assess the efficiency of urban governance. Based on the efficiency of urban governance, determine whether to retrieve the data of pending events and urban governance information. Divide the urban governance information into different governance information units and collect the current operating status data of each governance information unit. Step S2: Obtain task execution data of each governance information unit, construct a recurrent neural network model, obtain the resource potential characteristics of each governance information unit, and generate local strategy coefficients by combining the current operating status data; Step S3: Combine the event data to be processed with the local policy coefficients of all governance information units to perform policy analysis, and determine whether there is a conflict between action policies based on the policy analysis results of each governance information unit. In step S3, the set of manageable task types in each governance information unit is retrieved and matched with the event type of each event to be processed in the event data to be processed; Determine whether the set of manageable task types for each governance information unit includes event types; If the set of manageable task types of a governance information unit includes event types, then the scheduling feasibility value of the governance information unit corresponding to the event to be processed is set to 1. Conversely, set the scheduling feasibility value to 0; In step S3, based on the scheduling feasibility value, the processing progress of the events to be processed is fused and analyzed with the local policy coefficients of the governance information unit to obtain the policy adaptation score of the governance information unit for each event to be processed. The expression for the strategy fit score is: Where A is the scheduling feasibility value of the governance information unit. For the processing progress of each pending event, To manage the local strategy coefficients of information units, Score for strategy adaptation; Among all pending events, the governance information unit with the highest policy adaptation score is selected as the first governance information unit for the corresponding pending event. If the number of times a governance information unit appears as the first governance information unit exceeds a preset threshold, it is determined that there is an action strategy conflict, and the corresponding governance information unit is marked. Step S4: After detecting action policy conflict, match the event location of the event to be processed with the geographical location of the governance information unit to generate the optimal configuration result and re-evaluate the efficiency of urban governance.

2. The method for dynamic analysis of urban governance based on deep learning according to claim 1, characterized in that: In step S1, the ratio of the number of events in which resources are successfully scheduled within a preset unit time window to the total number of reported events is calculated as the resource scheduling response rate. The ratio of the number of events marked as completed within a preset unit time period to the total number of reported events is used as the event processing completion rate. The resource scheduling response rate and the event processing completion rate are weighted and combined to form a comprehensive efficiency value; If the overall efficiency value is less than the preset efficiency assessment threshold, the current urban governance efficiency is determined to be low; otherwise, the current urban governance efficiency is determined to be normal.

3. The method for dynamic analysis of urban governance based on deep learning according to claim 2, characterized in that: In step S1, when it is determined that the current urban governance efficiency is low, the event number, event type and event processing progress of the event to be processed are retrieved, and a two-dimensional data matrix is ​​constructed as the event data to be processed. Collect structured data from all governance resources to obtain urban governance information; Based on the business type of urban governance information, it is divided into different governance information units; For each governance information unit, the remaining amount of resources currently available for task allocation is collected to obtain the amount of schedulable resources.

4. The urban governance dynamic analysis method based on deep learning according to claim 1, characterized in that: In step S2, the amount of resources to be scheduled is obtained by statistically analyzing the total amount of various resources allocated by the governance information unit within a preset historical time interval. The non-duplication reporting rate is obtained by calculating the ratio of the number of unique and valid events processed by the governance information unit within a historical time interval to the total number of events received. Construct time series of scheduling resources and time series of non-repeating reports, and combine them into a time series feature vector.

5. The urban governance dynamic analysis method based on deep learning according to claim 4, characterized in that: In step S2, the time series feature vector is input into the recurrent neural network model, and the hidden state is obtained through recursive calculation; Map the hidden state to resource potential features; The resource potential characteristics are combined with the amount of schedulable resources to generate local strategy coefficients.

6. The method for dynamic analysis of urban governance based on deep learning according to claim 1, characterized in that: In step S4, after detecting a conflict in action strategies, the event location coordinates of the event to be processed and the governance information unit are obtained; Calculate the location distance of the event to be processed corresponding to the marked governance information based on the location coordinates; In the tagging governance information unit, the location distance between all associated pending events is calculated respectively; The pending events associated with the marked governance information units are sorted in ascending order based on their location distance to construct a sorted set of pending events.

7. The method for dynamic analysis of urban governance based on deep learning according to claim 6, characterized in that: In step S4, events to be processed are selected from the set to be processed in ascending order based on a preset number of times threshold, and then assigned to the tagging and governance information unit. The strategy adaptation scores and corresponding location distances of each unassigned event in the unassigned event set are normalized. The difference between the processed strategy adaptation scores and location distances is used to obtain logistic regression parameters. The comprehensive assignment score is obtained based on the logistic regression parameters through the Sigmoid function. The processing information unit corresponding to the maximum value of the comprehensive allocation score of the unassigned pending events is taken as the first processing information unit and assigned to the corresponding processing information unit. After all pending events have been assigned, the efficiency of urban governance will be reassessed.

8. A deep learning-based urban governance dynamic analysis system, used to implement the deep learning-based urban governance dynamic analysis method according to any one of claims 1-7, characterized in that: It includes an efficiency assessment module, a potential modeling module, a strategy analysis module, and a conflict decision-making module. The functions of each module are as follows: Efficiency Assessment Module: Detects the status of reported event processing to assess urban governance efficiency. Based on urban governance efficiency, it determines whether to retrieve pending event data and urban governance information, divides urban governance information into different governance information units, and collects the current operating status data of each governance information unit. Potential Modeling Module: Acquires task execution data from each governance information unit, constructs a recurrent neural network model, obtains the resource potential characteristics of each governance information unit, and generates local policy coefficients by combining current operating status data; Strategy Analysis Module: Combines the event data to be processed with the local strategy coefficients of all governance information units to perform strategy analysis, and determines whether there are action strategy conflicts based on the strategy analysis results of each governance information unit; Conflict Decision Module: After detecting a conflict in action strategies, it matches the event location of the event to be processed with the geographical location of the governance information unit, generates the optimal configuration result, and re-evaluates the efficiency of urban governance.

Images