A method and system for dispatching urban management events

Abstract

The present invention relates to the field of urban management technology, and discloses a method and system for dispatching and transferring urban management events. The method comprises: utilizing natural language processing technology to analyze urban management event description text, extracting urban management event semantic features, and constructing an urban management event semantic model library and a processing flow fragment library; constructing a processing intent recognition framework, extracting processing intent features based on historical successful cases, evaluating the matching degree between automatically generated processing flows and processing intents, and detecting and correcting potential defects in the processing flows; collecting feedback and evaluation data from processing personnel and parties involved in urban management events, constructing a processing flow quality evaluation system, and applying a differential evolution algorithm to discover the optimal processing flow model. The present invention achieves intelligent dispatching and efficient transfer of urban management events by integrating urban management event semantic modeling, automated assembly and verification of processing flows based on processing intent understanding, and adaptive optimization technology based on crowdsourcing feedback.

Description

Technical Field

[0001] The present invention relates to the technical field of urban management, and more specifically, to a method and system for dispatching and transferring urban management events. Background Art

[0002] Urban management event processing has high requirements for timeliness and accuracy, but traditional urban management event flow systems generally use predefined processing process templates, which have the following problems: lack of adaptability to complex and changing urban management scenarios, and difficulty in quickly responding to new urban management event types and processing requirements; adaptive processing process orchestration lacks an in-depth understanding of processing intent, resulting in the generated processing process being technically feasible but may not meet actual business processing needs; the automatically generated processing process lacks a feedback optimization mechanism for the actual execution effect, and it is difficult to continuously improve based on the processing results, resulting in similar problems recurring.

[0003] In existing technologies, most urban management event processing systems utilize a fixed library of process templates, selecting the appropriate template through simple rule matching. This approach works well for routine, standardized urban management events, but when faced with complex, unconventional urban management events, it often fails to generate an appropriate process, requiring manual intervention and adjustment, impacting processing efficiency and accuracy.

[0004] Some systems attempt to employ rule-based adaptive process generation techniques. However, these approaches rely on predefined rule bases, making them incapable of adapting to new types of urban management events and unable to self-optimize based on actual processing results. Other systems employ machine learning to recommend process flows, but due to a lack of in-depth understanding of the semantics and processing intent of urban management events, the recommended processes often fail to meet actual business needs and fail to establish a closed-loop feedback optimization mechanism.

[0005] Therefore, there is an urgent need for an urban management event scheduling and flow technology that can integrate urban management event semantic understanding, processing intention recognition and feedback optimization to achieve intelligent generation and self-evolution of urban management event processing processes. Summary of the Invention

[0006] The present invention provides a method and system for dispatching and transferring urban management events, which solve the technical problems in related technologies such as the rigidification of urban management event processing processes, poor adaptability, lack of understanding of processing intentions and feedback optimization mechanisms.

[0007] The present invention provides a method for dispatching and transferring urban management events, comprising:

[0008] Use natural language processing technology to analyze urban management event description texts, extract urban management event semantic features, and build an urban management event semantic model library and processing flow fragment library;

[0009] Build a processing intent recognition framework, extract processing intent features based on historical success cases, evaluate the match between automatically generated processing procedures and processing intent, and detect and correct potential defects in the processing procedures;

[0010] Collect feedback and evaluation data from processing personnel and stakeholders of urban management events, build a processing process quality evaluation system, apply the differential evolution algorithm to discover the optimal processing process model, and achieve continuous optimization and iterative evolution of the processing process.

[0011] In a preferred embodiment, the process of extracting semantic features of urban management events includes: constructing a domain vocabulary :

[0012] ;

[0013] in, Indicates the Field terms, Indicates the total number of terms in the vocabulary; Contains professional terms in the field of urban management; text describing urban management events Perform word segmentation and part-of-speech tagging to obtain word sequences :

[0014] ;

[0015] in, Indicates the first words, Indicates the total number of words after word segmentation;

[0016] Calculate the similarity score between each word and the domain vocabulary ; Attention weight The calculation is as follows:

[0017] ;

[0018] in, 、 Respectively represent the i-th and j-th words after word segmentation; Represents exponential operation; Represents the domain vocabulary; Represents the similarity calculation function; Indicates the total number of words after word segmentation. Expressive words The attention weight of a word reflects the importance of the word in the description of urban management events;

[0019] Generate semantic feature vectors of urban management events :

[0020] ;

[0021] in For words The word vector representation of ; Expressive words The attention weight of Indicates the total number of words after word segmentation.

[0022] In a preferred embodiment, the processing intent recognition framework includes: a multi-layer neural network model, which includes an input layer, an encoding layer, an attention layer, and an output layer, and its forward propagation process is:

[0023] ;

[0024] ;

[0025] ;

[0026] ;

[0027] ;

[0028] in, is the semantic feature vector of urban management events, To process the intent vector, 、 、 denote the first, second and third weight matrices respectively, 、 、 Represent the first, second and third bias vectors respectively, 、 are the outputs of the first and second hidden layers respectively, is the attention weight matrix, is the attention-weighted feature, for Dimensions; for Activation function; is the normalization function.

[0029] In a preferred embodiment, the process of verifying the processing flow based on the processing intent includes: establishing a semantic association matrix between the processing flow segments and the processing intent , the elements in the matrix Represents a processing flow fragment and processing intent categories The correlation degree; calculation process combination:

[0030] ;

[0031] in, Represents a specific processing flow combination; represents the i-th processing flow fragment, Indicates the total number of processing flow fragments;

[0032] and processing intent Overall match score:

[0033] ;

[0034] in Represents a specific processing flow combination; , Processing process fragments The weight of To process intent vectors Middle The probability of class processing intent, To process the total number of intent categories; Indicates the total number of processing flow fragments; Represents the i-th processing flow fragment and processing intent vectors Similarity function of Represents the i-th processing flow fragment in the processing flow combination and processing intent categories 's relevance.

[0035] In a preferred embodiment, the process of detecting and correcting potential defects in a process flow includes: defining a process flow defect pattern library:

[0036] ;

[0037] Contains common process design flaws; builds defect detection functions , identify the conditions that meet Defects ; Apply the repair rule set Generate the repair process plan:

[0038] ;

[0039] in is the combination of processing procedures to be detected, is the detection threshold, is the set of detected defects, represents the i-th detected defect, t represents the total number of detected defects, This is a bug fix function.

[0040] In a preferred embodiment, the execution process of the differential evolution algorithm includes: encoding the processing flow combination into a feature vector to construct an initial population; iteratively performing mutation, crossover and selection operations to update the population:

[0041] Mutations:

[0042] ;

[0043] Crossover: Generate trial vectors based on crossover probability ;

[0044] Selection: Compare the fitness of the test vector and the target vector, and select the better one to enter the next generation; select the individual with the highest quality score from the final population as the optimization process model;

[0045] in, For the Daidi The mutation vector of each individual, 、 、 are three different individuals randomly selected, is the scaling factor, is the test vector; is the number of generations in the current iteration.

[0046] In a preferred embodiment, the process quality evaluation system calculates the overall quality score of the process using the following formula: :

[0047] ;

[0048] in, For indicators The weight of Dynamic adjustment, is the indicator importance parameter, is the total number of evaluation indicators, For processing flow In the indicator The score on .

[0049] In a preferred embodiment, a method for dispatching and transferring urban management events further includes a memory-enhanced processing flow generation process:

[0050] Building an experience memory bank:

[0051] ;

[0052] Stores city management event features, corresponding high-quality processing procedures and quality scores; for new city management event features , retrieve similar historical experiences from the memory bank:

[0053] ;

[0054] The search results are integrated with the basic generation process to form an enhanced generation process:

[0055] ;

[0056] in is the balance parameter; Represents the basic generation process flow; Represents the similar historical processing flow of retrieval; represents the similarity function of urban management events; Represents the processing flow fusion function; Manage event characteristics for historical cities stored in memory banks; For the corresponding historical processing flow; Score the quality of historical processing; A feature vector for new city management events; Indicates obtaining the maximum parameter.

[0057] In a preferred embodiment, a method for dispatching and transferring urban management events further includes real-time monitoring of the execution of the processing flow, and triggering the process of regenerating and verifying the processing flow when an abnormality occurs in the execution of the processing flow.

[0058] In a preferred embodiment, a city management event dispatching and circulation system includes:

[0059] Urban management event semantic analysis module, used to analyze urban management event description text and extract urban management event features and attributes;

[0060] A processing flow fragment library management module is used to maintain and manage the processing flow fragment library;

[0061] Processing intention understanding and verification module, used to understand the processing intention and verify the processing flow plan;

[0062] Crowdsourcing feedback collection module, used to collect and process feedback information from personnel and stakeholders;

[0063] Process optimization and evolution module, used to optimize the quality of the process based on feedback data;

[0064] The processing flow execution and monitoring module is used for the specific execution and status monitoring of the processing flow.

[0065] The beneficial effects of the present invention are:

[0066] High adaptability: Through the semantic modeling of urban management events and the construction of a processing flow fragment library, the system can deeply understand the semantic characteristics of various urban management events and automatically generate customized processing flows for different types of urban management events. This overcomes the limitations of traditional fixed processing flow templates and greatly improves the adaptability and accuracy of urban management event processing.

[0067] Processing intent understanding capability: By introducing a processing intent recognition framework based on a multi-layer neural network, the system can learn processing intent patterns from historical successful cases, understand the real business needs behind urban management events, and ensure that the generated processing flow is not only technically feasible but also meets the actual business processing intent. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 This is a processing flow chart of a method for dispatching and transferring urban management events according to the present invention;

[0069] Figure 2 It is a process flow chart of the technology implementation of the city management event semantic modeling and process fragment library construction of the present invention;

[0070] Figure 3 It is a process flow chart for realizing the technology of automatic assembly and verification of process flow based on process intent understanding of the present invention;

[0071] Figure 4 It is a process flow chart of the technical implementation of adaptive optimization of the processing flow integrating crowdsourcing feedback of the present invention. DETAILED DESCRIPTION

[0072] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that these embodiments are discussed solely to enable those skilled in the art to better understand and implement the subject matter described herein, and that the functions and arrangements of the elements discussed may be varied without departing from the scope of this specification. Various examples may omit, substitute, or add various processes or components as needed. Furthermore, features described in some examples may be combined in other examples.

[0073] At least one embodiment of the present invention discloses a method for dispatching and transferring urban management events, such as Figure 1 As shown, the following steps are included:

[0074] Step 100: Analyze the urban management event description text using natural language processing technology, extract the semantic features of the urban management event, and build an urban management event semantic model library and a processing flow fragment library;

[0075] In one embodiment of the present invention, the urban management event description text is segmented and tagged with parts of speech, the semantic feature vector of the urban management event is extracted through word vector mapping and attention calculation, and the optimal processing flow combination is calculated according to the constraint satisfaction problem model to provide an initial solution for the subsequent processing flow generation.

[0076] Specifically include:

[0077] Step 101: extracting semantic features of urban management events;

[0078] Natural language processing technology is used to analyze the description text of urban management events and extract the core semantic features of urban management events. The implementation process is as follows:

[0079] Build a domain vocabulary:

[0080] ;

[0081] Contains professional terms in the field of urban management, such as garbage dumping, noise nuisance, illegal buildings, etc., totaling about 500 terms; among them, Indicates the Field terms, Indicates the total number of terms in the vocabulary.

[0082] Description text of urban management events Perform word segmentation and part-of-speech tagging to obtain word sequences:

[0083] ;

[0084] in, Indicates the first words, Indicates the total number of words after word segmentation.

[0085] For example, for the description of shared bicycles parked randomly on the sidewalk in front of Building A of the Financial Building, the word segmentation result is [Financial Building, Building A, in front, sidewalk, on, there, shared bicycles, parked randomly];

[0086] Calculate the similarity score of each word with the domain vocabulary:

[0087] ;

[0088] in, is the word vector similarity, indicating the word and domain terminology The semantic similarity between .

[0089] Calculate attention weights:

[0090] ;

[0091] in, 、 Respectively represent the i-th and j-th words after word segmentation; Represents exponential operation; Represents the domain vocabulary; Represents the similarity calculation function; Indicates the total number of words after word segmentation. Expressive words The attention weight of a word reflects the importance of the word in the description of urban management events;

[0092] Generate semantic feature vectors for urban management events:

[0093] ;

[0094] in, For words The word vector representation of Expressive words The attention weight, This is the final generated semantic feature vector of urban management events.

[0095] Step 102: constructing a semantic model library for urban management events;

[0096] A semantic model library is built based on the characteristics of urban management events for urban management event classification and similarity calculation. The implementation process is as follows:

[0097] Define a set of city management event categories :

[0098] ;

[0099] in, Indicates the Urban management event categories, Indicates the total number of urban management event categories; Includes major urban management event categories such as municipal facilities, environmental sanitation, and noise nuisance.

[0100] For each category , define its characteristic attribute set :

[0101] ;

[0102] in, Representation category No. characteristic attributes, Indicates the total number of characteristic attributes of the category; for example, the attributes of urban management events such as noise nuisance include noise source type, duration, and impact range.

[0103] Constructing attribute association rule sets , describing the constraints and associations between attributes.

[0104] Combining urban management event categories, feature attribute sets and association rules to form an urban management event semantic model :

[0105] ;

[0106] in, Representation category A complete semantic model of

[0107] Constructing a semantic model library for urban management events :

[0108] ;

[0109] Contains the semantic model of all city management event categories; among them, Represents the semantic model of the i-th urban management event category; The total number of semantic models representing urban management event categories.

[0110] Step 103, constructing a processing flow fragment library;

[0111] Establish a processing flow fragment library to store various operation units and processing modules. The implementation process is as follows:

[0112] Collect the basic operation units for urban management event processing and construct a set of processing flow fragments:

[0113] ;

[0114] It includes on-site investigation, information recording, problem classification, responsibility identification, processing and execution, etc. Indicates the Process flow fragments, Indicates the total number of processing flow fragments.

[0115] For each process segment , define its functional property set:

[0116] ;

[0117] Including input requirements, execution conditions, output results, etc.; among them, Represents a processing flow fragment No. Functional attributes, Indicates the total number of feature attributes.

[0118] By analyzing historical cases, a mapping matrix between process segments and urban management event categories is established. , where the elements Represents a processing flow fragment Urban management event categories The applicability is calculated as follows:

[0119] ;

[0120] in, Is an indicator function, indicating a processing flow fragment In the category Historical Cases It takes the value 1 when it is successfully applied, otherwise it takes the value 0; For category Total number of historical cases; Represents the total number of all historical cases.

[0121] Step 104, generating a processing flow combination based on constraint satisfaction;

[0122] Based on the characteristics and constraints of urban management events, the optimal initial processing flow combination is generated. The implementation process is as follows:

[0123] Define the set of processing flow combination constraints:

[0124] ;

[0125] Including preconditions, resource constraints, timing relationships, etc.; among them, Indicates the Constraints, Indicates the total number of constraints.

[0126] For a given city management event characteristics and categories , from the set of processing flow fragments Filter out the applicable subset ,satisfy:

[0127] ;

[0128] in, is the fitness threshold, which indicates the minimum fitness requirement for a process flow fragment to be selected into the subset; Represents a processing flow fragment Urban management event categories applicability.

[0129] The process combination problem is formalized as a constraint satisfaction problem (CSP), and the set of possible process combinations that satisfy all constraints is calculated. .

[0130] Design process fitness scoring function , evaluate the process combination Characteristics of urban management events The degree of adaptation. Represents a specific processing flow combination.

[0131] Select the optimal process combination:

[0132] ;

[0133] in, It represents the optimal process combination, which is the one with the highest fitness score among all process combinations that meet the constraints.

[0134] Output the optimal processing flow combination , as the initial processing flow plan.

[0135] Step 200: Build a processing intent recognition framework, extract processing intent features based on historical success cases, evaluate the matching degree between the automatically generated processing flow and the processing intent, and detect and correct potential defects in the processing flow;

[0136] Specifically include:

[0137] Step 201, processing intent recognition and understanding;

[0138] Build a processing intent recognition framework to understand the real business needs of urban management event processing. The implementation process is as follows:

[0139] Collect and annotate a large amount of historical processing records and expert-annotated processing intent datasets:

[0140] ;

[0141] in For urban management event characteristics, To correspond to the processing intention, Indicates the total number of samples in the dataset.

[0142] Design a multi-layer neural network model to handle intent recognition , the model structure is as follows:

[0143] Input layer: Receives the feature vector of urban management events , the dimension is ,in Represents the dimension of the feature vector of urban management events;

[0144] Encoding layer: consists of two fully connected layers. The first layer contains 128 neurons and uses the ReLU activation function. The second layer contains 64 neurons and also uses the ReLU activation function.

[0145] Attention layer: uses the self-attention mechanism to calculate the correlation between different dimensions of urban management event features and strengthen the representation of key features;

[0146] Output layer: contains neurons ( is the number of processing intent categories), and the softmax activation function is used to output the probability distribution of each processing intent category.

[0147] The forward propagation process of the model is:

[0148] ;

[0149] ;

[0150] ;

[0151] ;

[0152] ;

[0153] in, is the semantic feature vector of urban management events, To process the intent vector, 、 、 denote the first, second and third weight matrices respectively, 、 、 Represent the first, second and third bias vectors respectively, 、 are the outputs of the first and second hidden layers respectively, is the attention weight matrix, is the attention-weighted feature, for Dimensions; for Activation function; is the normalization function.

[0154] The model is trained on the urban management event dataset using the cross entropy loss function and the Adam optimizer, with the learning rate set to 0.001, the batch size to 64, and the number of training rounds to 100.

[0155] In application scenarios, such as urban management events such as square dance noise disturbing residents, the model can accurately identify that the main processing intention is noise control and resident mediation, rather than simple violation penalties, thereby guiding the subsequent processing process to be more in line with actual needs. For urban management events such as construction site dust pollution, the model can identify the composite processing intention of environmental protection and construction management, and accordingly generate a processing process that includes both environmental monitoring and construction specifications. This sub-step outputs the processing intention vector of the urban management event. , providing a reference for processing intent for process verification.

[0156] Step 202, processing flow verification based on processing intent;

[0157] Construct a process verification algorithm based on process intent to evaluate the matching degree between the automatically generated process and the process intent. First, establish a semantic association matrix between process fragments and process intent. , where the elements Represents a processing flow fragment and processing intent categories Then calculate the processing flow combination and processing intent Overall match score:

[0158] ;

[0159] in, Processing process fragments The weight of Represents the i-th processing flow fragment and processing intent vectors The similarity function is calculated as follows:

[0160] ;

[0161] in, To process intent vectors Middle The probability of class processing intent, Represents the i-th processing flow fragment in the processing flow combination and processing intent categories This sub-step outputs the process verification score , used to evaluate the compliance of processing process combinations with processing intentions.

[0162] Step 203, processing intent processing flow mapping optimization;

[0163] Based on the verification results, build a processing intent process mapping library and optimize the processing process combination. First, build a processing intent process mapping library:

[0164] ;

[0165] Record processing intent type , corresponding high-quality processing flow combination and validation scores ; Then for the new processing intention , find the most similar processing intent pattern in the mapping library :

[0166] ;

[0167] in, is the cosine similarity function, defined as , and are two vectors; Represents the L2 norm; finally, by adjusting the initial processing flow combination and high-quality processing pipelines in the mapping library , generate optimized processing flow combination :

[0168] ;

[0169] in, is the balance factor, and its value range is ,according to and Dynamic adjustment of similarity; Represents the initially generated processing flow combination; Represents the high-quality processing flow corresponding to similar processing intents retrieved from the mapping library. This sub-step outputs the optimized processing flow combination , better meet the processing intent requirements.

[0170] Step 204, detecting and correcting potential defects in the processing flow;

[0171] To achieve automatic detection and correction of potential defects in the processing process, first define the processing process defect pattern library:

[0172] ;

[0173] Contains common process design flaws, including Indicates the defect mode, Represents the total number of defect modes; then construct a defect detection function , the optimized processing flow combination To scan: ;

[0174] in, Represents a processing flow combination Contains defects The probability of is the detection threshold, which indicates the minimum probability requirement for determining it as a defect; Represents the set of all detected defects; finally, the repair rule set is applied to the detected defects , generate the repaired processing flow plan :

[0175] ;

[0176] in, For the defect repair function, the processing flow combination is adjusted according to the repair rules; Indicates the Repair rules, Indicates the total number of repair rules. This sub-step outputs the repaired processing flow plan , eliminating potential execution risks.

[0177] Step 300, adaptive optimization of the processing flow integrating crowdsourcing feedback;

[0178] Specifically include:

[0179] Step 301: crowdsourcing feedback data collection and preprocessing;

[0180] Build a crowdsourced feedback-based process evaluation framework to collect feedback from event handlers and stakeholders. First, design a multi-dimensional feedback collection interface to collect evaluation data from event handlers, event reporters, and stakeholders:

[0181] ;

[0182] in Rate your satisfaction. For text evaluation, For execution efficiency data, is the timestamp; then evaluate the text Perform sentiment analysis and topic extraction to obtain structured feedback features :

[0183] ;

[0184] in, It is a sentiment analysis function that outputs the sentiment polarity of text evaluation; The topic extraction function extracts the main topics in the text evaluation; finally, the feedback data is compared with the corresponding urban management event features. and processing plan Associate to form the evaluation dataset:

[0185] ;

[0186] in, Indicates the total number of samples in the evaluation dataset. This sub-step outputs a structured evaluation dataset. , providing a data basis for process quality evaluation and optimization.

[0187] Step 302, multi-dimensional process quality evaluation;

[0188] Build a multi-dimensional process quality evaluation system to comprehensively evaluate the effectiveness of process execution. First, define the evaluation indicator set:

[0189] ;

[0190] Including dimensions such as processing efficiency, satisfaction, and resource consumption, among which Indicates the evaluation indicators; then based on the evaluation dataset Calculate each treatment process plan Scores on each indicator ; Finally, the overall quality score of the processing process is calculated by combining the scores of each dimension:

[0191] ;

[0192] in, For indicators The weight is dynamically adjusted by the following formula:

[0193] ;

[0194] This sub-step outputs the process quality score. , providing the objective function for process optimization.

[0195] Step 303: Processing flow optimization based on differential evolution;

[0196] The differential evolution algorithm is used to discover the optimal processing flow pattern. The specific implementation of this algorithm in the optimization of urban management event processing flow is as follows:

[0197] Process flow encoding: combining process flows Encoded as a feature vector , using real number encoding, each dimension represents the selection probability or execution parameter of a processing flow segment. For example, for urban management events involving road damage, including road inspection, problem recording, classification and diversion, responsibility allocation, and processing tracking, it can be encoded as a vector , where each dimension represents the weight or parameter setting of the corresponding processing flow segment.

[0198] Initial population construction: Based on historical high-quality processing and random mutation, the initial population is constructed:

[0199] ;

[0200] Population size Set to 50, which is enough to maintain population diversity without excessively increasing the computational burden; Indicates the individuals (i.e., a process encoding vector).

[0201] Evolutionary operation: Iterate the following operations for 20 to 30 generations until convergence or the maximum number of iterations is reached:

[0202] Variation: for each individual Generate mutation vector :

[0203] ;

[0204] in Three randomly selected different individual indexes, none of which are equal to the current individual index ; Indicates the current algebra; Indicates the Daidi individual; is the scaling factor, set to 0.7, which is used to control the scaling degree of the differential vector; Represents the generated mutation vector.

[0205] Crossover: Mutation vector With the target vector Perform crossover to generate test vectors :

[0206] ;

[0207] in is the crossover probability, set to 0.5; Represents the dimension index of the vector; is a randomly selected dimension index, ensuring that at least one dimension comes from the mutation vector; Indicates generating a random number between 0 and 1; represents the generated trial vector.

[0208] Select: Compare test vectors With the target vector The fitness of , select the better one to enter the next generation:

[0209] ;

[0210] in The function decodes the encoded vector into a specific processing flow combination. is the quality scoring function; Indicates the Daidi Individuals.

[0211] Adaptive adjustment: During the iteration process, the weight of the evaluation function is dynamically adjusted according to the performance of the processing process in the actual environment. For example, the weight of timeliness is increased during peak hours, and the weight of accuracy is increased in the processing of complex urban management events.

[0212] Optimal solution selection: Select the individual with the highest quality score from the final population as the optimization process mode:

[0213] ;

[0214] in, Represents the final algebra population, Indicates the optimal processing flow mode.

[0215] In the application examples, the algorithm shows strong adaptability to different types of urban management events.

[0216] For example, for urban management incidents such as illegal emissions from dump trucks, the algorithm can automatically adjust the order of monitoring and evidence collection and on-site disposal, as well as the resource allocation ratio, to improve processing efficiency; for urban management incidents such as street waterlogging, the algorithm will dynamically adjust the processing process priority and resource allocation of emergency pumping, traffic control, and facility maintenance based on factors such as the degree of waterlogging, regional importance, and weather forecasts.

[0217] Through continuous optimization of the differential evolution algorithm, the system can generate the most suitable processing flow for urban management events in different areas, different time periods, and different complexities, thereby improving the accuracy and efficiency of urban management event processing. For example, in a commercial area of ​​a certain city, the algorithm-optimized process for handling urban management events for off-site operations increased the resolution rate by 27% and shortened the average processing time by 42%. This sub-step outputs the optimal processing flow model. , as a reference for processing flow library updates and generation.

[0218] Step 304, generating a memory-enhanced processing flow system;

[0219] Implement a memory-enhanced process generation system to achieve continuous optimization and iterative evolution of the process. First, build an experience memory library:

[0220] ;

[0221] Stores city management event characteristics , corresponding high-quality processing flow and quality rating ; Then for the new city management event features , retrieve similar historical experiences from the memory bank:

[0222] ;

[0223] in, is the urban management event similarity function, which calculates the similarity between the feature vectors of two urban management events; Indicates the characteristics of historical urban management events that are most similar to the new urban management event; finally, the search results are integrated with the basic generation processing flow to form an enhanced generation processing flow :

[0224] ;

[0225] in, Generate a processing flow based on Similar history processing flow for retrieval, is the fusion function, To balance the parameters, the similarity is dynamically adjusted to control the fusion ratio of historical experience and newly generated processing flow. This sub-step outputs the enhanced generated processing flow , organically combining historical successful experiences with newly generated processing processes to achieve continuous improvement in the quality of processing processes.

[0226] In one embodiment of the present invention, a city management event scheduling and circulation system mainly includes the following core modules:

[0227] The Urban Management Event Semantic Analysis Module is responsible for semantically analyzing input urban management events and extracting their features and attributes. It includes a natural language processing unit, an urban management event classification unit, and a feature extraction unit. This module uses deep learning technology to process urban management event descriptions using a pre-trained domain-specific language model and outputs a structured urban management event feature vector.

[0228] Processing Flow Fragment Library Management Module: Maintains and manages the processing flow fragment library, including a fragment storage unit, an update maintenance unit, and an index retrieval unit. This module uses a distributed database architecture that supports high-concurrency access and real-time updates, ensuring efficient management and rapid retrieval of processing flow fragments.

[0229] The Processing Intent Understanding and Verification module is responsible for understanding processing intent and verifying the processing flow plan. It includes a processing intent recognition unit, a processing flow verification unit, and a defect detection unit. This module integrates a multimodal deep learning model, which can learn processing intent patterns from historical cases and perform multi-dimensional verification of the generated processing flow.

[0230] Crowdsourcing Feedback Collection Module: This module collects feedback from processing personnel and stakeholders, including feedback collection, sentiment analysis, and topic extraction. This module provides a multi-channel feedback interface and supports the collection and analysis of structured ratings and unstructured text reviews.

[0231] Process Optimization and Evolution Module: This module optimizes process quality based on feedback data and includes a quality assessment unit, a differential evolution unit, and a memory management unit. This module implements an adaptive optimization algorithm that continuously learns from actual execution results and optimizes process design.

[0232] Process Execution and Monitoring Module: This module is responsible for the specific execution and status monitoring of the process, including the task distribution unit, status tracking unit, and exception handling unit. This module adopts a microservices architecture to ensure high availability and reliability of process execution.

[0233] Modules exchange data and collaborate through standardized interfaces, forming a complete closed-loop system for urban management event scheduling and circulation. The system utilizes a cloud-native design philosophy, supports horizontal scalability and elastic computing resource allocation, and can handle the high-concurrency urban management event processing requirements. Furthermore, the system implements end-to-end data encryption and access control mechanisms to ensure the security and compliance of sensitive information.

[0234] In one embodiment of the present invention, an application example of the aforementioned method for dispatching and transferring urban management events is provided:

[0235] The urban management event dispatching and transfer method of this embodiment focuses on the actual application in a certain city's urban management system, and mainly handles the following types of urban management events:

[0236] Municipal facilities and urban management incidents: including damaged roads, missing manhole covers, damaged public facilities, and other issues involving municipal infrastructure;

[0237] Environmental sanitation and urban management incidents: including garbage accumulation, sewage overflow, green space destruction and other issues affecting environmental sanitation;

[0238] Noise nuisance urban management incidents: including construction noise, commercial noise, square dance noise and other issues that affect residents' lives;

[0239] Illegal construction urban management incidents: including illegal construction, occupying the road for business, unauthorized reconstruction and other violations of urban planning;

[0240] Sudden urban management incidents: including sudden problems such as pipeline bursts and road collapses that require urgent handling.

[0241] A certain city's urban management system receives approximately 500,800 urban management incident reports daily. These incidents are distributed across different areas of the city and vary in urgency and complexity. Before implementing this method, the city used a traditional, fixed process template to handle urban management incidents. This process faced low processing efficiency, poor process adaptability, and an inability to effectively respond to new types of urban management incidents. The average processing time for urban management incidents was 72 hours, and the satisfaction rating was only 65 out of 100.

[0242] An example of semantic feature extraction for urban management events;

[0243] Taking the urban management incident of square dancing noise disturbing residents in a park as an example, the system processing flow is as follows:

[0244] Original urban management incident description text: There is a square dance group activity at the north gate square of a park from 7 to 9 pm every night. The volume is too loud and affects the rest of the surrounding residents. This has lasted for two weeks and the residents have complained many times but to no avail.

[0245] Text preprocessing and word segmentation: The system performs word segmentation and part-of-speech tagging on the text, obtaining the following word sequence: [a certain park (place name), north gate (direction), square (noun), every night (time), 7-9 am (time), square dance (noun), group (noun), activity (noun), volume (noun), too loud (adjective), influence (verb), surrounding (direction), resident (noun), rest (verb), continue (verb), two weeks (time), resident (noun), many times (frequency), reflect (verb), invalid (adjective)];

[0246] Domain vocabulary matching: The system identifies key domain terms: square dancing, noise (inferred from excessive volume), and disturbing the public (inferred from affecting residents' rest), and calculates similarity scores.

[0247] Attention weight calculation: The system calculates the similarity between each word in the word sequence and the domain vocabulary, and assigns higher attention weights to words such as square dance, volume, too loud, impact, residents, and rest.

[0248] Feature vector generation: The system generates semantic feature vectors of urban management events , including dimensional information such as urban management event type, location, time, and impact level.

[0249] As shown in Table 1, the attention weight distribution of domain keywords in the processing of urban management events is shown:

[0250] Table 1: Attention weight distribution of domain keywords in urban management event processing

[0251]

[0252] Automatic generation and verification examples of urban management event processing processes;

[0253] Based on the extracted semantic features, the system performs the following processing:

[0254] Urban management event classification: The system classifies the urban management event as noise nuisance, cultural and sports activities noise, and square dance noise.

[0255] Processing intention identification: Through the processing intention identification model, the system identifies the composite processing intention whose main processing intention is to mediate noise conflicts and standardize cultural and sports activities, rather than simply punishing violations.

[0256] Processing process segment selection: The system selects processing process segments suitable for this type of urban management event from the processing process segment library, including: on-site investigation, noise measurement, mediation and negotiation, formulation of activity specifications, follow-up visits, etc.

[0257] Process combination and verification: The system generates an initial process combination based on the constraint satisfaction model and uses the process intent verification algorithm to evaluate the matching degree between the process and the process intent. Table 2 shows the verification results of the two possible process solutions initially generated:

[0258] Table 2: Results of treatment process scheme verification

[0259]

[0260] Crowdsourcing feedback optimization examples;

[0261] After executing the generated processing flow plan, the system collects feedback from all parties and optimizes it:

[0262] Feedback data collection: The system collected feedback data from processing personnel, reporting residents and square dance groups, including satisfaction scores, text reviews and execution efficiency data.

[0263] The feedback data and process adjustments for handling square dance noise urban management incidents are shown in Table 3:

[0264] Table 3: Feedback data and process adjustments for handling square dance noise urban management incidents

[0265]

[0266] Feedback analysis and process adjustment: Through sentiment analysis and topic extraction, the system identified key points in the feedback: coordination time, clear time limits, and alternative venues. Based on this feedback, the system adjusted the process template for handling such urban management incidents:

[0267] Add a sub-step to clarify the time limit of the activity;

[0268] Add the sub-step of providing alternative site recommendations;

[0269] Optimize the mediation and negotiation process and introduce preset mediation plans to shorten the coordination time;

[0270] Differential Evolution Optimization: The system encodes the process flow for handling these urban management incidents into feature vectors and incorporates them into the initial population of the differential evolution algorithm, along with historical success cases. After 30 generations of iterative optimization, the system discovered that for urban management incidents such as square dancing noise, a communication-first, measurement-later approach resulted in higher satisfaction and a 22% improvement in processing efficiency, compared to measurement-first, communication-later approach. Consequently, the system adjusted the priority of these process segments.

[0271] Memory-enhanced learning: The system stores the optimized processing flow and its results in an experience memory bank. Later, when the system encounters a similar urban management incident involving noise disturbances from nighttime basketball courts, it uses similarity calculation to retrieve successful experiences in handling square dance noise and generate a more appropriate processing flow.

[0272] The improvement in urban management event processing efficiency is shown in Table 4:

[0273] Table 4: Changes in average processing time for different types of urban management incidents

[0274]

[0275] The processing accuracy and adaptability are shown in Table 5:

[0276] Table 5: Comparative analysis of urban management incident handling results

[0277]

[0278] The trend of system learning ability changing over time is shown in Table 6:

[0279] Table 6: Trends in system learning capabilities over time

[0280]

[0281] The above data can verify that the urban management event scheduling and circulation method of this embodiment has indeed achieved the expected technical effect, improved the efficiency and accuracy of urban management event processing, and has shown obvious advantages especially in handling complex and changeable non-conventional urban management events.

[0282] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Ordinary technicians in this field can also make more forms of equivalent embodiments based on the inspiration of this embodiment, all of which are protected by this embodiment.

Claims

1. A method for dispatching and transferring urban management events, characterized in that: The following steps are involved: Step 100: Analyze the urban management event description text using natural language processing technology, extract the semantic features of the urban management event, and build an urban management event semantic model library and a processing flow fragment library; generate the optimal initial processing flow combination based on the urban management event semantic features and constraints; Step 200: Build a framework for identifying processing intents for urban management events, extract processing intent features based on historical success cases, evaluate the matching degree between the automatically generated processing flow and the processing intent, and detect and correct potential defects in the processing flow. Specifically, the framework includes: Step 201, processing intent identification and understanding, includes: building a processing intent identification framework to understand the actual business needs of urban management event processing; Step 202, process flow verification based on process intent, includes: establishing a semantic association matrix between process flow segments and process intent , the elements in the matrix Represents a processing flow fragment and processing intent categories The correlation degree; calculation process combination: ； in, Represents a specific processing flow combination; represents the i-th processing flow fragment, Indicates the total number of processing flow fragments; and processing intent Overall match score: ； in, , Processing process fragments The weight of To process intent vectors Middle The probability of class processing intent, To process the total number of intent categories; Indicates the total number of processing flow fragments; Represents the i-th processing flow fragment and processing intent vectors Similarity function of Represents the i-th processing flow fragment in the processing flow combination and processing intent categories ; Step 203: Based on the verification result, a processing intention and processing flow mapping library is constructed and the processing flow combination is optimized, including: first, establishing a processing intention and processing flow mapping library: ;Record processing intention type , corresponding high-quality processing flow combination and validation scores ; Then for the new processing intention , find the most similar processing intent pattern in the mapping library ;Finally, by combining the initial processing flow and high-quality processing pipelines in the mapping library , generate optimized processing flow combination ; Indicates the high-quality processing flow corresponding to similar processing intentions retrieved from the mapping library; Step 204, detecting and correcting potential defects in the processing flow, including: combining the optimized processing flow Detect and correct potential defects in the processing flow and generate a repaired processing flow plan ; Step 300: Adaptive optimization of the processing flow based on crowdsourcing feedback. After executing the generated processing flow plan, the system collects feedback from all parties and performs optimization, including: Step 301, crowdsourcing feedback data collection and pre-processing, including: building a processing process evaluation framework based on crowdsourcing feedback, collecting feedback information from processing personnel and stakeholders of urban management events; Step 302, multi-dimensional process quality evaluation, builds a multi-dimensional process quality evaluation system to comprehensively evaluate the effectiveness of process execution, including: first, defining an evaluation indicator set: ; including processing efficiency, satisfaction, and resource consumption dimensions, among which Indicates the evaluation indicators; then based on the evaluation dataset Calculate each process combination Scores on each indicator ; Finally, the overall quality score of the process combination is calculated by combining the scores of each dimension ; Step 303: Optimize the processing flow based on differential evolution. Apply the differential evolution algorithm to find the optimal processing flow pattern. The specific implementation of this algorithm in optimizing the urban management event processing flow is as follows: Process flow encoding: combining process flows Encoded as a feature vector , using real number encoding, each dimension represents the weight or parameter setting of the corresponding processing flow segment; Initial population construction: Based on historical high-quality processing and random mutation, the initial population is constructed: NP is the population size, represents the i-th individual, i.e. a process encoding vector; Iteratively performs the following operations until convergence or the maximum number of iterations is reached: Variation: for each individual Generate mutation vector ; Indicates the Daidi Individuals, For the Daidi The mutation vector of each individual, is the number of generations in the current iteration; Crossover: Mutation vector and Perform crossover to generate test vectors ; Select: Compare test vectors and The fitness of , select the better one to enter the next generation: ； in The function decodes the encoded vector into a specific processing flow combination. is the quality scoring function; Indicates the Daidi individual; Optimal solution selection: Select the individual with the highest quality score from the final population as the optimization processing flow model.

2. A method for dispatching and transferring urban management events according to claim 1, characterized in that: The process of extracting semantic features of urban management events includes: constructing domain vocabulary : ； in, Indicates the Field terms, Indicates the total number of terms in the vocabulary; Contains professional terms in the field of urban management, including: garbage dumping, noise nuisance, illegal buildings; description text of urban management events Perform word segmentation and part-of-speech tagging to obtain word sequences : ； in, Indicates the first words, Indicates the total number of words after word segmentation; Calculate the similarity score between each word and the domain vocabulary ; The attention weight is calculated as follows: ； in, 、 Respectively represent the i-th and j-th words after word segmentation; Represents exponential operation; Represents the domain vocabulary; Represents the similarity calculation function; Indicates the total number of words after word segmentation. Expressive words The attention weight of a word reflects the importance of the word in the description of urban management events; Generate semantic feature vectors of urban management events : ； in For words The word vector representation of ; Expressive words The attention weight of Indicates the total number of words after word segmentation.

3. A method for dispatching and transferring urban management events according to claim 1, characterized in that: The intent recognition framework includes a multi-layer neural network model, which consists of an input layer, an encoding layer, an attention layer, and an output layer. The forward propagation process is as follows: ； ； ； ； ； in, is the semantic feature vector of urban management events, To process the intent vector, 、 、 denote the first, second and third weight matrices respectively, 、 、 Represent the first, second and third bias vectors respectively, 、 are the outputs of the first and second hidden layers respectively, is the attention weight matrix, is the attention-weighted feature, for Dimensions; for Activation function; is the normalization function.

4. A method for dispatching and transferring urban management events according to claim 1, characterized in that: The process of detecting and correcting potential defects in the processing flow includes: defining a processing flow defect pattern library: ； Contains common process design flaws; builds defect detection functions , identify the conditions that meet Defects ; Apply the repair rule set Generate the repaired processing flow plan: ； in is the combination of processing procedures to be detected, is the detection threshold, is the set of detected defects, represents the i-th detected defect, t represents the total number of detected defects, This is a bug fix function.

5. A method for dispatching and transferring urban management events according to claim 1, characterized in that: The execution process of the differential evolution algorithm includes: Mutations: ； Crossover: Generate trial vectors based on crossover probability ; in, 、 、 are three different individuals randomly selected, is the scaling factor.

6. A method for dispatching and transferring urban management events according to claim 1, characterized in that: The process quality evaluation system calculates the overall quality score of the process using the following formula: : ； in, For indicators The weight of Dynamic adjustment, is the indicator importance parameter, is the total number of evaluation indicators, Combining processes In the indicator The score on .

7. A method for dispatching and transferring urban management events according to claim 1, characterized in that: Also includes memory-enhanced process generation: Building an experience memory bank: ； Stores city management event features, corresponding high-quality processing procedures and quality scores; for new city management event features , retrieve similar historical experiences from the memory bank: ； The search results are integrated with the basic generation process to form an enhanced generation process: ； in is the balance parameter; Represents the basic generation process flow; Represents the similar historical processing flow of retrieval; represents the similarity function of urban management events; Represents the processing flow fusion function; Characterize historical urban management events stored in memory banks; For the corresponding high-quality processing flow; Rate the quality; A feature vector for new city management events; Indicates obtaining the maximum parameter.

8. A method for dispatching and transferring urban management events according to claim 1, characterized in that: The urban management event scheduling flow method also includes real-time monitoring of the execution of the processing flow, and triggering the process of regenerating and verifying the processing flow when an exception occurs in the execution of the processing flow.

9. A city management event scheduling and circulation system, used to implement the steps of the city management event scheduling and circulation method according to any one of claims 1 to 8, characterized in that: include: Urban management event semantic analysis module, used to analyze urban management event description text and extract urban management event features; A processing flow fragment library management module is used to maintain and manage the processing flow fragment library; Processing intention understanding and verification module, used to understand the processing intention and verify the processing flow plan; Crowdsourcing feedback collection module, used to collect feedback information from processing personnel and relevant parties; Process optimization and evolution module, used to optimize the quality of the process based on feedback data; The processing flow execution and monitoring module is used for the specific execution and status monitoring of the processing flow.

Images