Electronic government affair platform management method and system based on cloud data

Abstract

The invention discloses an e-government affair platform management method and system based on cloud data, and relates to the technical field of data processing, and the method comprises the steps: collecting multi-source data, carrying out the preprocessing, extracting keywords, carrying out the multi-scale time coding and multi-scale space coding of the data, and storing the data in a database; constructing a government affair knowledge graph based on the database, mining association relationships of department services and analyzing affair handling habits of users; generating a government affair examination and approval path based on the knowledge graph, optimizing the government affair examination and approval path through reinforcement learning, dynamically adjusting the examination and approval process, performing a simulation test by using a digital twinning technology, and verifying the user satisfaction after the examination and approval process is optimized; and storing the government affair data in a database and performing authority management. According to the method, the electronic government affair data processing capability can be improved, the examination and approval efficiency is improved, the verifiability of an optimization strategy is enhanced, and technical support is provided for digital and intelligent transformation of government affair services.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a cloud data-based e-government platform management method and system. Background Art

[0002] With the development of information technologies such as cloud computing, big data, and artificial intelligence, e-government is gradually evolving towards intelligence and data-driven. The traditional government management model mainly relies on offline approval, paper document storage, and manual process management, which leads to lengthy government approval processes and inefficient information sharing, making it difficult to meet the needs of modern society for intelligent, automated, and efficient government services. In recent years, the widespread application of cloud data technology has enabled government information systems to efficiently store and process massive amounts of data, and intelligent analysis technologies based on knowledge graphs, natural language processing (NLP), and reinforcement learning have also been gradually introduced to improve the intelligence level of government services. For example, using knowledge graphs to perform semantic association analysis on government data can mine the associations between cross-departmental businesses, optimize business processes, and reduce redundant approval links. In addition, the introduction of streaming computing frameworks (such as Apache Flink) has transformed government data processing from traditional approval processing modes to real-time computing modes, which can improve data flow efficiency and realize intelligent recommendations, process optimization, and automated approval of government services. In terms of government approval optimization, the digital twin technology that has emerged in recent years has provided a new solution for optimizing government business processes. By building a virtual government approval system, we can test the optimization effects of different approval paths in a simulation environment, and combine reinforcement learning to achieve dynamic optimization of the approval process. However, the existing technology still faces many challenges, such as the heterogeneity of government data and the difficulty of optimizing the approval process. These problems limit the intelligent development of the e-government platform. Summary of the invention

[0003] In view of the above existing problems, the present invention is proposed.

[0004] Therefore, the present invention provides an e-government platform management method and system based on cloud data to solve many challenges that still exist in the existing technology, such as the heterogeneity of government data and the difficulty of optimizing the approval process. These problems limit the intelligent development of the e-government platform.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a cloud data-based e-government platform management method, which comprises: Collect multi-source data, pre-process and extract keywords, perform multi-scale time coding and multi-scale spatial coding on the data and store the data in the database; The multi-source data includes user operation data, government business data, and IoT device data; Build a government knowledge graph based on the database, mine the correlation between departmental businesses and analyze users' handling habits; Generate government approval paths based on knowledge graphs, optimize government approval paths through reinforcement learning, dynamically adjust approval processes, and use digital twin technology for simulation testing to verify user satisfaction after the approval process is optimized; Store government data in the database and manage permissions.

[0006] As a preferred solution of the cloud data-based e-government platform management method of the present invention, extracting keywords includes the following steps: Use WordNet and Wikipedia semantic matching to determine synonyms in the government domain; Extract professional terms from government announcements, government websites, and policy and regulatory documents; Establish multi-level dictionaries, including basic dictionaries, institutional dictionaries, industry dictionaries, and action dictionaries; Count the words that appear frequently in government data texts but are not included in the dictionary, and calculate the weight of new words through TF-IDF, set the threshold Q, and if the weight of new words is greater than the threshold Q, filter out high-value new words and dynamically update the multi-level dictionary; Combined with the maximum forward matching and hidden Markov model, the dynamically updated multi-level dictionary is used for accurate word segmentation, and the Viterbi algorithm is used to calculate the optimal word segmentation path. During the calculation process, the state of the optimal choice at each step is recorded to generate a path storage table. The path storage table is backtracked to obtain the word boundary. New words are extracted through the backtracked boundary information, and the new words are integrated to form a predicted word set. Combine the known words of the maximum positive match and the predicted words of the hidden Markov model, calculate the confidence of the hidden Markov model word segmentation and the maximum positive match word segmentation, set the threshold A, if the confidence of the hidden Markov model predicted word is greater than the threshold A, then combine the hidden Markov model result with the maximum positive match word segmentation to get the final word segmentation result, and store the new word in the multi-level dictionary, otherwise directly use the maximum positive match result; Define government-specific part-of-speech categories and generate a government-specific annotation dataset, use the dataset to train a conditional random field model, and use the trained conditional random field model to predict part-of-speech tags; Based on Jaccard similarity calculation, duplicate removal is performed. A similarity threshold of O is set. If the similarity is greater than the threshold of O, two words are merged. A keyword set is generated based on the filtered and duplicate-free words. Count the co-occurrence frequency P(W,E) of keyword pairs (W,E) in the same government text, and calculate the independent occurrence probabilities P(W) and P(E) of single words W and E in all government documents. Use PMI to calculate the keyword correlation PWI(A,B), determine the semantic hierarchy of keywords W and E in WordNet, use Leacock-Chodorow similarity to calculate the semantic similarity between words, combine with Wikipedia knowledge base, extract the relevant context vectors of keywords W and E, and use cosine similarity to calculate the correlation between the two keywords. The similarity calculation results of WordNet and Wikipedia are combined using a weighted method to obtain the final keyword similarity score, and a similarity threshold y is set to retain only keyword pairs with a similarity greater than the threshold y. Define the Skip-Gram model and input keywords into the model to obtain word vectors; Use the K-Means clustering algorithm to perform cluster analysis on word vectors, use TextRank to calculate the importance of keywords, obtain the importance score of keywords, sort the keywords in descending order according to the importance score, and select the top M keywords with the highest scores as the final output.

[0007] As a preferred solution of the cloud data-based e-government platform management method of the present invention, wherein: building a government knowledge graph based on a database, mining the association relationship of departmental business and analyzing the user's handling habits include: Use hash mapping to parse time-coded data and map it back to specific business stages, including application stage, review stage, approval stage, and re-examination stage; Use nearest neighbor matching to parse spatially coded data and map it to the corresponding government agencies and approval areas; Use the triple model to establish a government data graph based on time, space and business. Use the cosine similarity method to calculate the correlation between different government businesses. Parse the user operation records in the knowledge graph to generate a user behavior dataset. Use the time window sliding method to organize user operations into time series, calculate the user business access frequency, and use Markov chain analysis to calculate the user business jump probability.

[0008] As a preferred solution of the cloud data-based e-government platform management method described in the present invention, wherein: generating a government approval path based on a knowledge graph, optimizing the government approval path through reinforcement learning, and dynamically adjusting the approval process include: Based on the time and space information of the user's applied business, obtain the stage of the current approval business, determine the starting stage of the approval, obtain the historical businesses similar to the current approval business according to the business relevance in the knowledge graph, calculate the access frequency of the approval nodes as the weights of the approval nodes, calculate the business transition probability of the user under the current approval business based on Markov chain analysis, calculate the total weight of the approval path using the normalized linear weighted method based on the approval node weights and the user's business transition probability, and select the approval path with the highest weight as the optimal approval path; Construct the state space X based on the nodes of the optimal approval path, calculate the posterior probability of the optimal approval path through Bayesian inference, and construct the initial strategy of reinforcement learning through Bayesian inference; Set the optimization actions that can be selected for reinforcement learning. Each state X must be able to map to at least one action A. Collect historical approval data, select the strategy with the highest approval passing rate and the shortest approval time to form a strategy set, calculate the strategy score using the Q-value function, use the generalization strategy to select the optimal strategy under the current approval path state. For the business that requires manual approval, set the constraint conditions and the minimum limit of the approval nodes. Set the optimization goal for the approval time. Only when the optimized approval time is less than the set target value will the optimization strategy be adopted; Take the set state space and optimization strategy set as the input of the Q-network, set the initial Q-value, and initialize the weights of the Q-network using a uniform distribution. Set the storage structure of the experience pool, and randomly extract samples of optimized approval paths from the experience pool each time for training; Calculate the Q-value using the Bellman equation, set the Q-value update rule, use the generalization strategy update to optimize the Q-value, select the best path from multiple historical optimal strategies and update the current approval path; Optimize the Q-network through gradient descent to obtain the optimized optimal approval path.

[0009] As a preferred solution of the e-government platform management method based on cloud data described in the present invention, the following steps are included: use digital twin technology for simulation testing to verify the user satisfaction after the approval process optimization, including determining the Petri net elements, setting the simulation parameters, running the optimized approval path in the simulation environment, collecting the execution data of the approval process, calculating the overall optimization rate of the approval process, recording the performance of the optimized approval path in the simulation environment, calculating the approval efficiency improvement rate based on the optimized approval success rate and the historical approval success rate, setting the target threshold U. If the improvement rate is greater than the threshold U, then deploy it. Otherwise, record the problems that do not meet the expectations and readjust the reinforcement learning model, and repeat the training process.

[0010] As a preferred solution of the electronic government affairs platform management method based on cloud data of the present invention, wherein: storing government affairs data in a database and performing permission management means storing the original government affairs data and the corresponding government affairs approval data in JSON format, storing them using a database, setting user access permissions, and only allowing authorized users to access.

[0011] As a preferred solution of the electronic government affairs platform management method based on cloud data of the present invention, wherein: storing the data in a database after performing multi-scale time encoding and multi-scale space encoding on the data includes the following steps: Based on the extracted keywords of government affairs data, determine the time granularity of each piece of data, perform time encoding on each piece of government affairs data, and determine the spatial attributes of the data according to the keywords of the government affairs data after time encoding. Convert the geographical coordinates of government affairs data into a one-dimensional index, use Morton encoding to store the spatial index, combine Hilbert encoding and Morton encoding, use Hilbert for large-scale data queries, use Morton for small-scale data queries, and store the encoded data in the database.

[0012] In a second aspect, the present invention provides an electronic government affairs platform management system based on cloud data, including: A data collection module, used to collect government affairs data from different data sources and perform preprocessing; A keyword extraction module, used to perform text analysis on government affairs data, extract core keywords, and perform time and space encoding on the data; A knowledge graph construction module, used to establish the association relationship of government affairs data using knowledge graph technology and analyze the business habits of users; An intelligent approval path optimization module, used to optimize the approval path using a reinforcement learning model and realize the dynamic adjustment of the approval process; A digital twin simulation test module, used to perform simulation tests on the optimized approval path using Petri net simulation technology to verify the optimization effect; A data storage module, used to store approval data and perform access permission control.

[0013] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and wherein: when the computer program is executed by the processor, any step of the electronic government affairs platform management method based on cloud data as described in the first aspect of the present invention is implemented.

[0014] Fourthly, the present invention provides a computer-readable storage medium, on which a computer program is stored, wherein: when the computer program is executed by a processor, any step of the e-government platform management method based on cloud data as described in the first aspect of the present invention is implemented.

[0015] The beneficial effects of the present invention are as follows: The present invention uses Kafka and Flink to achieve high-concurrency data access and stream computing, combines Morton coding and Hilbert coding to improve data storage and retrieval efficiency, constructs a government affairs business association network using a knowledge graph, and combines reinforcement learning to realize intelligent optimization and dynamic adjustment of the approval path. With the digital twin simulation environment modeled by Petri nets, the feasibility of the approval optimization plan is quantified to ensure the efficiency and reliability of the optimization path in practical applications. The present invention can improve the e-government data processing ability, enhance the approval efficiency, and strengthen the verifiability of the optimization strategy, providing technical support for the digital and intelligent transformation of government services. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0017] Figure 1 It is a flowchart of the e-government platform management method based on cloud data in Embodiment 1.

[0018] Figure 2 It is a structural diagram of the e-government platform management system based on cloud data in Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention in conjunction with the drawings in the specification.

[0020] Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar promotions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0021] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that excludes other embodiments.

[0022] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides an e-government platform management method based on cloud data, comprising the following steps: S1, collect multi-source data, pre-process and extract keywords, perform multi-scale time coding and multi-scale spatial coding on the data and store the data in the database; Specifically, Kafka is used as the message queue for high-concurrency data access to ensure smooth data transmission, and Flink is used for streaming computing to process user operation data (query, application, approval records), government business data (policy announcements, service guidelines), and IoT device data (queue status in government halls) in real time to ensure efficient data flow and unify and standardize the time format of the collected data.

[0023] Extracting keywords includes the following steps: Parse the encoding type of the data source and convert it uniformly, retain only the text description part of the IoT device data, remove structured fields (such as device ID, timestamp), and convert unstructured data into plain text format to ensure parsability, such as PDF conversion → OCR recognition of text, HTML conversion → remove tags, and only retain the text content; Use regular matching to remove invalid characters, including HTML tags, spaces, special symbols (non-Chinese characters, emoticons, special punctuation marks, only retaining text information) and delete duplicate data, remove common stop words and government-related stop words (such as "application number" and "approval unit"); Use WordNet + Wikipedia semantic matching to determine synonyms in the government affairs field; Extract professional terms from government announcements, government websites, and policy and regulatory documents, such as business, organization, and region; Establish a multi-level dictionary, including a basic dictionary (including common government terms), an institutional dictionary (including the names of government agencies and service departments), an industry dictionary, and an action dictionary (including high-frequency verbs in the process of government affairs (such as "submit", "apply", "approve")); Count the words that appear frequently in government data texts but are not included in the dictionary, and calculate the weight of new words through TF-IDF. Set the threshold Q based on experience. If the weight of a new word is greater than the threshold Q, filter out high-value new words and dynamically update the multi-level dictionary. Combine the maximum forward matching (FMM) and the hidden Markov model (HMM), use the dynamically updated multi-level dictionary to accurately segment words and ensure that new words can also be correctly recognized. The specific steps are as follows: Use the maximum forward matching method to sequentially match the longest entries in the text from left to right; If a short word fails to match, use the Hidden Markov Model for out-of-vocabulary word recognition. Define the key parameters of the Hidden Markov Model, including the state set, the observation set (consisting of single characters of out-of-vocabulary words), the initial state probability, the state transition probability (calculated using the frequency statistics method), and the observation probability (calculated using the maximum likelihood estimation method); The said state set , B is the starting character of a word, L is the middle character of a word, E is the ending character of a word, and S is a single character forming a word; Based on the defined key parameters, use the Viterbi algorithm to calculate the optimal word segmentation path. During the calculation process, record the state generation path storage table of the optimal choice at each step. Through backtracking using the path storage table, obtain the word boundaries. Based on the boundary information obtained from backtracking, extract new words and integrate the new words to form a predicted word set; Combine the known words of the maximum forward matching and the predicted words of the Hidden Markov Model, calculate the confidence levels of the Hidden Markov Model word segmentation and the maximum forward matching word segmentation. Set a threshold A through experiments. If the confidence level of the predicted words of the Hidden Markov Model is greater than the threshold A, then combine the results of the Hidden Markov Model with the maximum forward matching word segmentation to obtain the final word segmentation result, and store the new words in a multi-level dictionary. Otherwise, directly adopt the maximum forward matching result; Define the special part-of-speech categories for government affairs and generate a government affairs annotation dataset. Use the dataset to train a conditional random field model and use the trained conditional random field model to predict part-of-speech tags; The said special part-of-speech categories for government affairs include government affairs business categories (BIZ), organization categories (ORG), place name categories (LOC), and action categories (ACT); Screen the core keywords, only retain the words in the special part-of-speech categories for government affairs, and remove meaningless words (such as "de", "shi", "women", etc.) Calculate deduplication based on the Jaccard similarity. Set a similarity threshold O based on statistical analysis of historical government affairs data. If the similarity is greater than the threshold O, then merge the two words. Generate a keyword set based on the screened and deduplicated words; Statistically analyze the co-occurrence frequency P(W,E) of keyword pairs (W,E) in the same government affairs text and calculate the independent occurrence probabilities P(W) and P(E) of individual words W and E in all government affairs documents. Use PMI to calculate the correlation PWI(A,B) of keywords:

[0024] Set a threshold w based on statistical analysis and only retain the keyword pairs with PMI greater than w; Determine the semantic hierarchy of keywords W and E in WordNet, and calculate the semantic similarity between words using the Leacock-Chodorow similarity (LCH); Combine the Wikipedia knowledge base, extract the relevant context vectors of keywords W and E, and calculate the correlation between the two keywords using cosine similarity; Use a weighted method to synthesize the similarity calculation results of WordNet and Wikipedia to obtain the similarity score of the final keywords; Based on the actual application scenario, set the similarity threshold y, and only retain the keyword pairs with similarity greater than the threshold y; Define the Skip-Gram model, select the government affairs text data containing keywords (such as policy announcements, approval rules, user consultation records) as the training set, set the window size to k, extract k data from the training set for model training, use the negative sampling strategy to optimize the model calculation efficiency to obtain the trained Skip-Gram model, and input the keywords into the model to obtain word vectors; Use the K-Means clustering algorithm to perform clustering analysis on the word vectors, determine the categories corresponding to the keywords, construct a keyword graph based on the semantic clustering results, use TextRank to calculate the importance of the keywords, obtain the importance scores of the keywords, sort the keywords in descending order according to the importance scores, and select the top M keywords with the highest scores as the final output.

[0025] Through Kafka+Flink high-concurrency data access and streaming computing, realize the real-time processing of government affairs data; combine technologies such as OCR parsing, keyword extraction, multi-level dictionary matching, and HMM prediction of new words to improve the accuracy of government affairs data processing, and use algorithms such as TextRank, K-Means clustering, and PMI correlation calculation to perform semantic analysis and weight calculation on government affairs keywords to ensure the scientificity and reliability of keyword extraction. The combination of TextRank and the Skip-Gram model can automatically identify the core concepts in government affairs documents, reduce the interference of redundant information, and automatically extract the abstract of government affairs announcements through keyword weight sorting, improving the efficiency of obtaining government affairs information. FMM is suitable for known words, and HMM is suitable for unknown words. The combination of the two improves the recognition rate of emerging government affairs terms. Use Jaccard similarity to remove duplicates, eliminate data redundancy, and improve the accuracy of keyword extraction.

[0026] Furthermore, the steps of storing the data in the database after multi-scale time encoding and multi-scale spatial encoding of the data include the following: Based on the extracted keywords of government affairs data, determine the time granularity of each piece of data. Define government affairs data with an annual granularity as long-term valid data, such as policy documents and regulations. Define government affairs data with a daily and hourly granularity as short-term business data, such as approval processes and service records. Define government affairs data with a second and minute granularity as real-time data, such as the queuing situation in government service halls; Subtract the minimum time value in the business data (such as the first online time of the business) from the original timestamp of the data record and divide it by the time granularity to perform time encoding on each piece of government affairs data; Based on the keywords of government affairs data after time encoding, determine the spatial attributes of the data, including national government affairs data (such as laws and regulations), provincial-level data (such as provincial-level approval items), city-level data (such as municipal approvals), and real-time data in government service halls (such as queuing situations); Convert the geographical coordinates of government affairs data into a one-dimensional index, and use Morton encoding (Z-curve) to store the spatial index. Combine Hilbert encoding and Morton encoding, using Hilbert for large-scale data queries and Morton for small-scale data queries to ensure query efficiency. This method provides a geographical location index for the knowledge graph, enabling cross-regional analysis and optimization of the approval process; Store the encoded data in the database.

[0027] Through multi-scale time encoding and multi-scale spatial encoding techniques, the storage, query, and analysis of government affairs data are made more efficient. At the same time, it provides technical support for intelligent approval path optimization, cross-regional business recommendation, and intelligent analysis of government affairs processes. Combining the bit-interleaved index method of Morton encoding enables government affairs data queries to reduce disk I / O overhead and improve query speed. Hilbert encoding provides better spatial locality and is suitable for range queries, such as querying "the approval efficiency of all service halls in a certain province in the past three months".

[0028] S2. Build a government affairs knowledge graph based on the database, mine the correlation relationships of departmental operations, and analyze users' service habits; Specifically, building a government affairs knowledge graph based on the database, mining the correlation relationships of departmental operations, and analyzing users' service habits include: Use hash mapping to parse the time-encoded data and map it back to the specific business stages, including the application stage, review stage, approval passed stage, and re-review stage; Use nearest neighbor matching to parse the space-encoded data and map it to the corresponding government agencies and approval regions; Use a triple model to establish a government affairs data graph based on time, space, and business. Use the cosine similarity method to calculate the correlation degree of different government affairs operations. Analyze the user operation records in the knowledge graph to generate a user behavior dataset. Use the time window sliding method to organize user operations into a time series. Calculate the user business access frequency based on the number of times the user has handled the current business and the total number of the user's approval operations. Use Markov chain analysis to calculate the user business jump probability.

[0029] Through key technologies such as hash mapping, nearest neighbor matching, knowledge graph, time window sliding, and Markov chain analysis, the intelligent association of government affairs data, the analysis of user behavior patterns, and the intelligent optimization of the approval process are realized, effectively improving the management level of the e-government platform. Compared with the traditional government affairs approval method, the present invention can dynamically adjust the approval path, reduce approval redundancy, and optimize the user experience, making the e-government platform more efficient, intelligent, and user-friendly.

[0030] S3. Generate a government affairs approval path based on the knowledge graph, optimize the government affairs approval path through reinforcement learning, dynamically adjust the approval process, and use digital twin technology for simulation testing to verify the user satisfaction after the approval process is optimized; Specifically, generating a government affairs approval path based on the knowledge graph and optimizing the government affairs approval path through reinforcement learning and dynamically adjusting the approval process include: According to the time and space information of the user's applied business, obtain the stage of the current approval business, determine the starting stage of the approval, obtain the historical business similar to the current approval business according to the business correlation degree in the knowledge graph, and calculate the access frequency of the approval node as the weight of the approval node. Based on Markov chain analysis, calculate the user's business jump probability under the current approval business. Based on the approval node weight and the user's business jump probability, use the normalized linear weighting method to calculate the total weight of the approval path, and select the approval path with the highest weight as the optimal approval path; Construct a state space based on the nodes of the optimal approval path, wherein, is the starting point of the approval process, is the end point of the approval process. Through Bayesian inference, calculate the posterior probability of the optimal approval path, and through Bayesian inference, construct the initial strategy of reinforcement learning; Set the optimization actions that can be selected for reinforcement learning wherein, is to optimize the approval node order (adjust the business transfer logic), is to reduce redundant approval nodes (delete unnecessary approval steps), is to increase the proportion of automatic approval (reduce manual approval and improve automation), It is a dynamic merging approval node (merging similar or duplicate approval processes), and each state X must be able to map to at least one action A; Collect historical approval data, select the strategy with the highest approval passing rate and the shortest approval time to form a strategy set, calculate the strategy score using the Q-value function, use the generalization strategy to select the optimal strategy under the current approval path state. For the business that requires manual approval, set the constraint conditions and the minimum limit of the approval node. If the optimized path of a certain approval process results in the number of approval nodes less than the minimum value, the optimization strategy will not be executed. Set the optimization goal of the approval time. Only when the optimized approval time is less than the set target value, the optimization strategy will be adopted; Take the set state space and optimization strategy set as the input of the Q network, set the initial Q value, and initialize the Q network weights using a uniform distribution. Set the storage structure of the experience pool, including the current approval state, the currently executed optimization strategy, the current reward value, and the next approval state. Randomly extract samples of optimized approval paths from the experience pool each time during training to prevent the model from falling into a local optimal solution; Calculate the Q value using the Bellman equation, set the Q value update rule, use the generalization strategy update to optimize the Q value, select the best path from multiple historical optimal strategies and update the current approval path; Optimize the Q network through gradient descent to obtain the optimized optimal approval path.

[0031] The knowledge graph not only constructs the logical associations between approval operations, but also can automatically recommend the optimal approval path based on historical business data. The approval path based on the knowledge graph can adapt to the complex business requirements of cross-departmental collaboration, improve the collaborative efficiency between approval operations. By counting the jump behaviors of users in different operations, the common operation paths of users during the approval process can be predicted, reducing approval blocking points. Combining Markov chain analysis, the business nodes with higher approval frequencies can be preferentially optimized, thus reducing the waiting time of users. Bayesian inference can dynamically adjust the approval path, enabling the approval process to be optimized as policies are adjusted and user needs change. Through posterior probability calculation, the selection of the approval path is made more accurate, reducing ineffective approval links. The traditional approval process relies on preset rules, while reinforcement learning can dynamically optimize the approval path to adapt to different business requirements. Reinforcement learning can automatically recommend the optimal approval strategy through Q-value function calculation, improving the approval efficiency. Set the minimum limit for approval nodes to avoid over-simplifying the approval steps during the optimization process, resulting in business compliance issues. Set the optimization goal for approval time to ensure that the optimized approval path does not affect the approval quality while improving efficiency. The Bellman equation can ensure that the optimization process gradually approaches the optimal approval path, reducing unnecessary iteration processes. The optimized approval path by Q-value can clearly quantify the optimality of different paths, providing a reference basis for government affairs personnel. Through random sampling training, the experience pool can enhance the generalization ability of the model, enabling it to adapt to more approval business scenarios. The historical approval data stored in the experience pool can be used for subsequent optimization, improving the reusability of the approval optimization plan.

[0032] Furthermore, digital twin technology is used for simulation testing to verify the user satisfaction after the approval process optimization, including determining Petri net elements, including state sets, transition sets, flow relationships, and weight sets; The state set represents the key states of the approval process, such as "application submitted", "preliminary review completed", "final review completed", the transition set represents the state transitions between approval nodes, such as "approved", "supplementary materials required", the flow relationship describes the transfer rules between states, and the weight set represents the resource consumption of each approval task; Set simulation parameters, including approval processing time and passing rate, run the optimized approval path in the simulation environment, collect the execution data of the approval process, calculate the overall optimization rate of the approval process based on the duration of the optimized approval process and the average duration of the original approval process, record the performance of the optimized approval path in the simulation environment, including whether the approval time is significantly reduced, whether the proportion of automatic approvals is increased, and whether user satisfaction is improved (based on approval success rate and waiting time), calculate the approval efficiency improvement rate based on the optimized approval success rate and the historical approval success rate, statistically analyze the historical approval data to set the target threshold U. If the improvement rate is greater than the threshold U, then deploy; otherwise, record the problems that do not meet expectations and readjust the reinforcement learning model, and repeat the training process.

[0033] Refine the modeling of the approval process through Petri nets, enabling the optimization of the approval path to be fully verified in the simulation environment to ensure its applicability to real business scenarios. By setting parameters such as approval processing time and passing rate, the performance of different approval paths in different business scenarios can be accurately simulated, making the optimization plan more reliable. Through data collection, the effect of optimizing the approval process can be quantified, providing data support for subsequent optimization and ensuring the scientific nature of the decision-making on optimizing the approval path. By calculating the optimization rate of the approval process, it can intuitively reflect whether the optimized approval path truly improves the approval efficiency and avoid the problem that the optimization plan is only effective in specific scenarios. By setting the threshold U, it is ensured that the approval path optimization plan will only be officially deployed after meeting certain optimization criteria, avoiding the situation where the optimization effect is lower than expected. Through the iterative training of the reinforcement learning model, the approval path can be continuously optimized, enabling the system to adapt to different e-government business scenarios and achieve dynamic optimization.

[0034] S4. Store the e-government data in the database and perform permission management; Specifically, storing the e-government data in the database and performing permission management means storing the original e-government data and the corresponding e-government approval data in JSON format and using the database for storage, setting user access permissions, and only allowing authorized users to access.

[0035] In the e-government platform, the sources of e-government data are complex, including structured data (database records), semi-structured data (forms, logs), and unstructured data (text, pictures, etc.). Traditional relational databases are difficult to effectively store and manage these heterogeneous data. The JSON format has a hierarchical structure and can store different types of data in a nested manner, avoiding the problem of table splitting in relational databases, improving data query efficiency, and granting different access permissions to prevent data leakage.

[0036] This embodiment also provides an e-government platform management system based on cloud data, including: A data collection module, which is used to collect government affairs data from different data sources and perform preprocessing; A keyword extraction module, which is used to perform text analysis on government affairs data, extract core keywords, and perform time and space encoding on the data; A knowledge graph construction module, which is used to establish the association relationship of government affairs data by using knowledge graph technology and analyze the business habits of users; An intelligent approval path optimization module, which is used to optimize the approval path by using a reinforcement learning model and realize the dynamic adjustment of the approval process; A digital twin simulation test module, which is used to perform simulation tests on the optimized approval path by using Petri net simulation technology to verify the optimization effect; A data storage module, which is used to store approval data and perform access permission control.

[0037] This embodiment also provides a computer device, which is applicable to the situation of the electronic government affairs platform management method based on cloud data, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to implement the electronic government affairs platform management method based on cloud data proposed in the above embodiment.

[0038] The computer device can be a terminal. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner. The wireless manner can be achieved through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the shell of the computer device, or an external keyboard, a touchpad, or a mouse, etc.

[0039] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for managing an e-government platform based on cloud data as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, abbreviated as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, abbreviated as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, abbreviated as EPROM), programmable read-only memory (Programmable Red-Only Memory, abbreviated as PROM), read-only memory (Read-Only Memory, abbreviated as ROM), magnetic memory, flash memory, magnetic disk or optical disc.

[0040] In summary, the present invention uses Kafka and Flink to achieve high-concurrency data access and stream computing, combines Morton coding and Hilbert coding to improve data storage and retrieval efficiency, constructs a government affairs business association network using a knowledge graph, and combines reinforcement learning to achieve intelligent optimization and dynamic adjustment of the approval path. With the digital twin simulation environment modeled by Petri nets, the feasibility of the approval optimization plan is quantified to ensure the efficiency and reliability of the optimization path in practical applications. The present invention can improve the e-government data processing ability, enhance the approval efficiency, and strengthen the verifiability of the optimization strategy, providing technical support for the digital and intelligent transformation of government services.

[0041] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. An electronic government affairs platform management method based on cloud data, characterized in that: include, Collect multi-source data, pre-process and extract keywords, perform multi-scale time coding and multi-scale spatial coding on the data and store the data in the database; The multi-source data includes user operation data, government business data, and IoT device data; Build a government knowledge graph based on the database, mine the correlation between departmental businesses and analyze users' handling habits; Generate government approval paths based on knowledge graphs, optimize government approval paths through reinforcement learning, dynamically adjust approval processes, and use digital twin technology for simulation testing to verify user satisfaction after the approval process is optimized; Store government data in the database and manage permissions.

2. The management method of the e-government platform based on cloud data according to claim 1, characterized in that: The keyword extraction comprises the following steps: Use WordNet and Wikipedia semantic matching to determine synonyms in the government domain; Extract professional terms from government announcements, government websites, and policy and regulatory documents; Establish multi-level dictionaries, including basic dictionaries, institutional dictionaries, industry dictionaries, and action dictionaries; Count the words that appear frequently in government data texts but are not included in the dictionary, and calculate the weight of new words through TF-IDF, set the threshold Q, and if the weight of new words is greater than the threshold Q, filter out high-value new words and dynamically update the multi-level dictionary; Combined with the maximum forward matching and hidden Markov model, the dynamically updated multi-level dictionary is used for accurate word segmentation, and the Viterbi algorithm is used to calculate the optimal word segmentation path. During the calculation process, the state of the optimal choice at each step is recorded to generate a path storage table. The path storage table is backtracked to obtain the word boundary. New words are extracted through the backtracked boundary information, and the new words are integrated to form a predicted word set. Combine the known words of the maximum positive match and the predicted words of the hidden Markov model, calculate the confidence of the hidden Markov model word segmentation and the maximum positive match word segmentation, set the threshold A, if the confidence of the hidden Markov model predicted word is greater than the threshold A, then combine the hidden Markov model result with the maximum positive match word segmentation to get the final word segmentation result, and store the new word in the multi-level dictionary, otherwise directly use the maximum positive match result; Define government-specific part-of-speech categories and generate a government-specific annotation dataset, use the dataset to train a conditional random field model, and use the trained conditional random field model to predict part-of-speech tags; Based on Jaccard similarity calculation, duplicate removal is performed. A similarity threshold of O is set. If the similarity is greater than the threshold of O, two words are merged. A keyword set is generated based on the filtered and duplicate-free words. Count the co-occurrence frequency P(W,E) of keyword pairs (W,E) in the same government text, and calculate the independent occurrence probabilities P(W) and P(E) of single words W and E in all government documents. Use PMI to calculate the keyword correlation PWI(A,B), determine the semantic hierarchy of keywords W and E in WordNet, use Leacock-Chodorow similarity to calculate the semantic similarity between words, combine with Wikipedia knowledge base, extract the relevant context vectors of keywords W and E, and use cosine similarity to calculate the correlation between the two keywords. Use a weighted method to synthesize the similarity calculation results of WordNet and Wikipedia, obtain the similarity score of the final keywords, set the similarity threshold y, and only retain the keyword pairs with similarity greater than the threshold y; Define the Skip-Gram model and input the keywords into the model to obtain word vectors; Use the K-Means clustering algorithm to perform clustering analysis on the word vectors, use TextRank to calculate the importance of the keywords, obtain the importance scores of the keywords, sort the keywords in descending order according to the importance scores, and select the top M keywords with the highest scores as the final output.

3. The management method of the e-government platform based on cloud data according to claim 2, characterized in that: The construction of the government affairs knowledge graph based on the database and the mining of the association relationships of departmental operations and the analysis of users' handling habits include: Use hash mapping to parse the time-encoded data and map it back to the specific business stages, including the application stage, review stage, approval passed stage, and re-review stage; Use nearest neighbor matching to parse the space-encoded data and map it to the corresponding government agencies and approval regions; Use the triple model to establish a government affairs data graph based on time, space, and business, use the cosine similarity method to calculate the association degree of different government affairs operations, parse the user operation records in the knowledge graph to generate a user behavior data set, use the time window sliding method to organize the user operations into a time series, calculate the user business access frequency, and use Markov chain analysis to calculate the user business jump probability.

4. The method for managing an e-government platform based on cloud data according to claim 3, wherein: The generation of the government affairs approval path based on the knowledge graph and the optimization of the government affairs approval path through reinforcement learning and the dynamic adjustment of the approval process include: According to the time and space information of the user's applied business, obtain the current stage of the approval business, determine the starting stage of the approval, obtain the historical businesses similar to the current approval business according to the business association degree in the knowledge graph, and calculate the access frequency of the approval nodes as the weights of the approval nodes. Based on Markov chain analysis, calculate the user's business jump probability under the current approval business. Based on the approval node weights and the user business jump probability, use the normalized linear weighting method to calculate the total weight of the approval path, and select the approval path with the highest weight as the optimal approval path; Construct the state space X based on the nodes of the optimal approval path, calculate the posterior probability of the optimal approval path through Bayesian inference, and construct the initial strategy of reinforcement learning through Bayesian inference; Set the optimization actions that can be selected for reinforcement learning. Each state X must be able to map to at least one action A. Collect historical approval data, select the strategies with the highest approval passing rate and the shortest approval time to form a strategy set, use the Q-value function to calculate the strategy scores, use the generalized strategy to select the optimal strategy under the current approval path state. For the businesses that require manual approval, set the constraint conditions and the minimum limit of the approval nodes; Set the optimization goal for the approval time. Only when the optimized approval time is less than the set target value, the optimization strategy will be adopted; Take the set state space and optimization strategy set as the input of the Q network, set the initial Q value, and initialize the weights of the Q network using a uniform distribution. Set the storage structure of the experience pool, and randomly extract samples of the optimized approval path from the experience pool each time for training; Calculate the Q-value using the Bellman equation, set the Q-value update rule, use the generalization strategy update to optimize the Q-value, select the best path from multiple historical optimal strategies and update the current approval path; Optimize the Q-network through gradient descent to obtain the optimized optimal approval path.

5. The management method of the e-government platform based on cloud data according to claim 4, characterized in that: The use of digital twin technology for simulation testing to verify the user satisfaction after the approval process optimization includes determining Petri net elements, setting simulation parameters, running the optimized approval path in the simulation environment, collecting the execution data of the approval process, calculating the overall optimization rate of the approval process, recording the performance of the optimized approval path in the simulation environment, calculating the approval efficiency improvement rate based on the optimized approval success rate and the historical approval success rate, setting the target threshold U, if the improvement rate is greater than the threshold U, then deploy, otherwise record the problems that do not meet the expectations and readjust the reinforcement learning model, and repeat the training process.

6. The method for managing an e-government platform based on cloud data according to claim 5, wherein: The storing of government affairs data in the database and performing permission management means storing the original government affairs data and the corresponding government affairs approval data in JSON format, and using the database for storage, setting user access permissions, and only allowing authorized users to access.

7. The management method of the e-government platform based on cloud data according to claim 6, characterized in that: The storing of the data in the database after performing multi-scale time encoding and multi-scale space encoding on the data includes the following steps: Based on the extracted keywords of government affairs data, determine the time granularity of each piece of data, perform time encoding on each piece of government affairs data, and determine the spatial attributes of the data according to the keywords of the time-encoded government affairs data. Convert the geographical coordinates of the government affairs data into a one-dimensional index, and use Morton encoding to store the spatial index. Combine Hilbert encoding and Morton encoding, use Hilbert for large-scale data queries and Morton for small-scale data queries, and store the encoded data in the database.

8. An e-government platform management system based on cloud data, based on the e-government platform management method based on cloud data according to any one of claims 1 to 7, characterized in that: Including, A data collection module for collecting government affairs data from different data sources and performing preprocessing; A keyword extraction module for performing text analysis on government affairs data, extracting core keywords, and performing time and space encoding on the data; A knowledge graph construction module for using knowledge graph technology to establish the association relationship of government affairs data and analyze the business habits of users; An intelligent approval path optimization module for optimizing the approval path using a reinforcement learning model and realizing the dynamic adjustment of the approval process; A digital twin simulation testing module for using Petri net simulation technology to perform simulation testing on the optimized approval path and verifying the optimization effect; A data storage module for storing approval data and performing access permission control.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the method for managing an e-government platform based on cloud data according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the method for managing an e-government platform based on cloud data according to any one of claims 1 to 7.

Images