Public data objection processing evaluation system based on artificial intelligence

Abstract

The invention relates to the technical field of public data processing, and discloses a public data objection processing evaluation system based on artificial intelligence. The system comprises a data objection receiving module used for receiving and structurally storing objection information; the intelligent classification and grading module classifies and grades the objection information based on a machine learning model; the automatic checking module is used for determining an objection data source and a problem type through an algorithm and a model; the intelligent distribution module analyzes the objection content and dynamically distributes the objection content; the dynamic evaluation and feedback module is used for evaluating the processing progress, adjusting resource allocation and feeding back a result; and the continuous learning optimization module is used for optimizing the model by utilizing historical data. According to the system, efficient processing and evaluation of public data objection are realized, the data quality is improved, and the processing efficiency and public satisfaction are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of public data processing, and in particular to a public data objection processing evaluation system based on artificial intelligence. Background Art

[0002] In the digital age, public data, as an important information resource, is widely used in many fields such as government decision-making, social governance, and public services. With the continuous expansion of data scale and the increasing complexity of application scenarios, the quality of public data has gradually become prominent, and data objections have frequently occurred. If these objections are not handled in a timely and effective manner, they will not only affect the accuracy and reliability of the data, but may also have a negative impact on the decision-making of government departments, the quality of public services, and the stable development of the social economy.

[0003] The traditional way of handling public data objections mainly relies on manual operations, which has many disadvantages. First, manual processing is inefficient. Faced with massive public data and a large amount of objection information, manual investigation and processing one by one requires a lot of time and manpower costs. For example, in the process of census data processing, if data objections occur, it often takes a long time to manually check the data source and find the problem, which will lead to untimely data updates and affect the formulation and implementation of relevant policies. Secondly, the accuracy of manual processing is difficult to guarantee. The types of data objections are complex and diverse, involving multiple aspects such as the standardization, completeness, and accuracy of data. Manual judgment is easily affected by subjective factors, resulting in misjudgment or omission. Taking the business registration data of enterprises as an example, manual review may fail to find errors in the data due to negligence, thus affecting the normal operation of the enterprise and the effectiveness of market supervision.

[0004] In addition, the existing data objection handling lacks an effective classification and grading mechanism. The inability to quickly and accurately determine the nature of the objection and the attribution of responsibility leads to a chaotic objection handling process, unclear responsibilities between departments, and frequent buck-passing. For example, in the handling of urban planning data objections, due to the lack of clear classification and grading standards, data problems involving multiple departments may not be able to determine the responsible party in a timely manner, resulting in long-term unresolved problems and affecting the advancement of urban construction.

[0005] In terms of data verification, traditional methods lack systematicity and scientificity. It is difficult to fully and deeply trace the source and flow path of data, and it is impossible to accurately determine whether the data problem lies in the collection, storage or processing stage. For example, in a medical data sharing platform, if there is an objection to incorrect patient information, it is difficult to quickly locate the root cause of the problem with traditional methods, which affects the quality of medical services and the protection of patients' rights and interests.

[0006] Moreover, the existing objection handling system also has deficiencies in resource allocation and feedback mechanisms. It cannot dynamically adjust the resource allocation strategy according to the actual situation, resulting in low processing efficiency; at the same time, the feedback to the objection proposer is not timely and comprehensive, reducing the public's satisfaction and trust in data processing. For example, in the handling of traffic violation data objections, due to unreasonable resource allocation, some objections are not processed for a long time and the feedback is not timely, causing public dissatisfaction and doubts. Summary of the Invention

[0007] The purpose of the present invention is to provide an evaluation system for public data objection handling based on artificial intelligence to solve the problems raised in the above background technology.

[0008] To achieve the above purpose, the present invention provides the following technical solutions: An evaluation system for public data objection handling based on artificial intelligence, the system includes: A data objection receiving module, which is used to receive public data objection information through a government affairs data platform, a government affairs service platform and multi-channel interfaces, and structurally store the objection information. An intelligent classification and grading module, which automatically classifies and grades the received objection information based on a machine learning model. The classification includes objections to data standardization, integrity, and accuracy, and the grading includes responsibility grading and nature grading. An automated verification module, which determines the source and problem type of the objection data through a data lineage tracing algorithm and a data consistency verification model, and generates a technical verification report. An intelligent distribution module, which uses natural language processing technology to analyze the objection content, dynamically distributes it to the corresponding data provider or government affairs data platform in combination with the responsibility grading result, and establishes a multi-party collaborative processing channel. A dynamic evaluation and feedback module, which constructs an objection handling progress evaluation model based on real-time data streams, dynamically adjusts the resource allocation strategy, and feeds back the processing result to the objection proposer. A continuous learning and optimization module, which uses historical objection data to train classification models and verification models to optimize the classification accuracy and verification efficiency.

[0009] Preferably, in the intelligent classification and grading module, the machine learning model uses an ensemble learning algorithm, including random forest and gradient boosting decision tree, which is used to automatically match predefined classification labels according to the keywords, semantic features and metadata of the objection content, and output the probability distributions of responsibility grading and nature grading.

[0010] Preferably, in the automated verification module, the data lineage tracing algorithm traces the transfer path of the objection data in the collection, storage, and processing links by parsing the metadata graph of the data source, and performs differential comparison in combination with the version control log to locate the data abnormal node.

[0011] Preferably, in the intelligent distribution and collaborative processing module, a sequence model based on the attention mechanism is used to semantically parse the objection description, extract the keywords of the responsible entity, and generate the objection distribution path in combination with the predefined responsible entity priority table.

[0012] Preferably, in the dynamic evaluation and feedback module, the evaluation model is based on time series analysis and processing log data, calculates the response efficiency, processing duration, and problem-solving rate indicators of each department, and dynamically optimizes the resource allocation weights through the reinforcement learning algorithm.

[0013] Preferably, in the continuous learning and optimization module, incremental learning technology is used to update the classification model parameters, and synthetic simulation objection data is generated through the generative adversarial network.

[0014] Preferably, the data lineage tracing algorithm further includes: dynamically assigning weights to the nodes in the metadata graph, and adjusting the priority of the tracing path according to the complexity and correlation of the data processing link.

[0015] Preferably, the responsible entity priority table is generated through a game theory model, and the priority ranking of the responsible entities is dynamically adjusted based on the cooperation efficiency and problem-solving rate of each entity in the historical processing records.

[0016] Preferably, the reinforcement learning algorithm uses the Q-learning framework, defines the state space as the current resource allocation ratio, the action as adjusting the processing task quotas between departments, and the reward function is comprehensively calculated based on the processing efficiency improvement rate and the satisfaction of the objection submitter.

[0017] Preferably, the present invention further includes an electronic device, the device includes a memory and a processor, the memory stores a computer program, and is characterized in that when the processor executes the program, it realizes the functions of the above-mentioned public data objection processing and evaluation system based on artificial intelligence.

[0018] Compared with the prior art, the beneficial effects of the present invention are: The public data objection processing and evaluation system based on artificial intelligence proposed by the present invention innovates and optimizes from multiple key links, bringing significant beneficial effects.

[0019] In the data objection receiving and storing link, the system receives objection information through the government affairs data platform, government affairs service platform, and multi-channel interfaces, and performs structured storage. This method greatly broadens the objection collection channels, facilitates the public to feedback problems, and at the same time, the structured storage is convenient for the rapid retrieval and invocation of data, providing convenience for the subsequent processing process and improving the overall efficiency of data processing.

[0020] The intelligent classification and grading module uses machine learning models, especially random forests and gradient boosting decision trees in ensemble learning algorithms, to automatically classify and grade dissent information. This module can accurately identify dissent regarding data standardization, integrity, and accuracy, as well as responsibility grading and nature grading. Taking the handling of a large amount of financial data dissent as an example, by automatically matching predefined classification labels, the system can quickly determine the type of dissent, such as non-standard data format, data missing, or data error, and clarify the responsible department and the severity of the dissent, making subsequent processing more targeted, avoiding the subjectivity and low efficiency of manual classification and grading, and improving the accuracy and timeliness of processing.

[0021] The automated verification module plays a key role by means of data lineage tracing algorithms and data consistency verification models. The data lineage tracing algorithm traces the data flow path by parsing the metadata graph and combining version control logs, locates abnormal nodes, and assigns dynamic weights to the nodes to adjust the tracing priority. In the processing of financial transaction data, if there is a dissent regarding abnormal transaction amounts, this algorithm can quickly find the source of the data problem, whether it is an input error during collection or an error in subsequent processing, and can accurately locate it. The data consistency verification model further ensures the accuracy of the data, and the technical verification report generated by the combination of the two provides a strong basis for subsequent processing.

[0022] The intelligent distribution module uses natural language processing technology to analyze dissent content, dynamically distributes dissent in combination with the responsibility grading results, and establishes a multi-party collaborative processing channel. This effectively solves the problems of inaccurate dissent distribution and difficult collaboration between departments in traditional processing methods. In the scenario of handling dissent regarding urban construction data, the system can accurately extract the keywords of the responsible entity, quickly distribute the dissent to the corresponding department according to the responsibility entity priority table, and at the same time build a collaborative channel to promote information sharing and cooperation between departments and speed up the problem-solving process.

[0023] The dynamic evaluation and feedback module constructs an evaluation model based on real-time data streams, calculates indicators such as the response efficiency, processing duration, and problem-solving rate of each department, and uses reinforcement learning algorithms to dynamically optimize the resource allocation weights. In this way, the system can flexibly adjust resource allocation according to the actual processing situation, invest more resources in links with low processing efficiency or complex problems, and improve the overall processing efficiency. At the same time, the processing results are promptly feedback to the dissent submitter, enhancing the public's satisfaction and trust in data processing.

[0024] The continuous learning and optimization module uses historical dissent data to train classification models and verification models. Incremental learning technology updates the parameters of the classification model, and generative adversarial networks synthesize simulated dissent data to continuously optimize the classification accuracy and verification efficiency. As time goes by, the amount of dissent data processed by the system increases, and the model can continuously learn new features and rules, adapt to complex and changing data environments, and maintain high processing capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 FIG. 1 is a schematic working diagram of the public data objection handling and evaluation system based on artificial intelligence according to the present invention; Figure 2 FIG. 2 is a schematic working diagram of the semantic parsing and distribution path generation of the intelligent distribution module; Figure 3 FIG. 3 is a schematic working diagram of the dynamic evaluation and feedback module; Figure 4 FIG. 4 is a schematic working diagram of the continuous learning and optimization module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0027] Please refer to Figures 1 - 4 FIG. 1, the present invention provides a public data objection handling and evaluation system based on artificial intelligence, aiming to efficiently and intelligently handle public data objections and improve data quality and service level. The following details the specific implementation solutions and related embodiments of the system. The overall implementation solution is as follows: Data Objection Receiving Module: This module receives public data objection information through the government affairs data platform, government affairs service platform and multi-channel interfaces. In actual application scenarios, the government affairs data platform, as the core hub for aggregating government department data, may cover data in multiple fields such as population information and enterprise registration data; the government affairs service platform is oriented to public services, such as online service handling, business query and other functional modules. The multi-channel interfaces may include email, dedicated mobile APP feedback entrances, government affairs service hotlines, etc. When receiving objection information, the system will perform structured storage on it, for example, classify and store it according to fields such as the time when the objection is raised, the type of objection, and the field to which the data involved in the objection belongs, facilitating subsequent retrieval and processing.

[0028] Intelligent Classification and Grading Module: This module automatically classifies and grades the received objection information based on a machine learning model. The classification mainly covers objections to data normativity, integrity, and accuracy; the grading includes responsibility grading and nature grading. Through this classification and grading method, the system can quickly locate the key information of the objection, providing clear guidance for the subsequent processing process. For example, for an objection regarding an error in the registered capital information in enterprise registration data, the system can quickly determine it as an objection to data accuracy and determine the corresponding responsible department and the severity of the objection nature.

[0029] Automated Verification Module: By leveraging data lineage tracing algorithms and data consistency verification models, this module determines the source and problem type of the disputed data and generates a technical verification report. During the data processing process, data may go through multiple links, such as collection, storage, and processing. The data lineage tracing algorithm can trace the flow path of data through these links, and the data consistency verification model checks the consistency of the data. Taking census data as an example, if there are abnormal age data, the system can trace back to the data collection source through the algorithm to determine whether it is an input error during the collection process or a problem in the subsequent processing link, and generate a detailed verification report.

[0030] Intelligent Distribution Module: Using natural language processing technology to parse the content of the dispute, it is dynamically distributed to the corresponding data provider or government data platform in combination with the result of responsibility classification, and a multi-party collaborative processing channel is established. When parsing the content of the dispute, the system can extract key information, such as the data subject involved in the dispute, problem description, etc. For example, if the dispute involves the tax data of a certain enterprise, the system determines the responsible department according to the responsibility classification, accurately distributes the dispute information, and builds a communication bridge between the party raising the dispute, the data provider, and the government data platform to facilitate collaborative processing among all parties.

[0031] Dynamic Evaluation and Feedback Module: Based on real-time data streams, an evaluation model for the progress of dispute handling is constructed, the resource allocation strategy is dynamically adjusted, and the processing result is feedback to the party raising the dispute. During the process of handling disputes, the system monitors the progress of each link in real time, such as the response time of departments, processing duration, etc. According to this data, the system adjusts the resource allocation to improve the processing efficiency. After the processing is completed, the processing result is promptly feedback to the party raising the dispute, such as through text messages, emails, or message push on the government service platform.

[0032] Continuous Learning and Optimization Module: Using historical dispute data to train classification models and verification models to optimize the classification accuracy and verification efficiency. As the number of disputed data processed increases, the system continuously learns and optimizes. For example, by analyzing the dispute characteristics in historical data, the recognition ability of the classification model is improved; the verification model is optimized to locate data problems faster.

[0033] The following further details the specific implementation manners of the present invention through 5 embodiments: Embodiment

[0034] In the intelligent classification and grading module, the machine learning model adopts an ensemble learning algorithm, including random forest and gradient boosting decision tree. For the received dispute information, first extract its keywords, semantic features, and metadata. The keyword extraction can adopt an algorithm based on term frequency-inverse document frequency (TF-IDF). Assume the dispute information is in text form, use to represent this text collection, to represent a certain word, The frequency of occurrence of a word in a document represents the number of documents that contain the word represents the total number of documents in the document collection. Then, the TF-IDF value calculation formula for the word is as follows: ; Semantic features are obtained through word vector models in natural language processing technologies, such as Word2Vec or BERT, which convert the text into vector representations for easy model understanding and processing. Metadata includes information such as the data source and creation time.

[0035] The random forest algorithm constructs multiple decision trees to learn and classify these features. Let the number of decision trees be , and each decision tree is trained based on randomly selected samples and features. For a new objection information , the prediction result of the random forest is the synthesis of the prediction results of multiple decision trees. Assume that the decision tree predicts the result of as , the predicted category is , and the number of votes for the category is . Then, the predicted category of the random forest for is the category with the most votes, that is: ; Among them, is an indicator function that takes the value of 1 when is equal to and 0 otherwise.

[0036] The gradient boosting decision tree constructs the next decision tree based on the prediction error of the previous decision tree, continuously iterating and optimizing the model. By combining these two algorithms, the system can automatically match predefined classification labels according to the objection content and output the probability distributions of responsibility grading and nature grading, effectively improving the accuracy of classification and grading.

[0037] In practical applications, for example, for a large number of public transportation data objection information, by extracting keywords such as "incorrect station location" and "missing train number" through the above methods, combined with semantic features and metadata, the model can accurately classify the objections as data accuracy or integrity objections and give the corresponding probability of responsibility grading and nature grading. This helps to handle objections targeted later, improving the handling efficiency and quality.

[0038] Example 2: In the automated verification module, the data lineage tracing algorithm traces the transfer path of the disputed data by parsing the metadata graph of the data source. The metadata graph is a graphical representation of information such as the structure, relationships, and sources of data in the data source. Nodes in the graph represent data elements or data processing steps, and edges represent the flow or association relationships of data.

[0039] The algorithm first parses the metadata graph. Assume that the set of nodes in the graph is , and the set of edges is . For the disputed data, start tracing from its current node . During the tracing process, perform differential comparison in combination with the version control log. The version control log records the state changes of the data at different time points, represented by . Each record in it contains information such as the version number, modification time, and modification content of the data.

[0040] The algorithm determines the data anomaly nodes by comparing data of different versions. For example, in a financial data processing process, if it is found that the amount of a transaction data is abnormal. First, start from the current node of the data in the metadata graph, and trace back to the previous processing step according to the direction of the edge, and view the modification record of the data at this step in the version control log. Assume that at step , it is found that the data amount has an abnormal modification in a certain version. By comparing the data values before and after the modification, determine that this node is an abnormal node.

[0041] In addition, the data lineage tracing algorithm also dynamically assigns weights to the nodes in the metadata graph. Adjust the priority of the tracing path according to the complexity and correlation degree of the data processing link. Let the weight of node be , the complexity index be , and the correlation index be . Then the weight calculation formula is: ; where is the weight adjustment parameter, and its value range is , which can be adjusted according to the actual situation. The complexity index can be measured by calculating the number of processing steps included in the data processing link where node is located; the correlation index is measured by calculating the number of edges directly connected to node . In this way, the system can preferentially trace the paths with higher weights, locate the root cause of data problems faster, and improve the verification efficiency.

[0042] Example 3: In the intelligent distribution module, a sequence model based on the attention mechanism is used to semantically parse the objection description and extract the keywords of the responsible entity. Suppose the objection description is a text sequence , where represents the -th word in the sequence, is the length of the sequence.

[0043] The sequence model based on the attention mechanism first converts the text sequence into a vector representation through the embedding layer , where is the vector corresponding to the word . Then, the attention weights are calculated, and the formula is: ; where , , , are model parameters. The attention weight indicates the degree of attention of the model when processing the -th word.

[0044] The weighted sum of the word vectors is calculated through the attention weights to obtain the comprehensive representation of the objection description. The keywords of the responsible entity are extracted from this comprehensive representation, for example, through a keyword extraction algorithm or a pre-trained named entity recognition model.

[0045] Combined with the predefined responsible entity priority table, the objection distribution path is generated. The responsible entity priority table is generated through a game theory model, and based on the cooperation efficiency and problem-solving rate of each entity in the historical processing records, the priority ranking of the responsible entities is dynamically adjusted. Let the set of responsible entities be , and the cooperation efficiency of the responsible entity in the historical processing records is , and the problem-solving rate is . The priority score of the responsible entity is calculated through the game theory model, and the formula is: ; where is the weight adjustment parameter, and its value range is , which can be adjusted according to the actual situation. The responsible entities are sorted according to the priority scores to obtain the responsible entity priority table. When the keywords of the responsible entity of the objection are extracted, the target responsible entity for objection distribution is determined according to the priority table, and the objection distribution path is generated to achieve the accurate distribution of objections.

[0046] For example, when dealing with objections to urban planning data, the objection description is "The newly built road plan in a certain area does not match the actual terrain." Through the above semantic analysis process, the keyword of the responsible entity "urban planning department" is extracted. Combining with the priority table of the responsible entity, it is determined that this objection should be preferentially distributed to the urban planning department to ensure that the objection can reach the processing department in a timely and accurate manner, improving the processing efficiency.

[0047] Example 4: In the dynamic evaluation and feedback module, the evaluation model calculates the response efficiency, processing duration, and problem-solving rate indicators of each department based on time series analysis and processing log data. Assume that the processing log data contains the processing records of objections by each department, and each record contains information such as department name, reception time, processing completion time, and processing result.

[0048] For department , the response efficiency is measured by calculating the time interval from receiving the objection to starting to process. Let the time when department receives the objection be , and the time when starting to process be , then the calculation formula for the response efficiency is: ; The processing duration is measured by calculating the time interval from starting to process to processing completion. Let the time when department completes the processing be , then the calculation formula for the processing duration is: ; The problem-solving rate is measured by calculating the ratio of the number of objections successfully solved by department to the total number of received objections. Let the number of objections successfully solved by department be , and the total number of received objections be , then the calculation formula for the problem-solving rate is: ; Dynamically optimize the resource allocation weight through the reinforcement learning algorithm. The reinforcement learning algorithm adopts the Q-learning framework, defines the state space as the current resource allocation ratio, and assumes that the resource allocation ratio vector is , where represents the resource ratio allocated to department , and . The action is to adjust the processing task quota among departments, and the reward function is comprehensively calculated based on the processing efficiency improvement rate and the satisfaction of the dissenting party. Let the processing efficiency improvement rate be , and the satisfaction of the dissenting party be . The calculation formula of the reward function is: ; Among them, is the weight adjustment parameter, and its value range is , which can be adjusted according to the actual situation.

[0049] The system selects actions according to the current state, observes the new state and rewards after executing the actions, and optimizes the resource allocation weights through continuous iterative learning to improve the efficiency and quality of dissent handling. For example, when handling public data dissents in multiple fields, the system dynamically adjusts the resource allocation according to the response efficiency, processing duration, and problem-solving rate of each department, allocating more resources to the departments with high processing efficiency and high problem-solving rate, and at the same time improving the satisfaction of the dissenting party.

[0050] Example 5: In the continuous learning and optimization module, incremental learning technology is used to update the classification model parameters. Incremental learning means that on the basis of an existing model, new data is continuously added for learning without having to retrain the entire model. Assume that the classification model is , the existing training data is , and the new dissent data is .

[0051] First, preprocess the new data to make it meet the model input requirements. Then, divide the new data into multiple small batch data. Let the number of small batches be , and each small batch data is .

[0052] For each small batch data , optimization algorithms such as stochastic gradient descent are used to update the model parameters. Let the parameters of the model be , the loss function be , and the learning rate be . Then the model parameter update formula is: ; By gradually processing each small batch of data and continuously updating the model parameters, the model can adapt to the new data characteristics and improve the classification accuracy.

[0053] Synthesize simulated dissent data through a generative adversarial network. The generative adversarial network consists of a generator and a discriminator Composition. Generator takes random noise as input and outputs synthetic objection data . Discriminator takes real objection data and synthetic objection data as input and outputs the probability of judging whether the data is real.

[0054] During the training process, the goal of the generator is to generate synthetic data that can deceive the discriminator , that is, to maximize ; the goal of the discriminator is to accurately distinguish real data from synthetic data, that is, to maximize . Through continuous iterative training, the generator can generate more realistic objection data. The synthetic objection data is added to the training dataset to enrich the diversity of the training data, further optimize the classification model and verification model, and improve the performance of the system.

[0055] For example, when dealing with objections to public health data, as new objection data continuously emerges, the classification model is updated through incremental learning techniques so that it can better identify new types of objections. At the same time, adversarial generative networks are used to synthesize simulated objection data, such as simulating data objections that may occur in different regions and different scenarios, providing more diverse data for model training, and enhancing the generalization ability of the model and the ability to handle complex objections.

[0056] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0057] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A public data objection handling evaluation system based on artificial intelligence, characterized in that, It includes the following modules: A data objection receiving module, which is used to receive public data objection information through a government affairs data platform, a government affairs service platform and multi-channel interfaces, and perform structured storage on the objection information; An intelligent classification and grading module, which automatically classifies and grades the received objection information based on a machine learning model. The classification includes objections to data standardization, integrity, and accuracy, and the grading includes responsibility grading and nature grading; An automated verification module, which determines the source and problem type of the objection data through a data lineage tracing algorithm and a data consistency verification model, and generates a technical verification report; An intelligent distribution module, which uses natural language processing technology to parse the objection content, dynamically distributes it to the corresponding data provider or government affairs data platform in combination with the responsibility grading result, and establishes a multi-party collaborative processing channel; A dynamic evaluation and feedback module, which constructs an objection processing progress evaluation model based on real-time data streams, dynamically adjusts the resource allocation strategy, and feeds back the processing result to the objection submitter; A continuous learning and optimization module, which uses historical objection data to train classification models and verification models to optimize classification accuracy and verification efficiency.

2. The system according to claim 1, wherein In the intelligent classification and grading module, the machine learning model uses an ensemble learning algorithm, including random forest and gradient boosting decision tree, to automatically match predefined classification labels according to the keywords, semantic features and metadata of the objection content, and output the probability distributions of responsibility grading and nature grading.

3. The system according to claim 1, wherein In the automated verification module, the data lineage tracing algorithm traces the transfer path of the objection data in the collection, storage, and processing links by parsing the metadata graph of the data source, and performs differential comparison in combination with the version control log to locate the data anomaly nodes.

4. The system according to claim 1, characterized in that, In the intelligent distribution and collaborative processing module, a sequence model based on the attention mechanism is used to semantically analyze the objection description, extract the keywords of the responsible entity, and generate an objection distribution path in combination with a predefined responsible entity priority table.

5. The system according to claim 1, characterized in that, In the dynamic evaluation and feedback module, the evaluation model calculates the response efficiency, processing duration, and problem resolution rate indicators of each department based on time series analysis and processing log data, and dynamically optimizes the resource allocation weights through a reinforcement learning algorithm.

6. The system according to claim 1, wherein In the continuous learning and optimization module, incremental learning technology is used to update the parameters of the classification model, and synthetic simulation objection data is generated through a generative adversarial network.

7. The system according to claim 3, characterized in that, The data lineage tracing algorithm also includes: dynamically assigning weights to the nodes in the metadata graph, and adjusting the priority of the tracing path according to the complexity and correlation of the data processing link.

8. The system according to claim 4, characterized in that, The responsible entity priority table is generated through a game theory model, and the priority ranking of the responsible entities is dynamically adjusted based on the cooperation efficiency and problem resolution rate of each entity in the historical processing records.

9. The system according to claim 5, wherein The reinforcement learning algorithm uses the Q-learning framework, defines the state space as the current resource allocation ratio, the action as adjusting the processing task quota between departments, and the reward function is comprehensively calculated based on the processing efficiency improvement rate and the satisfaction of the objection submitter.

10. An electronic device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the program, it realizes the functions of the artificial intelligence-based public data objection processing and evaluation system as described in any one of claims 1 to 9.

Images