Intelligent checking system and method for customs affair compliance

Abstract

The invention belongs to the field of customs affair auditing, and particularly relates to a customs affair compliance intelligent auditing system and method, and the method comprises the steps: a data collection module synchronously generates an auditing rule updating table and a customs declaration service information table in real time through a multi-mode extraction algorithm; the compliance auditing module completes multi-dimensional compliance verification based on a transverse service auditing node network, a longitudinal service transmission node chain and a declaration node network in combination with a dynamic rule priority index and reinforcement learning algorithm; the evaluation optimization module calculates a service transmission risk probability, an auditing delay probability and an error probability in real time through a bidirectional risk evaluation model, and feeds back the probability to an auditing space to optimize a rule weight and a flow path; through a composite auditing architecture of'declaration clustering + longitudinal encryption transmission + transverse topological association ', dynamic adaptation of auditing rules, accurate risk traceability and intelligent process optimization are realized, and the problems of low efficiency in integration of multi-source heterogeneous data and lagging auditing of complex business scenes in cross-border trade are effectively solved.

Description

Technical Field

[0001] The present invention belongs to the field of customs audit, and in particular relates to a customs compliance intelligent audit system and method. Background Art

[0002] At present, customs declaration compliance audits face systemic challenges such as low efficiency in integrating multi-source heterogeneous data, weak dynamic adaptation capabilities of rules, and delayed tracing of hidden risks. Traditional audit systems rely on manual input and static rule engines, making it difficult to cope with complex scenarios such as cross-border e-commerce and multi-country transshipment, resulting in excessively long single-ticket audits and a greatly increased rate of missed detection of hidden violations. Although blockchain technology can improve data credibility, the insufficient throughput of public chains and the lack of deep collaboration with large models in vertical fields limit its practical value in real-time audit scenarios. Existing knowledge graph technologies are mostly limited to static rule mapping and cannot achieve dynamic reasoning. Cross-source alignment and real-time indexing technologies for text, images, and IoT sensor data are not yet mature, resulting in deviations in the extraction of key information and delays in audits. In addition, rule updates rely on manual configuration, and risk warnings lack causal traceability, further restricting the accuracy and agility of automated audits. For this reason, this application proposes a customs compliance intelligent audit system and method. Summary of the Invention

[0003] In response to the shortcomings of the existing technology, the present invention proposes a customs compliance intelligent audit system and system, which includes a data acquisition module, a compliance audit module and an evaluation and optimization module: the data acquisition module synchronizes customs inspection rules and customs declaration information in real time through a multimodal extraction algorithm, and generates a timestamp-aligned audit rule update table and a customs declaration business information table; the compliance audit module constructs an enhanced audit space, integrating the triple architecture of the horizontal business audit node network, the vertical business transmission node chain and the declaration node network, wherein the horizontal business audit node network establishes the topological association relationship of the business audit nodes based on the graph algorithm, and the vertical business transmission node chain adopts distributed reinforcement learning and encryption algorithm to construct the customs declaration process reinforcement. The three work together to combine dynamic rule priority indexing and reinforcement learning strategy to complete multi-dimensional compliance verification; the evaluation and optimization module calculates the business transmission risk probability, audit delay probability and error probability in real time through a two-way risk assessment model, and feeds back to the audit space to optimize the rule weights and process paths; the present invention breaks through the traditional single-link audit mode through the composite audit architecture of "horizontal topology-vertical transmission-declaration clustering", realizes dynamic rule adaptation, accurate risk tracing and intelligent process optimization, and effectively solves the problems of inefficient multi-source heterogeneous data integration and lagging audit of complex business scenarios in cross-border trade.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A customs compliance intelligent audit system, comprising: a data collection module, a compliance audit module, and an evaluation and optimization module;

[0006] The data collection module extracts the review rule update table and the customs declaration business information table in real time based on the review rules and business declaration information collected in real time in combination with the multimodal extraction model;

[0007] In the compliance audit module, the user obtains the audited customs declaration business information table based on the customs declaration classification information table through the preset enhanced audit space combined with the audit rule update table; the enhanced audit space includes a declaration node network, a vertical business transmission node chain and a horizontal business audit node network; the enhanced audit space is obtained by training the business information of the horizontal business audit node network and the vertical business transmission node chain, the preset audit rule index and the priority corresponding to the rules and business audits in combination with the reinforcement learning algorithm; the vertical business transmission node chain is constructed by combining each declaration node and the corresponding business audit node and the customs declaration business information transmission path with the distributed reinforcement learning algorithm and the encryption algorithm; the horizontal business audit node network is constructed by combining all business audit nodes with the correlation strength between business audit nodes through a graph algorithm; the declaration node network is constructed by combining all business declaration nodes with the similarity of the corresponding business declaration types; the declaration node network is connected to the horizontal business audit node network through the vertical business transmission node chain;

[0008] The evaluation optimization module, based on a preset two-way risk assessment model and combined with the real-time business audit information corresponding to the horizontal business audit node network and the vertical business transmission node chain, obtains the business audit transmission risk probability, audit delay probability and audit error probability and feeds back to the enhanced audit space to adjust the business transmission and audit process.

[0009] Specifically, the data acquisition module includes a data acquisition unit, a multimodal extraction unit, and an enhanced update unit;

[0010] The data collection unit is used to collect in real time the business declaration information uploaded by all business declaration nodes, the business declaration information that has not been completed in the review process, and the updated business review rule information, and classify the collected data information according to the data type;

[0011] The multimodal extraction unit obtains a customs declaration business information table and an audit rule update table based on the classified real-time collected information in combination with a multimodal feature extraction algorithm, and performs a secondary audit update of the unfinished declaration business based on the effective timestamp of the audit rules in the audit rule update table;

[0012] The enhanced update unit is used to update the audit rule information corresponding to each node in the declaration node network and the horizontal business audit node network according to the audit rule update table combined with the preset enhanced update model, and use the updated audit rules to conduct a secondary review and update of the unfinished declaration business review.

[0013] Specifically, the compliance review module includes a declaration clustering unit, a primary review unit, a transmission unit, and a secondary review unit;

[0014] The declaration clustering unit obtains a similar declaration node set based on the business information corresponding to all business declaration nodes in combination with a business similarity clustering model, and uses the cluster center corresponding to each similar declaration node subset in the similar declaration node set as the initial review node;

[0015] The initial review unit obtains the customs declaration business information table after initial review based on the first enhanced review model configured for each initial review node, combined with the customs declaration business information table in the corresponding similar declaration node subset and the corresponding review rules;

[0016] The transmission unit transmits and distributes the initially reviewed customs declaration business information table to the corresponding business review node through a transmission path optimization algorithm combined with a distributed enhanced allocation algorithm based on the reviewed customs declaration business information table corresponding to each initial review node, a preset encryption code, load information of the transmission channel corresponding to each vertical business transmission node chain, review business type information and load information corresponding to all the business review nodes, and initial review evaluation result information corresponding to each declaration node;

[0017] The secondary review unit obtains the customs declaration business information table after the secondary review is completed based on the priority of the review business corresponding to each business review node, the priority of the sub-business review process corresponding to each business, and the index information between the review rules and each sub-business review process, combined with the second enhanced review model configured for each business review node.

[0018] Specifically, the evaluation and optimization module includes a first evaluation unit, a second evaluation unit, and an evaluation feedback unit;

[0019] The first evaluation unit is configured to perform real-time evaluation using a first evaluation sub-model based on the clustering process corresponding to the declared clustering unit, the initial review process corresponding to the initial review unit, and the supply-demand relationship stability coefficient to obtain a first review risk probability; the first review risk probability is constructed from the clustering error probability and the initial false detection and missed detection risk probabilities; the supply-demand relationship stability coefficient is constructed from the ratio of the corresponding transaction completion frequency of the supply and demand parties to the total transaction frequency;

[0020] The second evaluation unit is configured to perform real-time evaluation using a second evaluation sub-model based on the data transmission process corresponding to the transmission unit, the secondary audit process corresponding to the secondary audit unit, and the audit anomaly coefficient to obtain a second audit risk probability; the second audit risk probability is constructed by combining the tampering risk probability corresponding to the data transmission process, the business allocation risk probability, and the risk probabilities of false detection, missed detection, and audit delay corresponding to the secondary audit with a Bayesian algorithm;

[0021] The audit anomaly coefficient is constructed by the abnormal frequency of historical declarations of the corresponding declaring user and the total declaration frequency;

[0022] The evaluation feedback unit optimizes the first enhanced audit model, the second enhanced audit model and the encrypted coding information based on the risk probability difference corresponding to the first audit risk probability and the second audit risk probability, combined with a preset multi-level risk discrimination threshold.

[0023] Specifically, the workflow of the evaluation and feedback unit includes:

[0024] Preset first audit judgment threshold , First audit difference discrimination threshold and the second audit difference discrimination threshold ,and , when the first audit risk probability is less than , the customs declaration business information table after the initial review is transmitted and assigned to the corresponding business review node. Otherwise, the first review risk probability is fed back to the first enhanced review model to update and optimize the clustering process and the initial review process until the first review risk probability is less than until;

[0025] When the risk probability difference is less than When the risk probability difference is greater than or equal to and less than When the second audit risk probability is fed back to the transmission unit and the second enhanced audit model, the data transmission process and the secondary audit process are optimized and updated until the risk probability difference is less than Until the time;

[0026] When the risk probability difference is greater than or equal to When the transmission encryption is abnormal, the business information corresponding to the business audit node is matched and compared with the business information of the corresponding declaration node to obtain comparison difference information. At the same time, the encryption coding level in the transmission unit and the business information for determining encryption abnormality are adjusted using the obtained comparison difference information, and the adjusted business information is updated and reviewed for the second time until the risk probability difference is less than Until then.

[0027] Specifically, the process of obtaining the customs declaration business information form after initial review includes:

[0028] Each business application user is regarded as a business application node. At the same time, based on the application user's business type information and the corresponding target customer information, a similarity algorithm is used to obtain the business similarity corresponding to different application users at the same time stamp;

[0029] Based on the similarity between the business declaration nodes and the corresponding businesses, a declaration node network is constructed through a graph algorithm;

[0030] Based on the declaration node network, the business declaration information of different declaration users at the same time stamp is collected in real time, and the collected business declaration information is input into the multimodal extraction model to obtain the customs declaration business information table corresponding to different declaration users;

[0031] The customs declaration business information table includes a unique identifier, business type, and structured business process information after feature extraction;

[0032] Based on the business information and business similarity corresponding to the declaration node network, clustering is performed through clustering algorithm combined with time series algorithm to obtain a subset of similar declaration nodes corresponding to each similar business type, and the cluster center node of the similar declaration node subset is used as the initial review node of the corresponding similar declaration node subset.

[0033] Specifically, the process of obtaining the customs declaration business information form after the initial review also includes:

[0034] A first enhanced audit model is constructed based on the first enhanced audit sub-model configured for each declaration node in each similar declaration node subset and the first evaluation sub-model configured for the corresponding initial audit node through a federated algorithm framework;

[0035] Based on the customs declaration business information table corresponding to each declaration node, the corresponding first enhanced review sub-model is used to review it, and the review result information is transmitted to the corresponding initial review node through the federation framework. The first evaluation sub-model is used to conduct review and evaluation. If the evaluation passes, the customs declaration business information table with the initial review completed is output;

[0036] If the review fails, the assessment results will be fed back to the corresponding declaration node through the federal framework for adjustment until the first review judgment threshold is met, and the adjusted customs declaration business information table of the initial review completion is obtained.

[0037] Specifically, the process of obtaining the customs declaration business information form after the second review includes:

[0038] Each auditing business type department of the auditing party is regarded as a business audit node, and the audit process of the corresponding business type under each business audit node and the audit priority corresponding to each audit sub-node in the audit process under the current business type audit process are used to construct a business audit process node chain corresponding to each business audit node through the blockchain algorithm; each audit sub-node corresponds one-to-one to each audit sub-process in the corresponding type of business audit process;

[0039] Based on the business review process node chain information corresponding to each business review node, the business overlap correlation degree corresponding to each business review node is obtained through the association algorithm;

[0040] Based on each business review node and the corresponding business overlapping correlation, a horizontal business review node network is constructed through graph algorithm.

[0041] Specifically, the process of obtaining the customs declaration business information form after the second review also includes:

[0042] The horizontal business review node network obtains the review result corresponding to each review sub-node under each business review node and the comprehensive result of the corresponding business based on the customs declaration business information table of the initial review completed and the corresponding review priority and initial review evaluation result assigned by the vertical business transmission node chain, and the review rule index information corresponding to the corresponding review sub-process stored in each review sub-node under each business review node, through the second enhanced review sub-model configured for each review sub-node;

[0043] Based on the comprehensive results of each business review node and the initial review results of the corresponding business, combined with the first review difference discrimination threshold, the second review difference discrimination threshold and the second evaluation sub-model, a secondary review judgment is performed. If the judgment passes, a customs declaration business information table with the secondary review completed is obtained;

[0044] Otherwise, the corresponding business, initial review process, transfer process and secondary review process will be adjusted in real time according to the secondary review results until all review sub-processes of each review business meet the review rules;

[0045] The second enhanced audit model is constructed by the second enhanced audit sub-model configured for each audit sub-node and the corresponding second evaluation sub-model through a federated framework.

[0046] A customs compliance intelligent audit method, comprising:

[0047] Based on the real-time collection of review rules and customs declaration information combined with a multimodal extraction algorithm, the review rule update table and customs declaration business information table are obtained;

[0048] According to the customs declaration classification information table, through the preset enhanced audit space combined with the audit rule update table, the audited customs declaration business information table is obtained;

[0049] The enhanced audit space includes a horizontal business audit node network and a vertical business transmission node chain; the enhanced audit space is obtained by training the business information of the horizontal business audit node network and the vertical business transmission node chain, the preset audit rule index and the priority corresponding to the rules and business audits in combination with the reinforcement learning algorithm; the vertical business transmission node chain is constructed by combining each declaration node and the corresponding business audit node and the customs declaration business information transmission path with the distributed reinforcement learning algorithm and the encryption algorithm; the horizontal business audit node network is constructed by combining all business audit nodes with the correlation strength between business audit nodes through a graph algorithm; the declaration node network is constructed by combining all business declaration nodes with the similarity of the corresponding business declaration types; the declaration node network is connected to the horizontal business audit node network through the vertical business transmission node chain;

[0050] Based on the preset two-way risk assessment model and the real-time business audit information corresponding to the horizontal business audit node network and the vertical business transmission node chain, the business audit transmission risk probability, audit delay probability and audit error probability are obtained and fed back to the enhanced audit space to adjust the business transmission and audit process.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] The present invention addresses the shortcomings of the existing technology and builds a dynamic and adaptive customs audit system through the deep collaboration of multi-source data fusion and intelligent architecture. The multimodal extraction model of the data acquisition module integrates the review rules and customs declaration information in real time to ensure the accurate alignment of the rules with the business data, and drives the dynamic iteration of the audit rules through the enhanced update mechanism. The compliance audit module integrates the topological architecture of the declaration node network, the vertical business transmission node chain and the horizontal business audit node network. Based on the federated learning framework and distributed reinforcement learning algorithm, it realizes clustering audit driven by business similarity, encryption transmission path optimization and collaborative decision-making of associated audit nodes. The horizontal business audit node network uses a graph algorithm to model the business correlation strength between audit nodes, and the vertical business transmission node chain combines load-aware strategy to dynamically allocate audit tasks. The bidirectional risk assessment model of the evaluation and optimization module analyzes transmission risk, delay and error probability in real time, and through a multi-level feedback mechanism, it links encryption strategy adjustment, enhanced model parameter optimization and audit process reconstruction to form a closed-loop control. The present invention has the capabilities of real-time rule adaptation, cross-modal conflict resolution and risk stratification traceability, while ensuring the efficient collaboration of audit tasks and the security and reliability of the entire data link, achieving a systematic leap in audit efficiency, accuracy and security. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a module diagram of a customs compliance intelligent audit system according to the present invention;

[0054] Figure 2 This is a flow chart of a unit of the customs compliance intelligent audit system of the present invention;

[0055] Figure 3 This is a flow chart of a customs compliance intelligent audit method according to the present invention. DETAILED DESCRIPTION

[0056] Example 1:

[0057] See also Figure 1 , an embodiment provided by the present invention: a customs compliance intelligent audit system, comprising: a data collection module, a compliance audit module, and an evaluation and optimization module;

[0058] The data collection module, based on the real-time collection of review rules and business declaration information combined with a multimodal extraction model, extracts the review rule update table and the customs declaration business information table in real time;

[0059] In the compliance audit module, users can obtain the audited customs declaration business information table based on the customs declaration classification information table through the preset enhanced audit space combined with the audit rule update table; the enhanced audit space includes the declaration node network, the vertical business transmission node chain and the horizontal business review node network; the enhanced audit space is constructed and trained by combining the business information of the horizontal business review node network and the vertical business transmission node chain, the preset audit rule index and the priority corresponding to the rules and business audits in combination with the reinforcement learning algorithm; the vertical business transmission node chain is constructed by combining each declaration node and the corresponding business review node and the customs declaration business information transmission path with the distributed reinforcement learning algorithm and the encryption algorithm; the horizontal business review node network is constructed by combining all business review nodes with the correlation strength between business review nodes through the graph algorithm; the declaration node network is constructed by combining all business declaration nodes with the similarity of the corresponding business declaration types; the declaration node network is connected to the horizontal business review node network through the vertical business transmission node chain;

[0060] The assessment and optimization module, based on the preset two-way risk assessment model and combined with the real-time business audit information corresponding to the horizontal business audit node network and the vertical business transmission node chain, obtains the business audit transmission risk probability, audit delay probability and audit error probability and feeds back to the enhanced audit space to adjust the business transmission and audit processes.

[0061] In this example, see Figure 2 ,The data acquisition module includes a data acquisition unit, a multimodal extraction unit, and an ,enhanced updating unit;

[0062] The data collection unit is used to collect in real time the business declaration information uploaded by all business declaration nodes, the business declaration information that has not been completed in the review process, and the updated business review rule information, and classify the collected data information according to the data type;

[0063] Furthermore, in this embodiment, the business declaration information specifically includes but is not limited to unique identification information, such as declaration form ID and user ID; commodity details, such as HS code and value; logistics information, such as transportation method and bill of lading number; transaction party information, such as importer and exporter information; and compliance certification documents, such as certificate of origin and quality inspection report;

[0064] Information on declarations that have not been completed during the review process includes, but is not limited to, recording process status, such as pending preliminary review; reasons for suspension, such as missing data or rule conflicts; and operation logs, such as rejection records and user re-upload actions;

[0065] Update audit rule information, including but not limited to rule attributes such as ID and effective time limit; logical descriptions such as conditional expressions and associated fields; dynamic parameters such as exchange rate tables and risk lists; and original audit rule documentation and explanations;

[0066] Data types include but are not limited to structured fields, unstructured files, and semi-structured logs, supporting full data storage and analysis.

[0067] The system ensures the compliance of customs declaration business, audit efficiency and data traceability through standardized definitions and dynamic rule management, and adapts to complex needs such as customs supervision and international trade.

[0068] The multimodal extraction unit obtains the customs declaration business information table and the audit rule update table based on the classified real-time collected information combined with the multimodal feature extraction algorithm, and performs a secondary review and update of the unfinished declaration business review based on the effective timestamp of the audit rules in the audit rule update table.

[0069] Furthermore, in this embodiment, the detailed process of obtaining the customs declaration business information table and the review rule update table includes:

[0070] First, the application node receives structured, semi-structured, and unstructured data in real time, and uses regular expression verification, OCR text extraction, image recognition models, and JSON parsing technology to classify, clean, and standardize the raw data.

[0071] It's important to note that in this example, structured data is verified for formatting rules and currency conversion to ensure field integrity and consistency. Unstructured data uses an OCR engine to extract text features, combined with a YOLOv8 model to identify danger signs in images and convert them into structured labels. The conditional expressions for semi-structured data parsing rules are executable triple logic. This stage outputs standardized multi-type data, providing high-quality input for subsequent processing. Key implementation details include review of abnormal data labeling, an OCR retry mechanism, and image recognition confidence threshold control. Formatting rule verification is preferably performed on the declaration form ID format.

[0072] Second, based on predefined field templates, the cleaned structured data is directly mapped to the customs declaration business information table. Key information is extracted from unstructured documents using OCR and NLP technologies. A ResNet model is used to identify brand trademarks in product images and link them to the intellectual property database.

[0073] Furthermore, in this embodiment, the semi-structured rule data is decomposed into triple logical conditions and dynamically associated with the external parameter interface; this step realizes the feature fusion and semantic unification of multi-source data, and outputs a standardized feature set. Attention should be paid to text semantic standardization, sensitive vocabulary matching and safe escape processing in rule parsing.

[0074] Third, the declaration form ID is used as the primary key to integrate multimodal features, ensuring data credibility through integrity checks and logical verification. It should be noted that in this embodiment, the cargo value verification uses asynchronous calculations to avoid blocking, and status conflicts trigger an automatic freezing process, outputting a complete and consistent customs declaration business information table to support subsequent review rule matching;

[0075] Fourth, parse rule attributes and design a table structure to store logical conditions and dynamic parameters. Synchronize external parameters and rule status management through scheduled tasks. Furthermore, in this embodiment, when outputting a dynamically effective audit rule table, attention should be paid to UTC timestamp storage, soft delete marking, and the abnormal circuit breaker mechanism of the multi-source parameter interface. Rule attributes include but are not limited to ID, type, and effective time; rule status management includes but is not limited to marking expired / pending rules.

[0076] Fifth, incomplete declarations are filtered based on the rule effective time and field relevance, rule verification is re-executed, and status is updated. Conflicts trigger manual review and audit log recording. It should be noted that the updated business status and audit trail are output, including time window matching, snapshot comparison analysis, and two-way conflict details push.

[0077] Sixth, business tables and rule tables are persistently stored in relational databases, and composite indexes are established to optimize query performance; message queues push rule change notifications and correction reminders, and priority channels are set to ensure emergency notifications; it should be noted that this embodiment uses blockchain to store key data change records, supports multi-institutional consensus audits, and simultaneously outputs queryable data tables, real-time notifications, and tamper-proof audit links, focusing on index design, message priority grading, and blockchain node disaster recovery deployment.

[0078] The enhanced update unit is used to update the audit rule information corresponding to each node in the declaration node network and the horizontal business review node network according to the audit rule update table combined with the preset enhanced update model, and use the updated audit rules to conduct a secondary review and update of the unfinished declaration business review.

[0079] In this embodiment, the compliance review module includes a declaration clustering unit, a primary review unit, a transmission unit, and a secondary review unit;

[0080] The declaration clustering unit obtains a set of similar declaration nodes based on the business information corresponding to all business declaration nodes and the business similarity clustering model, and uses the cluster center corresponding to each similar declaration node subset in the similar declaration node set as the initial review node;

[0081] The initial review unit obtains the customs declaration business information table after initial review based on the first enhanced review model configured for each initial review node and the customs declaration business information table in the corresponding similar declaration node subset and the corresponding review rules;

[0082] Furthermore, the process of obtaining the customs declaration business information form after the initial review includes:

[0083] Each business application user is regarded as a business application node. At the same time, based on the application user's business type information and the corresponding target customer information, a similarity algorithm is used to obtain the business similarity corresponding to different application users at the same time stamp;

[0084] Based on the similarity between the business declaration nodes and the corresponding businesses, a declaration node network is constructed through a graph algorithm;

[0085] Based on the declaration node network, the business declaration information of different declaration users at the same time stamp is collected in real time, and the collected business declaration information is input into the multimodal extraction model to obtain the customs declaration business information table corresponding to different declaration users;

[0086] The customs declaration business information table includes unique identification, business type, and structured business process information after feature extraction;

[0087] Based on the business information and business similarity corresponding to the declaration node network, clustering is performed through clustering algorithm combined with time series algorithm to obtain a subset of similar declaration nodes corresponding to each similar business type, and the cluster center node of the similar declaration node subset is used as the initial review node of the corresponding similar declaration node subset.

[0088] A first enhanced audit model is constructed based on the first enhanced audit sub-model configured for each declaration node in each similar declaration node subset and the first evaluation sub-model configured for the corresponding initial audit node through a federated algorithm framework;

[0089] Based on the customs declaration business information table corresponding to each declaration node, the corresponding first enhanced review sub-model is used to review it, and the review result information is transmitted to the corresponding initial review node through the federation framework. The first evaluation sub-model is used to conduct review and evaluation. If the evaluation passes, the customs declaration business information table with the initial review completed is output;

[0090] If the review fails, the assessment results will be fed back to the corresponding declaration node through the federal framework for adjustment until the first review judgment threshold is met, and the adjusted customs declaration business information table of the initial review completion is obtained.

[0091] For example, the process of obtaining the customs declaration business information form after the initial review needs further explanation:

[0092] First, a hybrid similarity algorithm is used to quantify the business relevance of the declared nodes and construct a node network diagram. It should be noted that this embodiment generates a similarity matrix by comprehensively calculating the structured business type, target country, and product text features, combined with adjacency table storage optimization. The hybrid similarity algorithm in this embodiment is obtained through Jaccard classification feature matching, TF-IDF text vectorization, and time window weighting.

[0093] Second, based on the DBSCAN algorithm combined with time-constrained clustering, highly similar nodes are divided into subsets, and nodes with high historical review pass rates and strong business consistency are selected as the initial review benchmark. By expanding time series features (declaration frequency, weekly trends) and adjusting dynamic density thresholds, the problem of traffic fluctuations during business peaks and valleys is resolved, and reliable subsets and initial nodes are output. The downgrade replacement mechanism of cluster centers and the adaptive expansion of neighborhood ranges are used to avoid the misaggregation of low-quality data; in this embodiment, the downgrade replacement mechanism is implemented by selecting suboptimal similar nodes with a priority pass rate greater than the preset pass threshold;

[0094] It should be further explained that the reason why this embodiment selects the suboptimal similar nodes with a priority pass rate greater than the preset pass threshold to implement the downgrade replacement mechanism is because of the core objectives of the audit business and the data quality requirements. In the customs declaration audit scenario, the historical audit pass rate directly reflects the accuracy and reliability of the node's processing business and is a key indicator for ensuring audit quality. If only similarity is relied upon, nodes with similar business characteristics but poor audit quality may be included, resulting in unreliable initial audit benchmarks and affecting subsequent audit results. Setting a pass threshold to filter suboptimal similar nodes can not only ensure the business relevance between nodes and maintain the effectiveness of clustering, but also ensure that the selected nodes have high audit capabilities, improve the quality of the audit benchmark from the source, reduce the risk of audit errors caused by node quality problems, and ensure the efficiency and accuracy of the entire audit process.

[0095] Third, under the federated framework, the reporting node deploys a local sub-model to perform rule verification, and the initial review node aggregates the global parameter update model to achieve collaborative optimization under privacy protection. It should be noted that the global parameter update model of this embodiment uses homomorphic encryption to transmit gradient parameters, combined with rule coverage monitoring to trigger incremental training, to ensure the real-time performance of the model and the completeness of the rules. Specific parameters include synchronous frequency control, secure multi-party computing protocol and uncovered rule warning, balancing efficiency and security.

[0096] Fourth, the local sub-model performs rule matching and data integrity verification. The assessment model generates audit recommendations by calculating risk levels and comparing them with historical consistency. This also triggers automated re-transmission guidance and manual review mechanisms, forming a closed loop of "initial review-feedback-re-review." Dynamic loading of mandatory fields should be strengthened, with verification rules and risk scoring algorithms switching based on business type to reduce missed and false positives. Furthermore, in this embodiment, rule matching is preferably based on HS code tax rate mapping, and data integrity verification is preferably based on mandatory field detection.

[0097] The transmission unit transmits and distributes the customs declaration business information table after the initial review to the corresponding business review node through a transmission path optimization algorithm combined with a distributed reinforcement allocation algorithm based on the customs declaration business information table after the initial review corresponding to each initial review node, a preset encryption code, the load information of the transmission channel corresponding to each vertical business transmission node chain, the review business type information and load information corresponding to all business review nodes, and the initial review assessment result information corresponding to each declaration node;

[0098] Furthermore, in this embodiment, the detailed implementation process of transmitting and allocating the customs declaration business information table after initial review to the corresponding business review node includes:

[0099] First, a dynamic encryption level matching algorithm is used to adaptively select encryption strength based on product sensitivity, and a real-time channel load quantification algorithm is used to evaluate the transmission link status. In this embodiment, the dynamic encryption level matching algorithm is preferably AES-128 / 256 and blockchain-based evidence storage. The load quantification algorithm is preferably constructed using a random forest pre-trained on bandwidth, latency, and error rate.

[0100] Second, based on the audit node capability matrix modeling and the Jaccard similarity algorithm to calculate the node matching degree, a transmission task graph model is constructed. In this embodiment, the audit node capability matrix modeling is constructed based on the audit type, load, and accuracy rate. The transmission task graph model is constructed based on the improved Dijkstra algorithm combined with the path cost.

[0101] Third, distributed reinforcement learning is used to dynamically optimize node and channel combinations to balance matching and load.

[0102] Fourth, data packet compression, SHA-256 hash checksum, and blockchain evidence storage technology are used to ensure transmission integrity, and an exception handling mechanism is used to trigger the selection of an alternative path.

[0103] Fifth, rely on real-time monitoring panels and dynamic parameter adjustment technology to iteratively optimize priority formulas and routing strategies.

[0104] This process significantly enhances the overall intelligence of the cross-border trade audit system by building a multimodal data fusion framework and a dynamic rule coordination mechanism. At the data governance level, a hybrid processing model combining structured feature mapping and unstructured semantic parsing achieves decoupling and reorganization of the multidimensional features of declaration information, effectively addressing the fragmented audit elements caused by data silos in traditional customs declaration systems. Through cross-modal correlation verification of HS codes and logistics documents, a spatiotemporal consistency verification chain is established between commodity attributes and logistics trajectories, enhancing the ability to identify anomalous declaration patterns. A timestamp-driven mechanism for audit rule update tables, combined with dynamic matching of rule effectiveness with the status of ongoing businesses, establishes a causal relationship model between rule versions and the audit process. This allows historical declaration data to be automatically retroactively re-evaluated as audit rules change, eliminating the risk of audit logic gaps caused by regulatory rule iterations. At the node coordination level, a federated learning framework based on business similarity maps heterogeneous declaration nodes into homogeneous business clusters through feature space projection. Secure aggregation of model parameters within clusters enables audit knowledge sharing, transcending the limitations of single-node samples while ensuring data privacy, and enhancing the model's generalization to regional trade characteristics. The dynamic transmission path optimization algorithm uses adaptive encryption level matching and real-time network topology awareness to construct transmission strategies under multi-dimensional constraints. Differentiated encryption strategies are implemented based on product sensitivity grading, ensuring the security of core data while avoiding resource waste caused by over-encryption. By combining the channel load index with node processing capacity, a balance function is established between transmission cost and audit efficiency, enabling flexible resource scheduling in high-concurrency scenarios. Deep integration of blockchain evidence storage technology creates an irreversible evidence chain linking rule file hash values ​​and key audit operations. A smart contract-driven automatic verification mechanism ensures the auditability and non-repudiation of cross-node data interactions. Regarding exception handling, an intelligent routing switching mechanism and a rapid backup node election strategy establish a fault-tolerant recovery system for transmission failures. Combined with real-time visual feedback from the transmission quality monitoring panel, this forms an iterative closed-loop optimization path decision mechanism. A time series clustering algorithm incorporates historical reporting frequency and periodic fluctuations to enhance the temporal correlation of business similarity calculations, enabling node clustering results to reflect both common business attributes and the dynamic patterns of reporting behavior. Homomorphic encryption of audit model parameters and a distributed gradient aggregation mechanism achieve a precise balance between data availability and privacy during federated learning, establishing a trusted computing environment for multi-agent collaborative audits. A dynamic priority calculation model quantifies business attributes such as risk level and correction history into transmission weight factors. Using a weighted decision tree, it allocates transmission resources to key businesses, ensuring the timely delivery of high-priority applications.The systematic integration of these technical means enables the customs declaration and review system to form a positive reinforcement loop in data collection, rule adaptation, node collaboration, secure transmission and other links, and build an intelligent review ecosystem with elastic expansion capabilities and continuous evolution characteristics, providing full-link technical support for compliance supervision in complex trade environments.

[0105] The secondary review unit obtains the customs declaration business information table after the secondary review is completed based on the priority of the review business corresponding to each business review node, the priority of the sub-business review process corresponding to each business, and the index information between the review rules and each sub-business review process, combined with the second enhanced review model configured for each business review node.

[0106] In this embodiment, the evaluation and optimization module includes a first evaluation unit, a second evaluation unit, and an evaluation feedback unit;

[0107] The first evaluation unit is configured to perform real-time evaluation using the first evaluation sub-model based on the clustering process corresponding to the declared clustering unit, the initial review process corresponding to the initial review unit, and the supply-demand relationship stability coefficient, to obtain a first review risk probability. The first review risk probability is constructed from the clustering error probability and the initial false detection and missed detection risk probabilities. The supply-demand relationship stability coefficient is constructed from the ratio of the corresponding transaction completion frequency of both the supply and demand parties to the total transaction frequency.

[0108] The second evaluation unit is used to perform real-time evaluation using a second evaluation sub-model based on the data transmission process corresponding to the transmission unit, the secondary audit process corresponding to the secondary audit unit, and the audit anomaly coefficient to obtain a second audit risk probability. The second audit risk probability is constructed by combining the tampering risk probability corresponding to the data transmission process, the business allocation risk probability, and the risk probabilities of false detection, missed detection, and audit delay corresponding to the secondary audit with a Bayesian algorithm. It should be further explained that a specific implementation method of the second audit risk probability is:

[0109] The risk probability of data transmission tampering is obtained based on blockchain evidence storage technology and the SHA-256 hash check algorithm, and is used as the prior input of the Bayesian fusion model.

[0110] Based on the LSTM-autoencoder time series anomaly detection algorithm, the transmission data feature deviation score is obtained as a conditional variable for tampering risk;

[0111] Based on the Jaccard similarity algorithm and the audit node capability matrix modeling, the business allocation risk probability is obtained and input into the Bayesian model;

[0112] Based on the historical confusion matrix of the second enhanced audit model and the time decay algorithm, the risk probability of false detection and missed detection is obtained as a priori input;

[0113] The audit delay risk probability is obtained based on the Weibull distribution model and maximum likelihood estimation algorithm, and integrated into Bayesian fusion;

[0114] The audit anomaly coefficient is obtained based on the time series clustering algorithm (K-Means, K=5) and the BERT pre-training model as a Bayesian conditional variable;

[0115] The Metropolis-Hastings algorithm based on Markov Chain Monte Carlo (MCMC) performs posterior sampling on the Bayesian model and iterates 10,000 times to obtain the integrated second-audit risk probability;

[0116] The channel load score is output based on the random forest algorithm, and the audit priority score is obtained based on the weighted decision tree algorithm. The product of the two is used as the dynamic weight adjustment factor;

[0117] The risk probability is converted into a visual heat map based on the HSL color space mapping algorithm and pushed to the monitoring panel based on the WebSocket protocol;

[0118] When the risk probability exceeds the preset threshold, the backup node switching is triggered based on the intelligent routing algorithm, forming a complete closed loop of data collection, risk quantification, model fusion, and dynamic prevention and control.

[0119] The audit anomaly coefficient is constructed by the abnormal frequency of historical declarations of the corresponding declaring user and the total declaration frequency;

[0120] The evaluation feedback unit optimizes the hierarchical feedback of the first enhanced audit model, the second enhanced audit model and the encrypted coding information based on the risk probability difference corresponding to the first audit risk probability and the second audit risk probability, combined with the preset multi-level risk discrimination threshold.

[0121] Furthermore, the workflow of the evaluation feedback unit includes:

[0122] Preset first audit judgment threshold , First audit difference judgment threshold and the second audit difference discrimination threshold ,and , when the first audit risk probability is less than , the customs declaration business information table after the initial review is transferred and assigned to the corresponding business review node. Otherwise, the first review risk probability is fed back to the first enhanced review model to update and optimize the clustering process and the initial review process until the first review risk probability is less than until;

[0123] When the risk probability difference is less than When the risk probability difference is greater than or equal to and less than When the second audit risk probability is fed back to the transmission unit and the second enhanced audit model, the data transmission process and the secondary audit process are optimized and updated until the risk probability difference is less than Until the time;

[0124] When the risk probability difference is greater than or equal to When the transmission encryption is abnormal, the business information corresponding to the business review node is matched and compared with the business information of the corresponding declaration node to obtain the comparison difference information. At the same time, the encryption coding level in the transmission unit and the business information for determining the encryption abnormality are adjusted using the obtained comparison difference information. The adjusted business information is then updated and reviewed for the second time until the risk probability difference is less than Until then.

[0125] Specifically, the process of obtaining the customs declaration business information form after the second review includes:

[0126] Each auditing business type department of the auditing party is regarded as a business audit node, and the audit process of the corresponding business type under each business audit node and the audit priority corresponding to each audit sub-node in the audit process under the current business type audit process are used to construct a business audit process node chain corresponding to each business audit node through the blockchain algorithm; each audit sub-node corresponds one-to-one to each audit sub-process in the corresponding type of business audit process;

[0127] Based on the business review process node chain information corresponding to each business review node, the business overlap correlation degree corresponding to each business review node is obtained through the association algorithm;

[0128] Based on each business review node and the corresponding business overlapping correlation, a horizontal business review node network is constructed through graph algorithm.

[0129] The horizontal business review node network obtains the review results and the comprehensive results of the corresponding business of each review sub-node under each business review node based on the customs declaration business information table of the initial review completed and the corresponding review priority and initial review evaluation results assigned by the vertical business transmission node chain, and the review rule index information corresponding to the corresponding review sub-process stored in each review sub-node under each business review node, through the second enhanced review sub-model configured for each review sub-node;

[0130] Based on the comprehensive results of each business review node and the initial review results of the corresponding business, combined with the first review difference discrimination threshold, the second review difference discrimination threshold and the second evaluation sub-model, a secondary review judgment is performed. If the judgment passes, a customs declaration business information table with the secondary review completed is obtained;

[0131] Otherwise, the corresponding business, initial review process, transfer process and secondary review process will be adjusted in real time according to the secondary review results until all review sub-processes of each review business meet the review rules;

[0132] The second enhanced audit model is constructed by the second enhanced audit sub-model configured for each audit sub-node and the corresponding second evaluation sub-model through a federated framework.

[0133] Furthermore, in this embodiment, the process of obtaining the customs declaration business information table that has completed the second review includes:

[0134] First, based on the business audit process node chain, a recursive decomposition algorithm is used to decompose complex business processes step by step into the smallest executable atomic sub-processes, with each sub-process as an independent unit. On this basis, blockchain evidence storage technology is used to connect the sub-processes into a chain through hash pointers. The key data of each sub-process is generated with a unique hash value using the SHA-256 hash algorithm, pointing to the hash value of the previous sub-process, forming an unalterable node relationship chain to ensure that the audit process is traceable and the data is authentic and reliable. Through recursive decomposition and hash chain construction, complex business processes are transformed into traceable, tamper-proof atomic sub-process chains, providing a foundation for subsequent audit collaboration. Key data includes but is not limited to operation records, timestamps, etc.

[0135] Second, based on the business review rule fields, the Jaccard similarity algorithm is used to calculate the overlapping correlation between rules. Specifically, the quantitative similarity value is obtained by statistically analyzing the ratio of the number of elements in the intersection and union of the rule fields. Rules with similarity values ​​higher than the preset threshold are regarded as association rules. Through the edge connection operation in graph theory, associated edges are established for them, and a horizontal business review node network is gradually constructed to achieve the correlation integration of review rules. Business review rule fields include but are not limited to HS code, place of origin, etc.

[0136] Third, based on graph algorithms, we identify highly correlated nodes in the horizontal business review node network and connect them to form a subgraph. Using a dynamic routing configuration algorithm, combined with real-time load data from review subnodes, we dynamically assign review tasks to idle or lightly loaded nodes, achieving load balancing across business reviews.

[0137] At the same time, a failover mechanism is set up. When a node goes down, the routing path is automatically switched through the backup path search algorithm to ensure the continuous execution of the audit task. This embodiment constructs a subgraph through a graph algorithm, combines dynamic routing with the set failover mechanism, and realizes dynamic balanced distribution and reliable execution of audit tasks according to the node load. The graph algorithm is preferably a depth-first search algorithm; the dynamic routing configuration algorithm is preferably an improved shortest path first algorithm; the real-time load data of the audit subnode includes but is not limited to the task backlog, processing rate, etc. The set failover mechanism is specifically: automatically switch the routing path when a node goes down;

[0138] Fourth, based on the federated learning algorithm, a local audit model is deployed on each audit sub-node, and each audit sub-node performs model training on local data. During the training process, the local gradient parameters are encrypted and aggregated using a secure aggregation protocol, and the encrypted parameters are uploaded to the central server. The central server uses a weighted average algorithm to fuse the parameters of each node, generate a global model, and feed it back to each sub-node to achieve dynamic model updates, thereby improving model audit capabilities while protecting data privacy. The preferred secure aggregation protocol is the SecureBoost algorithm, and the preferred weighted average algorithm is the FedAvg algorithm.

[0139] Fifth, based on a multimodal result aggregation algorithm, the audit results output by each audit sub-node are integrated and processed, and corresponding weights are assigned according to the credibility of different modal results. The comprehensive result is calculated and compared with the preset audit result threshold. When the difference exceeds the threshold, the manual review process is triggered; if it is within the threshold range, the process optimization is automatically promoted to achieve intelligent processing of the audit results. The multimodal result aggregation algorithm is preferably a weighted voting algorithm. The audit results include but are not limited to text conclusions, risk scores, image recognition results, etc.

[0140] Sixth, based on the two-way risk assessment model, the deviation probability is calculated by comprehensively considering the difference in the results of the initial review and the secondary review. According to the size of the deviation probability, the data encryption level and routing strategy are dynamically adjusted. When the deviation probability is high, the encryption level is increased to ensure data security, while the routing path is optimized, and more reliable nodes are selected for task transmission and review to reduce risks; the two-way risk assessment model is preferably a Bayesian risk assessment model; the dynamic adjustment of the data encryption level is achieved using the AES-128 or AES-256 algorithm; the routing strategy is to adjust the weight parameters of the shortest path algorithm according to the adjusted encryption level; this process is implemented by those skilled in the art based on existing technologies;

[0141] Seventh, a reinforcement learning algorithm is introduced, with audit efficiency and risk control as the objective functions, to establish a reward mechanism. The reinforcement learning algorithm learns the optimal load distribution and path selection strategies under different node loads and business scenarios through continuous trial and error. Based on the real-time node status and task attributes, the allocation plan and routing path are dynamically adjusted to achieve a balance between audit efficiency and risk control. Audit efficiency includes but is not limited to task processing time and throughput; risk control includes audit error rate and data leakage risk; node status includes but is not limited to load, audit capacity, and risk level; task attributes include but are not limited to urgency and complexity. This process is implemented by those skilled in the art based on existing technologies.

[0142] This process achieves full-link intelligent review from business decomposition to result output through node chain traceability assurance, horizontal business review node network construction, federated model iterative optimization and closed-loop feedback mechanism, thereby improving the accuracy, security and efficiency of customs declaration processing.

[0143] This process builds an intelligent customs declaration and audit system with elastic expansion and continuous optimization through the integration of multimodal audit processes and dynamic coordination mechanisms. The blockchain-based audit node chain technology chains the rule logic and execution path of the atomic audit sub-process, ensuring the immutability and full-link traceability of operations at each link, thereby enhancing the transparency and credibility of the audit process. The horizontal business audit node network achieves intelligent routing and load balancing of cross-departmental audit resources through business association quantification and graph network modeling. When the processing capacity of a single node is limited, tasks can be dynamically allocated to collaborative nodes based on association weights, effectively improving the audit throughput in high-concurrency scenarios. The sub-node model collaborative optimization mechanism driven by the federated learning framework achieves knowledge sharing of audit rule features and enhanced model generalization capabilities through encrypted parameter aggregation and distributed model updates, while protecting the data privacy of each node. For example, the ability to identify forgery patterns in cross-border certificates of origin can be migrated synchronously across nodes, significantly reducing the probability of missed detection of regional risks.

[0144] The dynamic priority-driven transmission encryption mechanism adaptively matches differentiated encryption strategies and transmission paths through multi-dimensional analysis of product attributes and risk levels, ensuring the security of sensitive data while avoiding resource waste caused by over-encryption of low-risk businesses. The bidirectional risk assessment model constructs a risk differential-driven hierarchical feedback system through probabilistic modeling of multi-source risk factors such as clustering errors, rule conflicts, and transmission leaks. When encryption anomalies or audit logic faults are detected, it automatically triggers transmission path rerouting, encryption level upgrades, and model parameter retraining, forming a complete control chain from risk perception to closed-loop optimization. The automated verification process enabled by smart contracts embeds audit rule logic into on-chain executable code, enabling seamless execution of rule matching and result determination, reducing response delays and operational deviations caused by human intervention. In addition, business overlap correlation analysis technology, through historical data mining and rule pattern matching, identifies the common characteristics and dependencies of cross-business audit processes, providing a logical basis for collaborative audits. For example, the optimization strategy of the origin verification process can be horizontally reused to related business nodes to improve the overall audit consistency; the multi-level discrimination threshold mechanism is combined with real-time risk assessment results to achieve dynamic resilience adjustment of the audit process, quickly release in low-risk scenarios, and filter medium and high-risk scenarios through multiple means such as model iteration, path switching, and manual review, achieving a precise balance between efficiency and security. This systematic interoperability design based on technical means enables the customs declaration audit system to adapt to changes in audit rules, business fluctuations and security threats, and can continuously optimize the customs compliance intelligent audit system.

[0145] Example 2:

[0146] See also Figure 3 Another embodiment provided by the present invention is a method for intelligent customs compliance audit, comprising:

[0147] Based on the real-time collection of review rules and customs declaration information combined with a multimodal extraction algorithm, the review rule update table and customs declaration business information table are obtained;

[0148] According to the customs declaration classification information table, through the preset enhanced audit space combined with the audit rule update table, the audited customs declaration business information table is obtained;

[0149] The enhanced audit space includes a horizontal business audit node network and a vertical business transmission node chain. The enhanced audit space is constructed and trained by combining the business information of the horizontal business audit node network and the vertical business transmission node chain, the preset audit rule index, and the priority corresponding to the rules and business audits, with a reinforcement learning algorithm. The vertical business transmission node chain is constructed by combining each declaration node with the corresponding business audit node and the customs declaration business information transmission path, with a distributed reinforcement learning algorithm and an encryption algorithm. The horizontal business audit node network is constructed by combining all business audit nodes with the strength of the association between business audit nodes through a graph algorithm. The declaration node network is constructed by combining all business declaration nodes with the similarity of the corresponding business declaration types. The declaration node network is connected to the horizontal business audit node network through the vertical business transmission node chain.

[0150] Based on the preset two-way risk assessment model and the real-time business audit information corresponding to the horizontal business audit node network and the vertical business transmission node chain, the business audit transmission risk probability, audit delay probability and audit error probability are obtained and fed back to the enhanced audit space to adjust the business transmission and audit processes.

[0151] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also change, modify, replace and modify the above-mentioned embodiments without departing from the scope of protection of the purpose of the present invention and the claims, and all of these are protected by the present invention.

Claims

1. A customs compliance intelligent audit system, characterized by: include: Data collection module, compliance review module, and evaluation and optimization module; The data collection module extracts the review rule update table and the customs declaration business information table in real time based on the review rules and business declaration information collected in real time in combination with the multimodal extraction model; In the compliance audit module, users obtain a completed customs declaration business information table based on the customs declaration classification information table through a preset enhanced audit space combined with an audit rule update table. The enhanced audit space includes a declaration node network, a vertical business transmission node chain, and a horizontal business audit node network. The enhanced audit space is constructed and trained using business information from the horizontal business audit node network and the vertical business transmission node chain, a preset audit rule index, and the priorities corresponding to the rules and business audits, combined with a reinforcement learning algorithm. The evaluation optimization module, based on a preset two-way risk assessment model and combined with the real-time business audit information corresponding to the horizontal business audit node network and the vertical business transmission node chain, obtains the business audit transmission risk probability, audit delay probability and audit error probability and feeds back to the enhanced audit space to adjust the business transmission and audit process.

2. The intelligent customs compliance audit system according to claim 1, characterized in that: The vertical business transmission node chain is constructed by combining each declaration node with the corresponding business review node and the customs declaration business information transmission path in combination with a distributed reinforcement learning algorithm and an encryption algorithm; the horizontal business review node network is constructed by combining all business review nodes with the strength of association between business review nodes through a graph algorithm; the declaration node network is constructed by combining all business declaration nodes with the similarity of corresponding business declaration types; the declaration node network is connected to the horizontal business review node network through the vertical business transmission node chain; The data acquisition module includes a data acquisition unit, a multimodal extraction unit and an enhanced update unit; The data collection unit is used to collect in real time the business declaration information uploaded by all business declaration nodes, the business declaration information that has not been completed in the review process, and the updated business review rule information, and classify the collected data information according to the data type; The multimodal extraction unit obtains a customs declaration business information table and an audit rule update table based on the classified real-time collected information in combination with a multimodal feature extraction algorithm, and performs a secondary audit update of the unfinished declaration business based on the effective timestamp of the audit rules in the audit rule update table; The enhanced update unit is used to update the audit rule information corresponding to each node in the declaration node network and the horizontal business audit node network according to the audit rule update table combined with the preset enhanced update model, and use the updated audit rules to conduct a secondary review and update of the unfinished declaration business review.

3. The intelligent customs compliance audit system according to claim 2, characterized in that: The compliance review module includes a declaration clustering unit, a primary review unit, a transmission unit, and a secondary review unit; The declaration clustering unit obtains a similar declaration node set based on the business information corresponding to all business declaration nodes in combination with a business similarity clustering model, and uses the cluster center corresponding to each similar declaration node subset in the similar declaration node set as the initial review node; The initial review unit obtains the customs declaration business information table after initial review based on the first enhanced review model configured for each initial review node, combined with the customs declaration business information table in the corresponding similar declaration node subset and the corresponding review rules; The transmission unit transmits and distributes the initially reviewed customs declaration business information table to the corresponding business review node through a transmission path optimization algorithm combined with a distributed enhanced allocation algorithm based on the reviewed customs declaration business information table corresponding to each initial review node, a preset encryption code, load information of the transmission channel corresponding to each vertical business transmission node chain, review business type information and load information corresponding to all the business review nodes, and initial review evaluation result information corresponding to each declaration node; The secondary review unit obtains the customs declaration business information table after the secondary review is completed based on the priority of the review business corresponding to each business review node, the priority of the sub-business review process corresponding to each business, and the index information between the review rules and each sub-business review process, combined with the second enhanced review model configured for each business review node.

4. The intelligent customs compliance audit system according to claim 3, characterized in that: The evaluation and optimization module includes a first evaluation unit, a second evaluation unit and an evaluation feedback unit; The first evaluation unit is configured to perform real-time evaluation using a first evaluation sub-model based on the clustering process corresponding to the declared clustering unit, the initial review process corresponding to the initial review unit, and the supply-demand relationship stability coefficient to obtain a first review risk probability; the first review risk probability is constructed from the clustering error probability and the initial false detection and missed detection risk probabilities; the supply-demand relationship stability coefficient is constructed from the ratio of the corresponding transaction completion frequency of the supply and demand parties to the total transaction frequency; The second evaluation unit is configured to perform real-time evaluation through a second evaluation sub-model based on the data transmission process corresponding to the transmission unit and the secondary audit process and audit anomaly coefficient corresponding to the secondary audit unit to obtain a second audit risk probability; The second audit risk probability is constructed by combining the tampering risk probability corresponding to the data transmission process, the business allocation risk probability, and the risk probabilities corresponding to the false detection, missed detection, and audit delay corresponding to the secondary audit with the Bayesian algorithm; The audit anomaly coefficient is constructed by the abnormal frequency of historical declarations of the corresponding declaring user and the total declaration frequency; The evaluation feedback unit optimizes the first enhanced audit model, the second enhanced audit model and the encrypted coding information based on the risk probability difference corresponding to the first audit risk probability and the second audit risk probability, combined with a preset multi-level risk discrimination threshold.

5. The intelligent customs compliance audit system according to claim 4, characterized in that: The workflow of the evaluation and feedback unit includes: Preset first audit judgment threshold , First audit difference discrimination threshold and the second audit difference discrimination threshold ,and , when the first audit risk probability is less than , the customs declaration business information table after the initial review is transmitted and assigned to the corresponding business review node. Otherwise, the first review risk probability is fed back to the first enhanced review model to update and optimize the clustering process and the initial review process until the first review risk probability is less than until; When the risk probability difference is less than When the risk probability difference is greater than or equal to and less than When the second audit risk probability is fed back to the transmission unit and the second enhanced audit model, the data transmission process and the secondary audit process are optimized and updated until the risk probability difference is less than Until the time; When the risk probability difference is greater than or equal to When the transmission encryption is abnormal, the business information corresponding to the business audit node is matched and compared with the business information of the corresponding declaration node to obtain comparison difference information. At the same time, the encryption coding level in the transmission unit and the business information for determining encryption abnormality are adjusted using the obtained comparison difference information, and the adjusted business information is updated and reviewed for the second time until the risk probability difference is less than Until then.

6. The intelligent customs compliance audit system according to claim 5, characterized in that: The process of obtaining the customs declaration business information form after the initial review includes: Each business application user is regarded as a business application node. At the same time, based on the application user's business type information and the corresponding target customer information, a similarity algorithm is used to obtain the business similarity corresponding to different application users at the same time stamp; Based on the similarity between the business declaration nodes and the corresponding businesses, a declaration node network is constructed through a graph algorithm; Based on the declaration node network, the business declaration information of different declaration users at the same time stamp is collected in real time, and the collected business declaration information is input into the multimodal extraction model to obtain the customs declaration business information table corresponding to different declaration users; The customs declaration business information table includes a unique identifier, business type, and structured business process information after feature extraction; Based on the business information and business similarity corresponding to the declaration node network, clustering is performed through clustering algorithm combined with time series algorithm to obtain a subset of similar declaration nodes corresponding to each similar business type, and the cluster center node of the similar declaration node subset is used as the initial review node of the corresponding similar declaration node subset.

7. The intelligent customs compliance audit system according to claim 6, characterized in that: The process of obtaining the customs declaration business information form after the initial review also includes: A first enhanced audit model is constructed based on the first enhanced audit sub-model configured for each declaration node in each similar declaration node subset and the first evaluation sub-model configured for the corresponding initial audit node through a federated algorithm framework; Based on the customs declaration business information table corresponding to each declaration node, the corresponding first enhanced review sub-model is used to review it, and the review result information is transmitted to the corresponding initial review node through the federation framework. The first evaluation sub-model is used to conduct review and evaluation. If the evaluation passes, the customs declaration business information table with the initial review completed is output; If the review fails, the assessment results will be fed back to the corresponding declaration node through the federal framework for adjustment until the first review judgment threshold is met, and the adjusted customs declaration business information table of the initial review completion is obtained.

8. The intelligent customs compliance audit system according to claim 7, characterized in that: The process of obtaining the customs declaration business information form after the second review includes: Each auditing business type department of the auditing party is regarded as a business audit node, and the audit process of the corresponding business type under each business audit node and the audit priority corresponding to each audit sub-node in the audit process under the current business type audit process are used to construct a business audit process node chain corresponding to each business audit node through the blockchain algorithm; each audit sub-node corresponds one-to-one to each audit sub-process in the corresponding type of business audit process; Based on the business review process node chain information corresponding to each business review node, the business overlap correlation degree corresponding to each business review node is obtained through the association algorithm; Based on each business review node and the corresponding business overlapping correlation, a horizontal business review node network is constructed through graph algorithm.

9. The intelligent customs compliance audit system according to claim 8, characterized in that: The process of obtaining the customs declaration business information form after the second review also includes: The horizontal business review node network obtains the review result corresponding to each review sub-node under each business review node and the comprehensive result of the corresponding business based on the customs declaration business information table of the initial review completed and the corresponding review priority and initial review evaluation result assigned by the vertical business transmission node chain, and the review rule index information corresponding to the corresponding review sub-process stored in each review sub-node under each business review node, through the second enhanced review sub-model configured for each review sub-node; Based on the comprehensive results of each business review node and the initial review results of the corresponding business, combined with the first review difference discrimination threshold, the second review difference discrimination threshold and the second evaluation sub-model, a secondary review judgment is performed. If the judgment passes, a customs declaration business information table with the secondary review completed is obtained; Otherwise, the corresponding business, initial review process, transfer process and secondary review process will be adjusted in real time according to the secondary review results until all review sub-processes of each review business meet the review rules; The second enhanced audit model is constructed by the second enhanced audit sub-model configured for each audit sub-node and the corresponding second evaluation sub-model through a federated framework.

10. A customs compliance intelligent audit method, which is implemented based on a customs compliance intelligent audit system according to any one of claims 1 to 9, characterized in that: include: Based on the real-time collection of review rules and customs declaration information combined with a multimodal extraction algorithm, the review rule update table and customs declaration business information table are obtained; According to the customs declaration classification information table, through the preset enhanced audit space combined with the audit rule update table, the audited customs declaration business information table is obtained; The enhanced audit space includes a horizontal business audit node network and a vertical business transmission node chain; The enhanced audit space is constructed and trained by combining the business information of the horizontal business audit node network and the vertical business transmission node chain, the preset audit rule index and the priority corresponding to the rules and business audits with the reinforcement learning algorithm; The vertical business transmission node chain is constructed by combining each declaration node with the corresponding business review node and the customs declaration business information transmission path in combination with a distributed reinforcement learning algorithm and an encryption algorithm; the horizontal business review node network is constructed by combining all business review nodes with the strength of association between business review nodes through a graph algorithm; the declaration node network is constructed by combining all business declaration nodes with the similarity of corresponding business declaration types; the declaration node network is connected to the horizontal business review node network through the vertical business transmission node chain; Based on the preset two-way risk assessment model and the real-time business audit information corresponding to the horizontal business audit node network and the vertical business transmission node chain, the business audit transmission risk probability, audit delay probability and audit error probability are obtained and fed back to the enhanced audit space to adjust the business transmission and audit process.

Images