A pre-warning method for customs clearance compliance management based on big data analysis

Abstract

The application relates to a pre-warning method for customs clearance compliance management based on big data analysis, which comprises a knowledge acquisition step, a feature matching step and a pre-warning analysis step; a knowledge graph is constructed through acquisition of regulation information and case information to generate key pre-warning items of customs for customs declaration, and when a certain declaration abnormal item corresponding to each pre-warning item appears, the relationship between the pre-warning item and the abnormal item is calculated to analyze the risk of each pre-warning item, so that the customs declaration personnel can be reminded to perform self-checking, verification or take remedial measures in advance.

Description

Technical Field

[0001] This invention relates to the field of customs import and export goods classification, and more specifically, to an early warning method for customs clearance compliance management based on big data analysis. Background Technology

[0002] Customs compliance management is a crucial aspect of maritime transport. To protect their national interests and fulfill international agreements, customs authorities in various countries employ compliance requirements, random inspections, testing, and analysis to manage cargo quality. Non-compliant goods are subject to measures such as port detention and fines. However, due to varying requirements, testing intensity, and document rigor across different countries, and the fact that customs inspection requirements are subject to policy adjustments, maritime cargo is susceptible to compliance risks due to human and freight-related factors. Failure to promptly identify and correct these risks can lead to significant hidden dangers. Currently, the main problems encountered are as follows: 1. Difficulty in standardizing the data format of customs declaration documents; 2. The complexity and diversity of product categories make searching difficult, and the semantic content is difficult to accurately convert, potentially leading to misunderstandings of customs clearance requirements; 3. The variety of customs declaration methods makes it difficult to understand customs clearance requirements. There is no unified standard for the format, type, and content of the generated data. While online review and customs declaration are possible, real-time early warning based on customs declaration information for compliance management still lacks a solid foundation. Therefore, patent CN109062872B proposes a method for unified processing of customs declaration documents in different formats to address the issue of data format uniformity. Patent CN112633006A proposes HS automatic classification and exclusive analysis compliance synchronization technology to achieve accurate numbering and classification of diverse products, enabling electronic data indexing of product types according to customs requirements. Patent CN108985718A presents a customs declaration data processing method based on draft customs declarations, completing the filling process management by coordinating the required declaration content at each port. Therefore, solving these three problems allows for big data analysis and judgment of declaration information, but it still cannot achieve compliance analysis of a specific transportation process. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide an early warning method for customs clearance compliance management based on big data analysis.

[0004] To address the aforementioned technical problems, the technical solution of this invention is: an early warning method for customs clearance compliance management based on big data analysis.

[0005] This includes knowledge acquisition steps, feature matching steps, and early warning analysis steps;

[0006] The knowledge acquisition steps include acquiring regulatory information and case information, generating early warning items and order anomaly items that match each early warning item based on the regulatory information using a preset risk construction strategy, and calculating the early warning threshold of the corresponding early warning item and matching the corresponding early warning analysis function based on the case information using a preset feature analysis strategy. The early warning analysis function reflects the correlation between the order anomaly value and the early warning risk value of the order anomaly item that matches the early warning item.

[0007] The anomaly analysis step includes analyzing customs declarations using a preset anomaly analysis strategy to identify corresponding declaration anomalies and calculate corresponding declaration anomaly values.

[0008] The early warning analysis step is configured with a preset early warning analysis function and early warning threshold for each early warning item. The early warning risk value of each early warning item is calculated through the early warning analysis function. When the early warning risk value is greater than the early warning threshold, the early warning information corresponding to the early warning item is output.

[0009] The risk construction strategy includes

[0010] Step A1: Identify regulatory information using a semantic recognition model to determine each risk feature word in the regulatory information, and classify the risk feature words according to an external semantic lexicon to generate several risk feature items;

[0011] Step A2: Retrieve historical customs declaration data, and extract the corresponding customs declaration items and historical declaration features corresponding to the customs declaration items from the historical customs declaration data based on the risk characteristics.

[0012] Step A3: Calculate the correlation matching value of each historical report feature and risk feature item using the preset first matching algorithm;

[0013] Step A4: Calculate the historical matching value of each customs declaration item and risk feature item using the preset first summation algorithm;

[0014] Step A5: Determine risk characteristic items as early warning items based on the preset matching constraints;

[0015] Step A6: Configure historical matching benchmarks, and designate customs declaration items whose historical matching values ​​with the warning items are greater than the historical matching benchmarks as declaration anomalies;

[0016] The anomaly analysis strategy includes a document verification sub-strategy, an information verification sub-strategy, a consistency matching sub-strategy, a risk identification sub-strategy, and a feedback verification sub-strategy.

[0017] Furthermore: In step A3, the first matching algorithm is as follows: Where u r For the associated matching value, b ig represents the matching correlation value between the i-th risk feature term and the historical order feature in the risk feature item. i Let k1 be the matching confidence value between the i-th risk feature word and the historical order feature in the risk feature item. The matching confidence value is positively correlated with the frequency of the risk feature word in the regulatory information. k1 is the total number of risk feature words in the risk feature item that have a matching relationship with the historical order feature.

[0018] In step A4, the first summation algorithm is: Among them, U r For the historical matching value, u ij For the j-th historical customs declaration data, g(T) represents the correlation matching value of the ith customs declaration item corresponding to the risk characteristic item. x () is a preset historical valid mapping function, which reflects the time interval T. x The mapping relationship between time validity values ​​and risk characteristics, where T0 is the current time, and T... j The historical moment when the historical customs declaration data was generated, kj is the total number of historical declaration features in the customs declaration item of the j-th historical customs declaration data, and k2 is the total number of historical customs declaration data with the customs declaration item.

[0019] In step A5, the matching constraint is a preset matching constraint range. Risk feature items whose historical matching values ​​fall into the corresponding matching constraint range are selected as warning items. When the historical matching value is higher than the corresponding matching constraint range, the risk feature item is marked, and the process returns to step A1 to generate a new risk feature item.

[0020] Furthermore: the aforementioned feature analysis strategy includes

[0021] Step B1: Extract the penalty conclusion from the case information and index the corresponding penalty risk value in the preset penalty association database using the penalty conclusion;

[0022] Step B2: Identify the corresponding report content data in the report anomaly items in the case information;

[0023] Step B3: Perform correlation analysis between the report content data and the penalty conclusion to obtain the corresponding conclusion correlation value, and remove report anomalies whose conclusion correlation value is lower than the preset correlation benchmark value;

[0024] Step B4: Divide the penalty conclusions by warning items to generate several penalty conclusion groups, and calculate the relevance weight of each abnormal item in the penalty conclusion group. The relevance weight is the weighted result of the relevance between the penalty conclusion and the report content data.

[0025] Step B5: Based on the proportional relationship of the relevance weights in the penalty conclusion group, index the corresponding early warning analysis function from the preset analysis function library as the early warning analysis function for this early warning item;

[0026] Step B6: Calculate the penalty risk value for each penalty conclusion group using a preset risk weighting algorithm, and generate the corresponding early warning threshold based on the penalty risk value.

[0027] Furthermore, step B1 also includes a risk update sub-strategy, which sorts similar penalty conclusions in the case information and adjusts the corresponding penalty risk value in the penalty association database according to the sorting results.

[0028] Step B3 also includes the aforementioned correlation analysis specifically being the PEN correlation analysis algorithm;

[0029] Furthermore: the document verification sub-strategy includes

[0030] Step Ca1: Retrieve the corresponding document verification pointer based on the report anomaly item;

[0031] Step Ca2: Retrieve the verification document based on the document verification pointer;

[0032] Step Ca3: Determine the corresponding document verification sub-item based on the type of the verification document;

[0033] Step Ca4: Verify each document verification sub-item and obtain the document abnormal sub-value corresponding to the document verification sub-item with the verification result being abnormal;

[0034] Step Ca5: Sum the document anomaly sub-values ​​to obtain the corresponding report anomaly values.

[0035] Furthermore: the information verification sub-strategy includes

[0036] Step Cb1: Obtain the HS code and verify its match with the basic product information;

[0037] Step Cb2: Generate the corresponding product verification sub-item based on the report anomaly item using the HS code;

[0038] Step Cb3: Verify each product verification sub-item and obtain the product risk sub-value corresponding to the product verification sub-item with an abnormal verification result;

[0039] Step Cb4: Sum the product risk sub-values ​​to obtain the corresponding order exception values.

[0040] Furthermore: the consistency matching sub-strategy includes

[0041] Step Cc1: Determine the corresponding declaration content in the customs declaration through the preset declaration anomaly items;

[0042] Step Cc2: Compare the report content according to the preset consistency conditions;

[0043] Step Cc3: Obtain the content of the report that does not meet the consistency conditions and generate a consistency exception sub-value;

[0044] Step Cc4: Sum the consistency anomaly sub-values ​​to obtain the corresponding report anomaly values;

[0045] The risk identification sub-strategy includes

[0046] Step Cd1: Retrieve the corresponding risk-sensitive keywords through the preset abnormal reporting items;

[0047] Step Cd2: Determine if there are any corresponding risk-sensitive words in the customs declaration;

[0048] Step Cd3: Generate sensitive anomaly sub-values ​​based on existing risk-sensitive words;

[0049] Step Cd4: Sum the sensitive anomaly sub-values ​​to obtain the corresponding report anomaly values.

[0050] Furthermore: the feedback verification sub-strategy includes

[0051] Step Ce1: Determine the monitoring feature sub-items and corresponding monitoring anomaly conditions through the preset report anomaly items;

[0052] Step Ce2: Obtain monitoring feedback data from the target monitoring terminal based on the monitoring feature sub-items;

[0053] Step Ce3: Compare the contents of the customs declaration with the monitoring feedback data based on the abnormal monitoring conditions;

[0054] Step Ce4: Generate corresponding feedback anomaly sub-values ​​based on the abnormal monitoring feedback data;

[0055] Step Ce5: Sum the feedback anomaly subvalues ​​to generate the report anomaly value.

[0056] Furthermore, the monitoring feedback data includes the appearance characteristics of the goods, the parameter characteristics of the goods, and the characteristics of the freight data.

[0057] The main technical effects of this invention are reflected in the following aspects: By setting it up in this way, a knowledge graph is constructed by acquiring regulatory information and case information to generate key early warning items for customs declarations, and when a certain declaration anomaly item corresponding to each early warning item occurs, the relationship between the early warning item and the corresponding early warning item is calculated to analyze the risk of each early warning item, and to remind customs declarants to conduct self-inspection, verification or take remedial measures in advance. Attached Figure Description

[0058] Figure 1The present invention provides a flowchart of the steps of an early warning method for customs clearance compliance management based on big data analysis. Detailed Implementation

[0059] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings, so that the technical solution of the present invention can be more easily understood and mastered.

[0060] An early warning method for customs clearance compliance management based on big data analytics:

[0061] This includes knowledge acquisition steps, feature matching steps, and early warning analysis steps;

[0062] The knowledge acquisition steps include acquiring regulatory information and case information. Firstly, the regulatory information refers to publicly available information on laws and policies from the customs or other external databases. Case information includes external data such as publicly available customs administrative penalty information, judgment information, and announced seizure information. Since these information formats and content types differ, traditionally, judgments are made through manual research and analysis. However, a pre-set risk construction strategy generates early warning items and matching report anomalies based on the regulatory information.

[0063] Specifically, the risk construction strategy includes the following:

[0064] Step A1: Identify regulatory information using a semantic recognition model to determine each risk feature word within the regulatory information. The purpose of the semantic recognition model is to identify valid information within the regulatory information and output it in a formatted manner, thereby facilitating the retrieval of textual features in the regulatory information to identify risk feature words. Based on an external semantic lexicon, the risk feature words are categorized to generate several risk feature items. Since similar risk feature words may exist, similar risk feature words are identified through the semantic lexicon and then classified to generate corresponding risk feature items. In this step, risk feature items will have inclusion and intersection relationships because regulations and words have hierarchical relationships, thus forming inclusion and intersection relationships.

[0065] Step A2: Retrieve historical customs declaration data and extract the corresponding customs declaration items and their corresponding historical declaration features from the historical customs declaration data based on the risk characteristics. Each customs declaration item and its corresponding historical declaration features can be extracted from the historical customs declaration data. For example, a combination of multiple sets of data for a customs declaration item, where one item is the quantity of goods, can be represented by the specific quantity details and the order of magnitude range. This allows for the splitting of customs declarations and the acquisition of corresponding splitting information.

[0066] Step A3: Calculate the correlation matching value of each historical report feature and risk feature item using a preset first matching algorithm; in step A3, the first matching algorithm is... Where u r For the associated matching value, b i This is the matching correlation value between the i-th risk feature term and historical order features in the risk feature item. Based on the matching degree between the risk feature term and the historical order features (e.g., if the risk feature term involves units, whether it is the same as or similar to the units in the historical order features (e.g., kilograms and tons), a corresponding value will be assigned if there is a match. Different matching correlation values ​​can be pre-assigned for different matching situations. i This is the matching confidence value between the i-th risk feature term and historical declaration features in the risk feature item. The matching confidence value is positively correlated with the frequency of the risk feature term in the regulatory information. If the frequency of a risk feature term is higher, it means that a historical declaration item will generally be measured by this risk feature. Therefore, the matching confidence value is converted by the frequency of occurrence. The conversion relationship and method are pre-configured, for example, by a certain ratio. k1 is the total number of risk feature terms that have a matching relationship with the historical declaration feature in the risk feature item. It should be noted that, theoretically, all historical declaration features are normal values. The customs declaration items corresponding to the risk feature items are pre-entered in the background. That is to say, each risk feature item may theoretically be related to which customs declaration items. This is a coarse screening process. Then, by calculating the matching result of each risk feature item and the historical declaration feature, the corresponding association matching value can be calculated.

[0067] Step A4: Calculate the historical matching value of each customs declaration item and risk feature item using a preset first summation algorithm; in step A4, the first summation algorithm is... Among them, U r For the historical matching value, u ij For the j-th historical customs declaration data, g(T) represents the correlation matching value of the ith customs declaration item corresponding to the risk feature item. x () is a preset historical valid mapping function, which reflects the time interval T. x The mapping relationship between historical and time validity values ​​is defined by a preset function. For example, time intervals within two months correspond to larger time validity values, while time intervals over two years correspond to smaller time validity values. This avoids the significant impact of historical customs declaration data with large time spans. The time validity value reflects the influence of historical declaration data on risk characteristics. T0 represents the current time, and T... jHere, kj represents the historical moment when the historical customs declaration data was generated, k2 represents the total number of historical declaration features in the j-th historical customs declaration data for that customs declaration item, and k2 represents the total number of historical customs declaration data with that customs declaration item. A customs declaration item may include several historical declaration features of different dimensions. After obtaining the correlation and matching values, the matching degree between the risk feature item and the customs declaration item can be calculated, which is the historical matching value. Since there may be historical customs declaration data generated at different times in the customs declaration, by matching all relevant historical customs declaration data with the risk feature item, the historical customs declaration data related to the risk feature item can be determined. This range is actually quite large.

[0068] Step A5: Determine risk features that meet preset matching constraints as early warning items. For example, in step A5, the matching constraints are preset matching constraint ranges. Risk features whose historical matching values ​​fall within the corresponding matching constraint range are selected as early warning items. Risk features with historical matching values ​​higher than the corresponding matching constraint range are marked, and the process returns to step A1 to generate new risk features. By setting different matching constraints, risk features corresponding to each customs declaration item can be filtered out as early warning items. This means further filtering of risk features to determine those with a high degree of correlation with the declaration, and also identifying related customs declaration items. This is specifically implemented through step A6.

[0069] Step A6: Configure historical matching benchmarks, and designate customs declaration items whose historical matching values ​​with the warning items are greater than the historical matching benchmarks as declaration anomalies.

[0070] And, based on case information, calculates the warning threshold for the corresponding warning item and matches the corresponding warning analysis function using a preset feature analysis strategy. The feature analysis strategy includes...

[0071] Step B1: Extract the penalty conclusions from the case information and index the corresponding penalty risk values ​​in a preset penalty association database using the penalty conclusions. Step B1 also includes a risk update sub-strategy, which sorts similar penalty conclusions in the case information and adjusts the corresponding penalty risk values ​​in the penalty association database based on the sorting results. Case information is generally based on penalty conclusions, and penalty conclusions are generally easy to classify and identify. Therefore, by retrieving the corresponding penalty risk values ​​in the preset association database using the penalty conclusions, and then determining the results in the corresponding penalty association database using the penalty risk values, the risk of different case information can be obtained.

[0072] Step B2: Identify the corresponding report content data in the report anomaly items in the case information; the report content data corresponding to the report anomaly items appearing in the case information is extracted through the identification model to ensure uniform format.

[0073] Step B3: Perform correlation analysis between the order content data and the penalty conclusion to obtain the corresponding conclusion correlation value, and remove order anomalies whose conclusion correlation value is lower than the preset correlation benchmark value; Step B3 also includes the aforementioned correlation analysis specifically using the PEN correlation analysis algorithm; Since there may be identical penalty conclusions with abnormal order content in the case information, and there may also be identical penalty conclusions with normal order content data, the correlation between the order content data and the penalty conclusion can be determined based on consistency analysis. If an anomaly results in a penalty conclusion, the correlation is high; if a normal penalty conclusion results in a penalty conclusion, the correlation is low or negative. The order content data represents the degree of anomaly, and the corresponding conclusion correlation value can be obtained, thus determining the relationship between the order anomaly and the penalty conclusion.

[0074] Step B4: Divide the penalty conclusions into several penalty conclusion groups based on the warning items. Calculate the relevance weight of each abnormal item in the penalty conclusion group. The relevance weight is the weighted result of the relevance between the penalty conclusion and the report content data. Since the relationship between the warning items and the penalty conclusions is known or can be pre-associated through purchasing, but a single penalty conclusion may have multiple warning items, the warning items can be used to divide the penalty conclusions.

[0075] Step B5: Based on the proportional relationship of the relevance weights in the penalty conclusion group, index the corresponding early warning analysis function from the preset analysis function library as the early warning analysis function for this early warning item. Since the relevance weights reflect the relationship between a report anomaly and the penalty conclusion, but the penalty conclusion may be associated with multiple report anomalies, the preset early warning analysis function is called based on the association relationship as the early warning analysis function for this early warning item. The early warning analysis function includes weighted, vector sum, square sum and other association relationships, which are preset. The report anomaly value is used as the input variable in the early warning analysis function to obtain the corresponding early warning risk value. That is to say, the degree of anomaly of a report anomaly is analyzed by the early warning analysis function to analyze its impact on the degree of penalty.

[0076] Step B6: Calculate the penalty risk value for each penalty conclusion group using a preset risk weighting algorithm, and generate the corresponding early warning threshold based on the penalty risk value. The risk weighting algorithm includes calculating the corresponding outlier value for each order anomaly, then weighting each outlier value according to the penalty risk value to obtain the weighted result for each type of order anomaly, and then substituting the weighted result into the early warning analysis function according to the order anomaly category to obtain the corresponding penalty risk value. The penalty risk value can be converted into an early warning threshold by a certain ratio.

[0077] The early warning analysis function reflects the correlation between the abnormal value of the order report and the early warning risk value of the abnormal order report item that has a matching relationship with the early warning item.

[0078] The anomaly analysis step includes analyzing customs declarations using a preset anomaly analysis strategy to identify corresponding declaration anomalies and calculate corresponding declaration anomaly values; the anomaly analysis strategy includes a document verification sub-strategy, which includes...

[0079] Step Ca1: Retrieve the corresponding document verification pointer based on the report exception item; pre-build a database for different document types, which stores the document verification sub-items for the corresponding documents.

[0080] Step Ca2: Retrieve the verification document based on the document verification pointer;

[0081] Step Ca3: Determine the corresponding document verification sub-item based on the type of document; for example, the document verification sub-item could be verifying the certificate barcode, verifying whether the document information is consistent with the declaration information, etc.

[0082] Step Ca4: Verify each document verification sub-item and obtain the document abnormal sub-value corresponding to the document verification sub-item with an abnormal verification result; if the verification result is abnormal, a corresponding risk is generated, and this risk is recorded as a report risk item.

[0083] Step Ca5: Sum the document anomaly sub-values ​​to obtain the corresponding order anomaly value. By summing the anomaly sub-values, the order anomaly value for the corresponding order risk item can be obtained.

[0084] Furthermore: the anomaly analysis strategy includes an information verification sub-strategy, which includes...

[0085] Step Cb1: Obtain the HS code and verify its match with the basic product information;

[0086] Step Cb2: Generate corresponding product verification sub-items based on the declaration anomaly items using the HS code. Since the HS code is a universal code and an important indexing tool for customs to query product categories, the corresponding product verification sub-items can also be confirmed through the HS code to quickly index the product. For example, whether the corresponding description of the product, whether the unit is consistent, or the possibility of a certain parameter being checked can all be generated through the declaration anomaly items to generate product verification sub-items.

[0087] Step Cb3: Verify each product verification sub-item and obtain the product risk sub-value corresponding to the product verification sub-item with the verification result being abnormal; if a product, for example, has a large deviation from other similar products, the abnormality is high, and the corresponding product risk sub-value is also high. The product risk sub-value is the deviation degree multiplied by the preset basic risk value of the product verification sub-item. If there is no deviation degree, only the verification of whether it is consistent, the deviation degree is represented by 0 and 1.

[0088] Step Cb4: Sum the product risk sub-values ​​to obtain the corresponding order exception values.

[0089] Furthermore: the anomaly analysis strategy includes a consistency matching sub-strategy and a risk identification sub-strategy, wherein the consistency matching sub-strategy includes...

[0090] Step Cc1: Determine the corresponding declaration content in the customs declaration through the preset declaration anomaly items; since some declaration contents are related, if they are inconsistent, they may be considered as anomalies.

[0091] Step Cc2: Compare the report content with the preset consistency conditions; the consistency comparison conditions not only compare whether they are the same, but also whether they correspond. Therefore, by using the consistency comparison conditions, you can check for text data anomalies in the report content.

[0092] Step Cc3: Obtain the content of the report that does not meet the consistency conditions and generate a consistency exception sub-value;

[0093] Step Cc4: Sum the consistency anomaly sub-values ​​to obtain the corresponding report anomaly values;

[0094] The risk identification sub-strategy includes

[0095] Step Cd1: Retrieve the corresponding risk-sensitive words through the preset abnormal order items; abnormal order items, such as the names of goods prohibited from import and export, will generate corresponding sensitive risk words.

[0096] Step Cd2: Determine if there are any corresponding risk-sensitive words in the customs declaration;

[0097] Step Cd3: Generate sensitive anomaly sub-values ​​based on existing risk-sensitive words; if a sensitive risk word appears, it indicates that the order is abnormal.

[0098] Step Cd4: Sum the sensitive anomaly sub-values ​​to obtain the corresponding report anomaly values.

[0099] The anomaly analysis strategy includes a feedback verification sub-strategy.

[0100] Step Ce1: Determine the monitoring feature sub-items and corresponding monitoring anomaly conditions through the preset report anomaly items;

[0101] Step Ce2: Obtain monitoring feedback data from the target monitoring terminal based on the monitoring feature sub-items; the monitoring feedback data includes cargo appearance characteristics, cargo parameter characteristics, and freight data characteristics. Verification is performed using data generated during the actual freight process or docking, obtaining information in real time to avoid inspection risks. For example, situations such as prolonged cargo spoilage can be predicted in advance to prevent credit issues. Discrepancies between the actual loaded cargo and the cargo recorded on the declaration will also be judged as abnormal. This data is fed back through the corresponding pre-associated target monitoring terminal.

[0102] Step Ce3: Compare the contents of the customs declaration with the monitoring feedback data based on the monitoring anomaly conditions; determine whether the corresponding contents and monitoring feedback data are the same and correspond to each other based on the monitoring anomaly conditions.

[0103] Step Ce4: Generate corresponding feedback anomaly sub-values ​​based on the abnormal monitoring feedback data;

[0104] Step Ce5: Sum the feedback anomaly subvalues ​​to generate the report anomaly value.

[0105] The early warning analysis step is configured with a preset early warning analysis function and early warning threshold for each early warning item. The early warning analysis function calculates the early warning risk value for each early warning item. When the early warning risk value is greater than the early warning threshold, the corresponding early warning information for the early warning item is output. By analyzing through the early warning analysis function, the corresponding early warning risk value can be obtained, thereby calculating the risk of each early warning item and outputting the corresponding early warning information in a timely manner. The early warning information includes the early warning item, the corresponding report feature item, and the corresponding report content, etc.

[0106] Of course, the above are just typical examples of the present invention. In addition, the present invention may have many other specific embodiments. All technical solutions formed by equivalent substitution or equivalent transformation fall within the scope of protection claimed by the present invention.

Claims

1. A pre-warning method for customs clearance compliance management based on big data analysis, characterized in that: it comprises a knowledge acquisition step, an abnormality analysis step, a feature matching step and a pre-warning analysis step; the knowledge acquisition step comprises acquiring regulation information and case information, generating a pre-warning item and a declaration abnormality item having a matching relationship with each pre-warning item according to the regulation information through a preset risk construction strategy, and calculating a pre-warning threshold value of the corresponding pre-warning item and a matching pre-warning analysis function according to the case information through a preset feature analysis strategy, the pre-warning analysis function reflecting the correlation between the declaration abnormality value and the pre-warning risk value of the declaration abnormality item having a matching relationship with the pre-warning item; the abnormality analysis step comprises analyzing the customs declaration through a preset abnormality analysis strategy to identify the corresponding declaration abnormality item and calculate the corresponding declaration abnormality value; the pre-warning analysis step is configured with a preset pre-warning analysis function and a pre-warning threshold value for each pre-warning item, and the pre-warning risk value of each pre-warning item is calculated through the pre-warning analysis function, and when the pre-warning risk value is greater than the pre-warning threshold value, the pre-warning information corresponding to the pre-warning item is output; the risk construction strategy comprises steps A1 to A6: step A1, identifying each risk feature word in the regulation information through a semantic recognition model, and classifying the risk feature words according to an external semantic word library to generate a plurality of risk feature items; step A2, calling historical customs declaration data, and extracting corresponding customs declaration items and historical declaration features corresponding to the customs declaration items from the historical customs declaration data according to the risk feature items; step A3, calculating the correlation matching value of each historical declaration feature and risk feature item through a preset first matching algorithm; step A4, calculating the historical matching value of each customs declaration item and risk feature item through a preset first summation algorithm; step A5, determining the risk feature item as a pre-warning item that meets the preset matching constraint condition; step A6, being configured with a historical matching reference, and taking the customs declaration item with a historical matching value greater than the historical matching reference as a declaration abnormality item; the abnormality analysis strategy comprises a document verification sub-strategy, an information verification sub-strategy, a consistency matching sub-strategy, a risk identification sub-strategy and a feedback inspection sub-strategy; in step A5, the matching constraint condition is a preset matching constraint range, and the risk feature item with a historical matching value falling within the corresponding matching constraint range is selected as a pre-warning item, and when the historical matching value is higher than the risk feature item within the corresponding matching constraint range, it is marked and returned to step A1 to generate a new risk feature item; the feature analysis strategy comprises steps B1 to B3: step B1, extracting the penalty conclusion in the case information, and indexing the corresponding penalty risk value in the preset penalty correlation database through the penalty conclusion; step B2, identifying the corresponding declaration content data in the declaration abnormality item in the case information; step B3, performing correlation analysis on the declaration content data and the penalty conclusion to obtain a corresponding conclusion correlation value, and eliminating the declaration abnormality item with a conclusion correlation value lower than a preset correlation reference value. ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ 2. The pre-warning method for customs compliance management based on big data analysis according to claim 1, characterized in that: The first matching algorithm in step A3 is wherein u r is the correlation matching value, b i is the matching correlation value of the i-th risk characteristic word in the risk characteristic item and the historical order characteristic, g i is the matching credibility value of the i-th risk characteristic word in the risk characteristic item and the historical order characteristic, the matching credibility value is positively correlated with the frequency of the risk characteristic word appearing in the regulation information, and k1 is the total number of risk characteristic words in the risk characteristic item that have a matching relationship with the historical order characteristic. In the step A4, the first summation algorithm is wherein U r is the historical matching value, u ij is the associated matching value of the i-th customs declaration item corresponding to the risk characteristic item in the j-th historical customs declaration data, g(T x ) is a preset historical effective mapping function, the historical effective mapping function reflects the mapping relationship between the time interval T x and the time effective value, the time effective value reflects the influence degree of the historical declaration data on the risk characteristic item, T0 is the current time, T j is the historical time when the historical customs declaration data is generated, kj is the total number of historical declaration characteristics in the customs declaration item in the j-th historical customs declaration data, and k2 is the total number of historical customs declaration data having the customs declaration item. ​ 3. The pre-warning method for customs compliance management based on big data analysis according to claim 1, characterized in that: ​ ​ ​ ​ Step B4, dividing the penalty conclusion into several penalty conclusion groups by the pre-warning item, calculating the correlation weight of each abnormal item in the penalty conclusion group, the correlation weight being the weighted result of the correlation between the penalty conclusion and the content data of the declaration; Step B5, indexing the corresponding pre-warning analysis function from the pre-set analysis function library as the pre-warning analysis function of the pre-warning item according to the proportion relationship of the correlation weight in the penalty conclusion group; Step B6, calculating the penalty risk value of each penalty conclusion group by the pre-set risk weighting algorithm, and generating the corresponding pre-warning threshold according to the penalty risk value.

4. The early warning method for customs compliance management based on big data analysis according to claim 3, characterized in that: In step B1, the risk updating sub-strategy is further included, which sorts the same type of penalty conclusions in the case information, and adjusts the corresponding penalty risk value in the penalty association database according to the sorting result; In step B3, the correlation analysis is specifically a person correlation analysis algorithm.

5. The pre-warning method for customs compliance management based on big data analysis according to claim 1, characterized in that: The document verification sub-strategy includes Step Ca1, retrieving the corresponding document verification pointer according to the abnormal item of the declaration; Step Ca2, retrieving the verification document according to the document verification pointer; Step Ca3, determining the corresponding document verification sub-item according to the type of the verification document; Step Ca4, verifying each document verification sub-item and obtaining the document abnormal sub-value corresponding to the document verification sub-item with an abnormal verification result; Step Ca5, summing the document abnormal sub-values to obtain the corresponding abnormal value of the declaration.

6. The early warning method for customs compliance management based on big data analysis according to claim 5, characterized in that: The information verification sub-strategy includes Step Cb1, obtaining the HS code and verifying the matching with the commodity basic information; Step Cb2, generating the corresponding commodity verification sub-item according to the abnormal item of the declaration through the HS code; Step Cb3, verifying each commodity verification sub-item and obtaining the commodity risk sub-value corresponding to the commodity verification sub-item with an abnormal verification result; Step Cb4, summing the commodity risk sub-values to obtain the corresponding abnormal value of the declaration.

7. The pre-warning method for customs compliance management based on big data analysis according to claim 6, characterized in that: The consistency matching sub-strategy includes Step Cc1, determining the corresponding declaration content in the customs declaration through the pre-set abnormal item of the declaration; Step Cc2, comparing the declaration content through the pre-set consistency condition; Step Cc3, obtaining the declaration content not meeting the consistency condition and generating a consistency abnormal sub-value; Step Cc4, summing the consistency abnormal sub-values to obtain the corresponding abnormal value of the declaration; The risk identification sub-strategy includes Step Cd1, retrieving the corresponding risk sensitive word through the pre-set abnormal item of the declaration; Step Cd2, judging whether the corresponding risk sensitive word exists in the customs declaration; Step Cd3, generating a sensitive abnormal sub-value according to the existing risk sensitive word; Step Cd4, summing the sensitive abnormal sub-values to obtain the corresponding abnormal value of the declaration.

8. The pre-warning method of customs compliance management based on big data analysis according to claim 7, characterized in that: The feedback inspection sub-strategy includes Step Ce1, determining the monitoring feature sub-item and the corresponding monitoring abnormal condition through the pre-set abnormal item of the declaration; Step Ce2, obtaining the monitoring feedback data from the target monitoring terminal according to the monitoring feature sub-item; Step Ce3, comparing the content of the customs declaration with the monitoring feedback data according to the monitoring abnormal condition; Step Ce4, generating the corresponding feedback abnormal sub-value according to the abnormal monitoring feedback data; Step Ce5, summing the feedback abnormal sub-values to generate the abnormal value of the declaration.

9. The pre-warning method for customs compliance management based on big data analysis according to claim 8, characterized in that: The monitoring feedback data includes appearance features of the goods, parameter features of the goods, and freight data features.

Images