Intelligent analysis method for customs declaration data

Abstract

The invention relates to the technical field of data processing, in particular to a customs declaration data intelligent analysis method, which comprises the following steps: acquiring the export quantity and price of all customs declarations of any article of any customer; calculating the data similarity between each customs declaration and other customs declaration; constructing a first coordinate system; determining a representative number and a representative price; calculating the deviation degree of each customs declaration; constructing a second coordinate system; calculating a weighted Euclidean distance between each customs declaration and the feature point of any one of the rest customs declaration; and according to the weighted Euclidean distance, obtaining an abnormal score of each customs declaration by using an LOF algorithm so as to realize intelligent analysis of customs declaration data. According to the method, the accuracy of customs declaration anomaly detection and the effectiveness of risk assessment are remarkably improved, a supervision mechanism can be helped to find potential risks in time, the supervision cost is reduced, and the overall risk control level is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and more specifically, to a method for intelligent analysis of customs declaration form data. Background Art

[0002] In international trade, historical customs declaration data, as a core resource for recording detailed information on import and export activities, has irreplaceable value for comprehensively evaluating customer business conditions, predicting market trends, and analyzing potential risks. However, the current customs declaration data entry process relies heavily on manual operations by managers, which may result in input errors, negligence, or possible malicious tampering, which may lead to outliers in the customs declaration data, seriously damaging the accuracy and credibility of the data. In order to effectively address this problem and improve the overall quality of customs declaration data, an intelligent analysis method for customs declaration data is needed. By deeply mining and carefully analyzing historical customs declaration data, it can automatically identify abnormal customs declarations, making it convenient for the review department to check abnormal customs declarations in a timely manner.

[0003] The patent document with announcement number CN116307125B discloses a seasonal change analysis system for the number of customs declarations based on traffic data. The system includes: a prediction processing device, which establishes a corresponding artificial intelligence prediction body for each cargo loading and unloading port in a set area, so as to predict the monthly traffic data of the port this year based on the port's historical traffic data, port operating parameters, and the traffic data of the related ports in the previous year; a total amount analysis device, which obtains the total amount of traffic data of each cargo loading and unloading port in the set area every month of the year; a season judgment device, which determines the total amount of customs declarations of each cargo loading and unloading port in the set area every month of the year and performs seasonal change analysis and processing.

[0004] However, the above patent document is an analysis of seasonal changes in the number of customs declarations based on traffic data, not an analysis of the customs declaration data itself. Since the entry of customs declaration data is often manual, there may be human input errors or tampering, resulting in abnormal customs declaration data. The traditional Local Outlier Factor (LOF) algorithm uses the Euclidean distance of each indicator to calculate the local density to obtain abnormal points when processing multi-indicator data. For a certain customer, the historical customs declarations of a certain type of goods often have a certain degree of similarity, and it is impossible to capture the relationship between the various indicators of the historical customs declaration data well, resulting in misjudgment or omission of abnormal data, thereby reducing the accuracy of risk declaration assessment. Summary of the invention

[0005] In order to solve the problem that the LOF algorithm uses the Euclidean distance of each indicator to calculate the local density to obtain the abnormal score when processing multi-indicator data, and the historical customs declarations of a certain type of goods of any customer have a certain similarity, so that the relationship between the indicators of the historical customs declarations cannot be well captured, resulting in misjudgment or omission of abnormal data and reducing the accuracy of risk customs declaration assessment, the present invention proposes a method for intelligent analysis of customs declaration data, which includes the following steps: Obtain the export quantity and price of all customs declarations; calculate the data similarity between each customs declaration and the other customs declarations; construct a first coordinate system based on the customs declaration date of each customs declaration and the data similarity to obtain the similarity points of the corresponding customs declarations; construct a second coordinate system based on the export quantity and price of each customs declaration to obtain the feature points of the corresponding customs declarations; cluster each similarity point as the center to obtain several clusters, and take the cluster containing the largest number of customs declarations as the representative cluster, and then obtain the representative quantity and representative price of the representative cluster; record any customs declaration as the target The target customs declaration form is selected, and the deviation degree of the target customs declaration form is calculated based on the difference between the export quantity and price of the target customs declaration form and the representative quantity and representative price respectively; the Euclidean distance between the feature points of the target customs declaration form and the other customs declaration forms and the deviation degree are used as weights to obtain the weighted Euclidean distance between the feature points of the target customs declaration form and the other customs declaration forms; and the LOF algorithm is used to obtain the abnormal score of the target customs declaration form according to the weighted Euclidean distance to realize intelligent analysis of the target customs declaration form data.

[0006] The beneficial effects are: by comprehensively considering the export quantity and price of the customs declarations as well as the similarity and deviation between the data, it is possible to more accurately identify abnormal customs declarations that are obviously deviated from the norm in quantity and price; through the LOF algorithm combined with weighted Euclidean distance and cluster analysis, it is possible to effectively screen out potential high-risk customs declarations from a large number of customs declarations, thereby enhancing the reliability of risk assessment and more comprehensively reflecting the inherent relationship between data when faced with complex and multi-dimensional data; it is able to adapt to the different types and multi-dimensional characteristics of customs declaration data, so that when faced with diversified data, the analysis results are still stable and accurate, with strong robustness.

[0007] Furthermore, all the customs declaration forms are all customs declaration forms for the same item of the same user.

[0008] Furthermore, the data similarity satisfies the following relationship: ; In the formula, For the The data similarity between the customs declaration form and the other customs declaration forms, For the The export quantity of each customs declaration, For the remaining customs declarations The export quantity of each customs declaration, is the standard deviation of the export quantity of all customs declarations, For the The price of a customs declaration, For the remaining customs declarations The price of a customs declaration, is the standard deviation of the prices of all customs declarations, is the number of regional declarations, is a natural exponential function.

[0009] The beneficial effects are: by comprehensively considering the differences in export quantity and price and standardizing them, the similarity calculation between data is made more accurate. By using the standard deviation, the influence of different data dimensions on the similarity calculation can be eliminated, so that the relative relationship and difference between various types of data (such as quantity, price) can be measured more fairly; by quantifying the relative difference in export quantity and price, the deviation between a single customs declaration and the rest of the customs declarations can be effectively measured, so that when faced with data with large fluctuations in quantity or price, those abnormal customs declarations that deviate from the normal state can be more accurately identified without being affected by the extreme value of a single data point.

[0010] Furthermore, the abscissa of the first coordinate system is the declaration date of the customs declaration, and the ordinate is the data similarity; the abscissa of the second coordinate system is the export quantity of the customs declaration, and the ordinate is the price of the customs declaration.

[0011] Furthermore, the clustering adopts DBSCAN algorithm.

[0012] Furthermore, the representative quantity and the representative price of the representative cluster are respectively the mean of the export quantity and the mean of the price of all customs declarations in the representative cluster.

[0013] Furthermore, the degree of deviation satisfies the following relationship: ; In the formula, For the The degree of deviation of the customs declaration, For the The export quantity of each customs declaration, represents the number, For the The price of a customs declaration, To represent the price, is the standard normalization function.

[0014] The beneficial effects are: the degree of deviation is quantified by calculating the Euclidean distance between the export quantity and price of the customs declaration and the representative quantity and representative price, thereby quantifying the degree of deviation of the quantity and price of each customs declaration; through standard normalization processing, the deviation value can eliminate the impact of dimensional differences and different data ranges, ensure the consistency and fairness of the calculation results, and can more accurately evaluate which customs declarations have significant deviations from the normal pattern, thereby identifying potential abnormal and high-risk data.

[0015] Furthermore, the weighted Euclidean distance satisfies the following relationship: ; In the formula, No. The first customs declaration and the remaining customs declarations The weighted Euclidean distance between the characteristic data points of the customs declarations, No. The degree of deviation of the customs declaration, No. The first customs declaration and the remaining customs declarations The cosine similarity between the feature data points of the customs declarations, No. The first customs declaration and the remaining customs declarations The Euclidean distance between the characteristic data points of the customs declarations, is the standard normalization function, is the hyperbolic tangent function.

[0016] The beneficial effects are: comprehensively considering the similarities and differences between different features to optimize the distance calculation of customs declarations; compressing extreme values ​​through the hyperbolic tangent function to avoid their excessive influence on the calculation results; combining Euclidean distance and cosine similarity to more comprehensively reflect the similarities and differences between data.

[0017] Furthermore, the intelligent analysis of the target customs declaration data is achieved, including: in response to the abnormality score being greater than the abnormality threshold, determining that the data of the target customs declaration is abnormal, and issuing an early warning prompt to complete the intelligent analysis of the customs declaration data.

[0018] The beneficial effects are: by setting abnormal scores and thresholds, abnormal situations in customs declaration data can be monitored and detected in real time to ensure the accuracy and reliability of the data; when customs declaration data is abnormal, the system can automatically issue an early warning prompt and promptly notify relevant personnel to intervene or verify. Automated processing greatly improves the efficiency and response speed of data processing, and reduces the possibility of human errors and underreporting.

[0019] The present invention has the following beneficial effects: The present invention calculates the similarity between the customs declaration form and the historical average level, finds the densest cluster through clustering, determines the representative features of the normal customs declaration form, and weights the distance of the LOF algorithm by using the degree of deviation between each customs declaration form and the representative features combined with the cosine similarity of the spatial distribution of the feature data points, so as to more accurately evaluate the abnormal score of each customs declaration form, reduce the possibility of misjudgment and missed judgment, and improve the accuracy and reliability of risk assessment; the present invention significantly improves the accuracy of customs declaration form anomaly detection and the effectiveness of risk assessment, can help regulatory agencies to timely discover potential risks, reduce regulatory costs, and improve the overall risk control level. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] By reading the detailed description below with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals are the same or corresponding parts, wherein: Figure 1 The present invention is a flowchart of a method for intelligently analyzing customs declaration data according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0022] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0023] See also Figure 1 , which shows a flow chart of a method for intelligent analysis of customs declaration data provided by an embodiment of the present invention, the method comprising the following steps: S1: Get the export quantity and price of all customs declarations.

[0024] Use the customer ID to retrieve all the customs declarations of the customer in all the historical customs declarations, and use the item number to retrieve all the customs declarations of the customer's item in all the customs declarations of the customer, and obtain the data of each customs declaration in the history of the customer's item. Each customs declaration data includes: customs declaration date, export quantity, price, weight, taxes, origin and destination, etc. Since the actual customs declaration contains multiple parameter data, such as customs declaration date, export quantity, price, etc., directly comparing all the data is more complicated and not all parameters have a direct impact on the main focus of the analysis. Therefore, extract the main indicators of the core content of the trade activities reflected in the actual customs declaration, such as the quantity and price of exported items.

[0025] Specifically, all the customs declaration forms are all customs declaration forms for the same item of the same user.

[0026] S2: Calculate the data similarity between each customs declaration and the rest of the customs declarations.

[0027] It should be noted that by evaluating the similarity between the key indicators of each customs declaration and the rest of the customs declarations, the patterns and trends in the historical data of the item are analyzed to determine whether there is a fixed trading pattern for the same customer when exporting a certain item.

[0028] Specifically, the data similarity satisfies the following relationship: ; In the formula, For the The data similarity between the customs declaration form and the other customs declaration forms, For the The export quantity of each customs declaration, For the remaining customs declarations The export quantity of each customs declaration, is the standard deviation of the export quantity of all customs declarations, For the The price of a customs declaration, For the remaining customs declarations The price of a customs declaration, is the standard deviation of the prices of all customs declarations, is the number of regional declarations, is a natural exponential function.

[0029] in, Indicates The product of the absolute difference between the export quantity of a customs declaration and the average export quantity of all other customs declarations except itself and the absolute difference in price. The smaller the value, the higher the export quantity of the customs declaration. The closer the export quantity and price of a customs declaration are to the average export quantity and price of other customs declarations except itself, the more similar the data of this customs declaration is to the other customs declarations. It represents the product of the standard deviation of export quantity and the standard deviation of price of all customs declarations. Through standardization, the influence caused by the absolute numerical differences in quantity and price of different customs declarations can be eliminated, making the similarity calculation more accurate.

[0030] S3: Construct the first coordinate system and the second coordinate system.

[0031] A first coordinate system is constructed based on the declaration date of each customs declaration and the data similarity to obtain the similarity points of the corresponding customs declaration; a second coordinate system is constructed based on the export quantity and price of each customs declaration to obtain the characteristic points of the corresponding customs declaration.

[0032] Specifically, the horizontal axis of the first coordinate system is the declaration date of the customs declaration (for example, the earliest customs declaration is numbered 1, the second is numbered 2, and so on), and the vertical axis is the data similarity; the horizontal axis of the second coordinate system is the export quantity of the customs declaration, and the vertical axis is the price of the customs declaration.

[0033] S4: Determine the number of representatives and the representative price.

[0034] It should be noted that in the actual transaction process, customers may have certain trade patterns when exporting certain items; by analyzing the similarity characteristics of any historical customs declarations, the density clustering algorithm is used to cluster based on the similarity characteristics. The density clustering algorithm can identify high-density areas in the data; if most customs declarations are concentrated in a certain cluster, then it can be considered that these customs declarations represent the normal trade pattern of the customer. Therefore, based on further analysis of the clustering results, the point closest to the customer's normal trade pattern can be found as the representative of the customer's historical customs declaration.

[0035] Clustering is performed with each similarity point as the center to obtain several clusters. The cluster containing the largest number of customs declarations is taken as the representative cluster, and then the representative quantity and representative price of the representative cluster are obtained.

[0036] Specifically, the clustering adopts the DBSCAN algorithm.

[0037] A density-based spatial clustering of applications with noise (DBSCAN) algorithm is to cluster similar points of each customs declaration if its All similarity points in the neighborhood are greater than or equal to (Default ), then this point is the core point; with the core point as the center, all similarity data points in its Eps neighborhood form a temporary cluster, and traverse each similarity data point in the temporary cluster. If a similarity point is a core point and other similarity data points in its neighborhood have not been classified into any cluster, then these points are also classified into the current cluster, and this process is repeated until they can no longer be merged, until all similarity data points are divided into different clusters. Use the traditional DBSCAN algorithm to cluster the similarity data points of each historical customs declaration, divide the historical customs declarations into different clusters, and traverse the number of customs declarations in each cluster.

[0038] Specifically, the representative quantity and the representative price of the representative cluster are respectively the mean of the export quantity and the mean of the price of all customs declarations in the representative cluster.

[0039] S5: Calculate the degree of deviation of each customs declaration.

[0040] It should be noted that the customer's historical customs declaration data for a certain item reflects its long-term export activity pattern. The representative characteristics of the customer's historical customs declarations analyzed in combination with the historical customs declaration data are used to reflect the degree to which each customs declaration deviates from the normal trade pattern by considering the distance between each customs declaration and the representative characteristics of the customer's historical customs declarations.

[0041] Any customs declaration is recorded as a target customs declaration, and the deviation degree of the target customs declaration is calculated based on the difference between the export quantity and price of the target customs declaration and the representative quantity and representative price, respectively.

[0042] Specifically, the degree of deviation satisfies the following relationship: ; In the formula, For the The degree of deviation of the customs declaration, For the The export quantity of each customs declaration, represents the number, For the The price of a customs declaration, To represent the price, is the standard normalization function.

[0043] in, Indicates The square of the difference between the export quantity of the declaration form and the representative quantity. The larger the value, the The more the export quantity of a customs declaration deviates from the export quantity of the normal trade pattern; Indicates The square of the difference between the price of the first declaration and the representative price. The larger the value, the The more the price of a customs declaration deviates from the normal trade pattern price; Indicates The Euclidean distance between the representative characteristic data points of the first customs declaration and the historical customs declaration. The larger the value, the The export quantity and price of the customs declaration form deviate more from the normal trade pattern. The greater the deviation of the customs declaration.

[0044] S6: Calculate the weighted Euclidean distance between the feature points of each customs declaration and the other customs declarations.

[0045] It should be noted that the traditional LOF algorithm calculates the Euclidean distance between each data point and other points, and finds the distance between each data point and other points. The neighborhood points calculate the local density to get the anomaly score, but the data in the customs declaration form are often distributed in high-dimensional space. The relative positions between the data points become blurred, making it difficult for the Euclidean distance metric to accurately reflect the actual relationship between the data points, resulting in inaccurate calculation of the local density. Therefore, it is necessary to analyze the proximity of the two customs declaration forms in the feature space, and the cosine similarity is a method to measure the size of the angle between two vectors in the vector space. It is not affected by the length of the vector and only focuses on the angle between the vectors, that is, the angle between the two points relative to a certain point in the horizontal axis direction. The implementer can select this point according to the specific implementation situation, such as the origin of the coordinates, so as to more accurately measure the similarity of the customs declaration data; if the cosine similarity between any customs declaration form and the feature vectors of other customs declaration forms is lower, and the deviation of the customs declaration form is greater, the data in the customs declaration form is considered to be less reliable; therefore, it is necessary to weight the distance between the customs declaration form and other customs declaration form data points, that is, the feature points, obtained by the traditional algorithm, and the distance is corrected to be farther away from the customs declaration form, so as to isolate the customs declaration form and improve the accuracy of anomaly detection.

[0046] The cosine similarity of the feature points of the target customs declaration form and the remaining customs declaration forms and the degree of deviation are used as weights to weight the Euclidean distance between the feature points of the target customs declaration form and the remaining customs declaration forms to obtain the weighted Euclidean distance between the feature points of the target customs declaration form and the remaining customs declaration forms.

[0047] Specifically, the weighted Euclidean distance satisfies the following relationship: ; In the formula, No. The first customs declaration and the remaining customs declarations The weighted Euclidean distance between the characteristic data points of the customs declarations, No. The degree of deviation of the customs declaration, No. The first customs declaration and the remaining customs declarations The cosine similarity between the feature data points of the customs declarations, No. The first customs declaration and the remaining customs declarations The Euclidean distance between the characteristic data points of the customs declarations, is the standard normalization function, is the hyperbolic tangent function.

[0048] in, Indicates The customs declaration and The weighted value of the Euclidean distance between the feature points of the two customs declarations. Since the cosine similarity value range is , the larger its value is, the more similar the distribution of two feature points in the feature space is considered; The function will The value is limited to between; if The customs declaration and The smaller the cosine similarity between the feature points of the two declarations, the larger the angle between the feature points of the two declarations in the feature space, and the farther the distribution distance between the two declarations in the feature space. The more the data of a customs declaration deviates from the normal trade pattern (i.e., the greater the degree of deviation), the less reliable the data in the customs declaration is, and there may be anomalies. Therefore, it is necessary to isolate the customs declaration and distinguish it from the rest by giving a larger weight to the initial Euclidean distance.

[0049] S7: Based on the weighted Euclidean distance, the LOF algorithm is used to obtain the abnormal score of the target customs declaration form to achieve intelligent analysis of the target customs declaration form data.

[0050] Specifically, the method for realizing intelligent analysis of target customs declaration data includes: In response to the abnormal score being greater than the abnormal threshold, the data of the target customs declaration is determined to be abnormal, and an early warning is issued. Furthermore, the percentage of the number of historical early warnings to the total number of historical customs declarations is counted, and the risk level of the customer is divided in combination with the specified risk level to complete the intelligent analysis of the customs declaration data.

[0051] Implementers can set the exception threshold according to specific implementation circumstances, for example, 1.

[0052] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for intelligent analysis of customs declaration data, characterized in that: include: Obtain export quantities and prices for all customs declarations; Calculate the data similarity between each customs declaration and the rest of the customs declarations; A first coordinate system is constructed based on the customs declaration date of each customs declaration and the data similarity to obtain the similarity points of the corresponding customs declaration; a second coordinate system is constructed based on the export quantity and price of each customs declaration to obtain the characteristic points of the corresponding customs declaration; Clustering is performed with each similarity point as the center to obtain several clusters, and the cluster containing the largest number of customs declarations is taken as the representative cluster, and then the representative quantity and representative price of the representative cluster are obtained; any customs declaration is recorded as the target customs declaration, and the deviation degree of the target customs declaration is calculated based on the difference between the export quantity and price of the target customs declaration and the representative quantity and representative price respectively; Using the cosine similarity of the characteristic points of the target customs declaration and the other customs declarations and the degree of deviation as weights, weighting the Euclidean distance between the characteristic points of the target customs declaration and the other customs declarations to obtain the weighted Euclidean distance between the characteristic points of the target customs declaration and the other customs declarations; According to the weighted Euclidean distance, the LOF algorithm is used to obtain the abnormal score of the target customs declaration form, so as to realize intelligent analysis of the target customs declaration form data.

2. The method for intelligent analysis of customs declaration data according to claim 1, characterized in that: The said all customs declaration forms are all customs declaration forms for the same item of the same user.

3. The method for intelligent analysis of customs declaration data according to claim 1, characterized in that: The data similarity satisfies the following relationship: ； In the formula, For the The data similarity between the customs declaration form and the other customs declaration forms, For the The export quantity of each customs declaration, For the remaining customs declarations The export quantity of each customs declaration, is the standard deviation of the export quantity of all customs declarations, For the The price of a customs declaration, For the remaining customs declarations The price of a customs declaration, is the standard deviation of the prices of all customs declarations, is the number of regional declarations, is a natural exponential function.

4. The method for intelligent analysis of customs declaration data according to claim 1, characterized in that: The abscissa of the first coordinate system is the declaration date of the customs declaration, and the ordinate is the data similarity; the abscissa of the second coordinate system is the export quantity of the customs declaration, and the ordinate is the price of the customs declaration.

5. The method for intelligent analysis of customs declaration data according to claim 1, characterized in that: The clustering adopts the DBSCAN algorithm.

6. The method for intelligent analysis of customs declaration data according to claim 1, characterized in that: The representative quantity and representative price of the representative cluster are respectively the mean of the export quantity and the mean of the price of all customs declarations in the representative cluster.

7. The method for intelligent analysis of customs declaration data according to claim 1, characterized in that: The degree of deviation satisfies the following relationship: ； In the formula, For the The degree of deviation of the customs declaration, For the The export quantity of each customs declaration, represents the number, For the The price of a customs declaration, To represent the price, is the standard normalization function.

8. The method for intelligent analysis of customs declaration data according to claim 1, characterized in that: The weighted Euclidean distance satisfies the following relationship: ； In the formula, No. The first customs declaration and the remaining customs declarations The weighted Euclidean distance between the characteristic data points of the customs declarations, No. The degree of deviation of the customs declaration, No. The first customs declaration and the remaining customs declarations The cosine similarity between the feature data points of the customs declarations, No. The first customs declaration and the remaining customs declarations The Euclidean distance between the characteristic data points of the customs declarations, is the standard normalization function, is the hyperbolic tangent function.

9. The method for intelligent analysis of customs declaration data according to claim 1, characterized in that: The method for realizing intelligent analysis of target customs declaration data includes: In response to the abnormality score being greater than the abnormality threshold, the data of the target customs declaration is determined to be abnormal, and an early warning prompt is issued to complete the intelligent analysis of the customs declaration data.

Images