Cross-border supply chain whole-link data fusion and intelligent abnormal early warning management and control method

Abstract

This invention relates to the field of digital management technology for cross-border supply chains, specifically a method for full-link data fusion and intelligent anomaly early warning and control in cross-border supply chains. The method includes the following steps: acquiring cross-border order records, customs declaration records, warehousing and inventory records, transportation trajectory records, port status records, customs clearance status records, delivery records, and cost records; converting these records into link node data; generating a full-link data sequence based on a multi-identifier association index; obtaining time-series deviation nodes according to node order and node time windows; cross-source verification of time-series deviation nodes to form candidate anomaly nodes; generating anomaly risk levels after confirming anomaly nodes; outputting early warning information or control instructions; and updating the anomaly status based on the handling results or node recovery status. This invention, through link node conversion, multi-identifier association, and cross-source verification, enables scattered records to form a continuous processing basis, ensuring a correspondence between anomaly early warning and control feedback.

Description

Technical Field

[0001] This invention relates to the field of digital management technology for cross-border supply chains, and in particular to a method for data fusion and intelligent anomaly early warning and control across the entire cross-border supply chain. Background Technology

[0002] The field of digital management technology for cross-border supply chains typically revolves around the digital collection, transmission, storage, association, and retrieval of cross-border orders, customs documents, warehousing inventory, transportation routes, port inspections, customs clearance status, and delivery records. Participants include enterprise management systems, logistics terminals, customs data interfaces, warehouse barcode scanners, vehicle positioning devices, and servers. Data forms a business chain based on order number, waybill number, cargo identifier, timestamp, geographical location, and node status. The traditional cross-border supply chain end-to-end data fusion and intelligent anomaly early warning and control method involves big data analysis and processing of business records at each node of the cross-border supply chain. Business systems and field terminals obtain order status, document fields, inventory changes, transportation locations, port releases, and fee records, which are then transmitted via network to servers or data warehouses. The data is then cleaned, matched, summarized, compared, and anomaly-marked according to subject identifier, node sequence, time difference, status fields, and preset thresholds. Early warning records are then written to the management interface for viewing by control personnel.

[0003] Existing digital management processes for cross-border supply chains often rely on order numbers, waybill numbers, cargo identifiers, and preset thresholds to summarize and compare cross-border orders, customs documents, warehousing inventory, transportation routes, port status, customs clearance status, delivery, and fee records. When records from different sources differ in terms of field definitions, node time, node location, and node status, the correspondence between links can easily be broken. There is a lack of continuous evidence between anomaly markers and subsequent control targets, and control personnel need to review the source and scope of anomalies in scattered records. Summary of the Invention

[0004] The purpose of this invention is to provide a method for full-chain data fusion and intelligent anomaly early warning and control in cross-border supply chains. This method is used to establish continuous correlations between cross-border order records, customs declaration records, warehousing and inventory records, transportation trajectory records, port status records, customs clearance status records, delivery records, and expense records, even when there are differences in field definitions, node time, node location, and node status among multiple sources in the cross-border supply chain. Furthermore, it enables traceable early warning and control processing for abnormal nodes. To achieve the above objectives, this invention provides a method for cross-border supply chain end-to-end data fusion and intelligent anomaly early warning and control, comprising the following steps: Obtain cross-border order records, customs declaration records, warehousing and inventory records, transportation tracking records, port status records, customs clearance status records, delivery records, and fee records; The acquired records are converted into link node data containing cargo identifier, business node, node time, node location, node status and source system identifier; A multi-identifier association index is established based on order number, waybill number, customs declaration number, container number, batch number and cargo identifier. The link node data pointing to the same cargo or transportation batch are merged according to the multi-identifier association index to generate a full-link data sequence. Based on the order of business nodes and the time window of nodes, the business nodes in the full-link data sequence are compared for missing, reversed and timed out to obtain the time sequence deviation nodes. Cross-validate the time-series deviation nodes with cross-border order records, customs declaration records, warehousing and inventory records, transportation trajectory records, port status records, customs clearance status records, delivery records, and fee records to obtain the cross-source verification results corresponding to the time-series deviation nodes. When the cross-source verification results indicate inconsistencies between node location, node status, source system identifier, document fields, delivery status, or fee occurrence status, the corresponding time-series deviation nodes will be generated as candidate abnormal nodes. Candidate abnormal nodes are confirmed to have multi-source record consistency anomalies, and candidate abnormal nodes that meet the conditions for multi-source record consistency anomalies are identified as abnormal nodes. An abnormal risk level is generated based on the abnormal node type, abnormal duration, quantity of goods involved, quantity of related documents, and scope of impact of subsequent nodes. Early warning information or control instructions are generated based on the abnormal risk level and abnormal node. Receive the handling results or node recovery status, and update the abnormal status of the corresponding link node data based on the handling results or node recovery status.

[0005] Furthermore, when converting the acquired records into link node data, the order number, planned node and planned time are extracted from the cross-border order records; the customs declaration number, document fields and declaration status are extracted from the customs declaration document records; the inventory status and warehouse time are extracted from the warehousing inventory records; the trajectory location and trajectory time are extracted from the transportation trajectory records; and the corresponding node status and node occurrence time are extracted from the port status records, customs clearance status records, delivery records and fee records. The extraction results are written into the link node data according to the cargo identifier, business node, node time, node location, node status, and source system identifier; When any record is missing a cargo identifier, the corresponding cargo identifier is supplemented based on the order number, waybill number, customs declaration number, box number, or batch number in the same record, and a field integrity flag is written to the corresponding link node data.

[0006] Furthermore, the process of generating field integrity tags includes: The cargo identifier, business node, node time, and node status in the link node data are set as basic fields, and the node location and source system identifier are set as verification fields; when all basic fields exist and the verification fields are missing, an integrity tag to be verified is generated. When any field in the base field is missing, a completeness tag to be supplemented is generated; when both the base field and the check field exist, a valid completeness tag is generated, and the field completeness tag is used as the filtering basis when building a multi-identifier associated index.

[0007] Furthermore, when establishing a multi-identifier association index, first filter the link node data with valid integrity tags, and use the order number, waybill number, customs declaration number, box number, batch number and cargo identifier in the link node data as association fields; When the same cargo identifier corresponds to at least two identical associated fields, and the node time is within the time range of the same transportation batch, the corresponding link node data will be grouped into the same index group. When a related field is missing, the link node data with the integrity tag to be verified is called to supplement the matching, and the matching result is written to the multi-identifier related index.

[0008] Furthermore, when comparing missing, inverted, and timed-out business nodes in the full-link data sequence, the order of business nodes is determined according to the planned nodes in the cross-border order records, and adjacent business nodes in the full-link data sequence are sorted by node time. When a pre-defined business node is missing between adjacent business nodes, a missing node marker is generated; When the node time of the subsequent business node is earlier than the node time of the previous business node, an inverted node marker is generated. When the time difference between adjacent business nodes exceeds the corresponding node time window, a timeout node marker is generated, and the timing deviation node is obtained based on the missing node marker, inverted node marker, or timeout node marker.

[0009] Furthermore, during cross-validation, the cargo identifier and business node of the time-series deviation node are used as search conditions to extract the corresponding node location, node status, source system identifier, document fields, delivery status and cost occurrence status from cross-border order records, customs declaration records, warehousing and inventory records, transportation trajectory records, port status records, customs clearance status records, delivery records and cost records. The results are compared and extracted based on the corresponding business nodes to obtain the cross-source verification results. When the cross-source verification result indicates that there are inconsistencies in fields between at least two types of records, and the inconsistent fields correspond to the same business node, the corresponding time-series deviation node is identified as a candidate abnormal node.

[0010] Furthermore, when a candidate abnormal node is identified as an abnormal node, the inconsistent fields corresponding to the candidate abnormal node are grouped according to the source system identifier, and the number of records and node time corresponding to each group are counted. When records corresponding to at least two source system identifiers point to the same inconsistent field within the same node time window, the candidate abnormal node is determined to meet the multi-source record consistency abnormality condition, and the corresponding candidate abnormal node is identified as an abnormal node. When the number of records does not meet the multi-source record consistency exception condition, the pending confirmation status of the corresponding candidate exception node is retained.

[0011] Furthermore, when generating anomaly risk levels, the node risk category is determined according to the type of anomaly node, the continuous risk category is determined according to the duration of the anomaly, the object risk category is determined according to the quantity of goods involved and the number of related documents, and the link impact category is determined according to the scope of impact of subsequent nodes. The candidate risk levels are obtained by prioritizing and comparing the node risk category, the persistent risk category, the object risk category, and the link impact category. When the number of abnormal sources corresponding to the candidate risk level is consistent with the number of abnormal sources corresponding to the cross-source verification results, the candidate risk level is determined as the abnormal risk level.

[0012] Furthermore, during priority coverage comparison, when the number of subsequent nodes corresponding to the link impact category reaches the preset number of impact nodes, the link impact category covers the node risk category, the ongoing risk category, and the object risk category. When the number of affected nodes in the link does not reach the preset number of affected nodes, and the number of goods or related documents involved in the object risk category reaches the preset number of objects, the object risk category covers the node risk category and the ongoing risk category. When neither the link impact category nor the object risk category meets the corresponding preset conditions, the candidate risk level is determined based on the category with the highest risk ranking among the node risk category and the continuous risk category.

[0013] Furthermore, when updating the abnormal status, if the handling result indicates that the abnormality has been confirmed, the abnormal status of the corresponding abnormal node will be updated to the confirmed status, and the corresponding early warning information or control instructions will be retained. When the node recovery status indicates that the business node has once again met the node time window and cross-source verification conditions, the abnormal status of the corresponding link node data will be updated to the recovered status. When the handling result indicates an abnormality that has not been confirmed and the node recovery status does not meet the recovery conditions, the abnormal status of the corresponding link node data will be updated to a pending review status.

[0014] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this invention, cross-border order records, customs declaration records, warehousing and inventory records, transportation trajectory records, port status records, customs clearance status records, delivery records, and fee records are converted into link node data. A full-link data sequence is generated using a multi-identifier association index, enabling continuous comparison of records from different sources under the same cargo or transportation batch dimension. Candidate abnormal nodes are formed through time-series deviation nodes and cross-source verification results. Abnormal nodes are then confirmed and an abnormal risk level is generated. This allows early warning information or control instructions to simultaneously associate the abnormal source, duration, cargo quantity, document quantity, and the impact range of subsequent nodes. After the handling result or node recovery status is written back, the abnormal state is kept consistent with the actual handling process. Attached Figure Description

[0015] Figure 1 This is the main flowchart of the end-to-end anomaly early warning and control of the present invention; Figure 2 This is a schematic diagram of the link node data structure of the present invention; Figure 3 This is a schematic diagram illustrating the generation of the multi-identifier association index according to the present invention; Figure 4 This is a schematic diagram of the timing deviation node and cross-source verification status of the present invention; Figure 5 This is a schematic diagram illustrating the update of abnormal risk levels and abnormal states in this invention. Detailed Implementation

[0016] This embodiment focuses on cross-border order records, customs declaration records, warehousing and inventory records, transportation trajectory records, port status records, customs clearance status records, delivery records, and fee records in the digital management of cross-border supply chains. The so-called link node data refers to data records formed by converting the above-mentioned types of records and containing cargo identification, business nodes, node time, node location, node status, and source system identification. The so-called multi-identifier association index refers to the association relationship established based on order number, waybill number, customs declaration number, container number, batch number and cargo identifier; the so-called full-link data sequence refers to the sequence of link node data pointing to the same cargo or transportation batch, organized according to the business node order and node time. The term "timing deviation node" refers to a business node identified in comparisons of missing, inverted, and timeout conditions. The cross-source verification result refers to the field consistency result formed after cross-verifying the time-series deviation node with eight types of records; The term "candidate abnormal node" refers to a node that is triggered by cross-source verification results and is waiting for confirmation of multi-source record consistency anomalies. The term "abnormal node" refers to a candidate abnormal node that meets the conditions for multi-source record consistency anomaly.

[0017] In subsequent embodiments, the Sxx class number is used to distinguish method flows and does not constitute a limitation on the number of executions, parameter range, or protection boundaries.

[0018] Please see Figures 1 to 5 This embodiment provides a method for full-link data fusion and intelligent anomaly early warning and control in cross-border supply chains. It is designed for processing scenarios where records of orders, documents, warehousing, transportation, ports, customs clearance, delivery and fees are scattered in cross-border supply chains. Data processing is performed in the following order: data acquisition, node conversion, identifier association, time sequence comparison, cross-source verification, anomaly confirmation, risk classification and feedback update. This enables the node status of the same goods or transportation batch in different source records to form a continuous processing link.

[0019] S1 retrieves cross-border order records, customs declaration records, warehousing and inventory records, transportation tracking records, port status records, customs clearance status records, delivery records, and fee records.

[0020] In S1, cross-border order records provide the order number, planned node, and planned time; customs declaration records provide the customs declaration number, document fields, and declaration status; warehousing and inventory records provide the inventory status and in-warehouse time; transportation track records provide the track location and track time; port status records provide the node status and node occurrence time corresponding to the port node; customs clearance status records provide the node status and node occurrence time corresponding to the customs clearance node; delivery records provide the delivery status and node occurrence time; and fee records provide the fee occurrence status and node occurrence time.

[0021] In S1, when retrieving various records, fields reflecting the business entity and node status are retained for the purpose of associating the same goods or the same transportation batch. In cross-border order records, the order number and planned node are used to define the planned order of the entire data sequence; in customs declaration records, the customs declaration number and document fields are used to define the correspondence between goods and the declaration process; in warehousing and inventory records, the inventory status is used to define the status of goods at warehousing nodes; in transportation trajectory records, the trajectory location and trajectory time are used to define the node location and node time during transportation; port status records, customs clearance status records, and delivery records are used to define the status of subsequent business nodes; and expense records are used to define the correspondence between expense occurrence status and business nodes.

[0022] S2 converts the acquired records into link node data containing cargo identifier, business node, node time, node location, node status, and source system identifier.

[0023] S201 extracts the order number, planned node, and planned time from the cross-border order record, uses the planned node as the business node, the planned time as the node time, and writes the order source into the source system identifier.

[0024] S202: Extract the customs declaration number, document fields, and declaration status from the customs declaration document records. Take the declaration stage corresponding to the document field as the business node, the declaration status as the node status, and write the customs declaration source into the source system identifier.

[0025] S203 extracts inventory status and in-warehouse time from warehouse inventory records, taking the corresponding links of inbound, outbound or inventory change as business nodes, in-warehouse time as node time, and inventory status as node status.

[0026] S204: Extract the trajectory location and trajectory time from the transportation trajectory record, take the transportation link corresponding to the change of transportation location as the business node, take the trajectory time as the node time, and take the trajectory location as the node location.

[0027] S205: Extract the corresponding node status and node occurrence time from the port status record, customs clearance status record, delivery record and cost record, take the port, customs clearance, delivery and cost occurrence stages as business nodes, take the node occurrence time as node time, and write the corresponding status into the node status.

[0028] S206. Write the extraction results formed in S201 to S205 into the link node data according to the cargo identifier, business node, node time, node location, node status and source system identifier.

[0029] In the link node data, the cargo identifier is used to distinguish the processing object, the business node is used to distinguish the supply chain links, the node time is used to establish the order, the node location is used to reflect the spatial status, the node status is used to reflect the business execution status, and the source system identifier is used to record the record source corresponding to the link node data.

[0030] S207: When any record is missing a cargo identifier, supplement the corresponding cargo identifier based on the order number, waybill number, customs declaration number, box number or batch number in the same record, and write a field integrity flag for the corresponding link node data.

[0031] When supplementing the corresponding cargo identifier, first determine whether there is at least one of the following business identifiers in the same record: order number, waybill number, customs declaration number, box number, or batch number; When a business identifier exists, the association between the business identifier and known goods identifiers is used to supplement the goods identifier; If no corresponding business identifier exists, the record will not be directly entered into the subsequent full-link data sequence generation process, but the record status will be limited to "to be supplemented".

[0032] S208, the process of generating field integrity markers includes: setting the cargo identifier, business node, node time and node status in the link node data as basic fields, and setting the node location and source system identifier as verification fields; when all basic fields exist and the verification fields are missing, an integrity marker to be verified is generated. When any field in the basic fields is missing, an integrity tag to be supplemented is generated; A valid integrity flag is generated when both the base field and the validation field exist.

[0033] Field integrity flags are used to distinguish whether link node data can directly participate in index grouping when building multi-identifier association indexes later.

[0034] S3 establishes a multi-identifier association index based on order number, waybill number, customs declaration number, container number, batch number, and cargo identifier, and merges the link node data pointing to the same cargo or transportation batch according to the multi-identifier association index to generate a full-link data sequence.

[0035] S301, when establishing a multi-identifier association index, first filter the link node data with valid integrity tags, and use the order number, waybill number, customs declaration number, box number, batch number and cargo identifier in the link node data as association fields.

[0036] The association field is used to establish a correspondence between records from different sources for the same goods or shipping batch. The association field is not used as a separate conclusion for anomaly judgment, but as an index basis for subsequent merging and sorting.

[0037] S302, when the same cargo identifier corresponds to at least two identical associated fields, and the node time is within the time range of the same transportation batch, the corresponding link node data will be grouped into the same index group.

[0038] The link node data within the same index group is organized according to the business node order and node time to form the basic data set for the subsequent full-link data sequence.

[0039] S303, when the associated field is missing, call the link node data with the integrity mark to be verified for supplementary matching.

[0040] When performing supplementary matching, the existing order number, waybill number, customs declaration number, container number, batch number, or cargo identifier is used as the matching entry point to search for link node data corresponding to the same business node, the same time range of the same node, or the same transportation batch. When the supplemented matching field relationship can point to the same goods or transportation batch, the matching result is written to the multi-identifier association index; When supplementary matching cannot form a corresponding relationship, the integrity flag to be verified is retained, and the failed matching records are not forcibly merged into the index group.

[0041] S304 When generating the full-link data sequence, the link node data within the same index group are arranged according to the planning node in the cross-border order record, the declaration stage in the customs declaration record, the inventory stage in the warehousing and inventory record, the transportation stage in the transportation trajectory record, the port stage in the port status record, the customs clearance stage in the customs clearance status record, the delivery stage in the delivery record, and the cost occurrence stage in the cost record, and sorted according to the node time within the same business node.

[0042] After sorting, a full-link data sequence is generated that can characterize the flow status of the same goods or transportation batch throughout the entire cross-border supply chain.

[0043] S4, according to the business node order and node time window, compare the missing, reversed and timed-out business nodes in the full-link data sequence to obtain the time sequence deviation nodes.

[0044] S401 determines the order of business nodes according to the planned nodes in the cross-border order record, and sorts adjacent business nodes in the full-link data sequence by node time.

[0045] The order of business nodes is used to define the relative sequence of orders, customs declarations, warehousing, transportation, ports, customs clearance, delivery, and fees in the cross-border supply chain. The node time window is used to define the acceptable time relationship between adjacent business nodes.

[0046] S402, when a preset business node is missing between adjacent business nodes, a missing node marker is generated. The business node corresponding to the missing node marker is used as one of the sources of timing deviation nodes. The missing node marker is not directly equivalent to an abnormal node, but is used for subsequent cross-source verification.

[0047] S403, when the node time of the subsequent business node is earlier than the node time of the previous business node, generate an inverted node marker.

[0048] The inverted node marker is used to reflect the state where the node time order is inconsistent with the business node order. The business node corresponding to the inverted node marker enters the subsequent cross-validation process.

[0049] S404: When the time difference between adjacent service nodes exceeds the corresponding node time window, a timeout node marker is generated.

[0050] The timeout node marker is used to reflect the state where the time interval between adjacent business nodes exceeds the node time window. The business node corresponding to the timeout node marker enters the subsequent cross-validation process.

[0051] S405: Obtain the timing deviation node based on the missing node marker, inverted node marker, or timeout node marker. When the same business node corresponds to more than two markers, retain the source of each marker and the corresponding node time in the timing deviation node so that subsequent cross-source verification can distinguish the source of the deviation.

[0052] S5 cross-validates the time-series deviation node with cross-border order records, customs declaration records, warehousing and inventory records, transportation trajectory records, port status records, customs clearance status records, delivery records, and fee records to obtain the cross-source validation results corresponding to the time-series deviation node.

[0053] S501 uses the cargo identifier and business node of the time-series deviation node as search criteria to extract the corresponding node location, node status, source system identifier, document fields, delivery status, and cost occurrence status from cross-border order records, customs declaration records, warehousing and inventory records, transportation trajectory records, port status records, customs clearance status records, delivery records, and cost records. The cargo identifier in the search criteria is used to determine the processing object, and the business node is used to determine the supply chain link that needs to be verified.

[0054] S502, compare and extract results according to the business node correspondence to obtain cross-source verification results. When comparing and extracting results, node position is used to compare with transportation trajectory records and business node correspondence, node status is used to compare with warehousing inventory records, port status records, customs clearance status records and delivery records, source system identifier is used to limit the source of inconsistent fields, document fields are used to compare with the correspondence between customs declaration document records and cross-border order records, delivery status is used to compare with the status of the preceding node, and cost occurrence status is used to compare with the status of the business node.

[0055] S503 When there is an inconsistency between the cross-source verification result and the node location, node status, source system identifier, document field, delivery status, or cost occurrence status, the corresponding time-series deviation node will be identified as a candidate abnormal node.

[0056] Inconsistent fields are only used as the basis for generating candidate abnormal nodes when they correspond to the same business node; differences in fields that do not belong to the same business node do not directly generate candidate abnormal nodes.

[0057] S504 When the cross-source verification result does not represent a field inconsistency, the verification pass status of the corresponding time-series deviation node is retained, and the verification pass status is retained together with the link node data of the business node for subsequent full-link data sequence status tracing.

[0058] S6, perform multi-source record consistency anomaly confirmation on candidate abnormal nodes, and determine the candidate abnormal nodes that meet the multi-source record consistency anomaly conditions as abnormal nodes.

[0059] S601, group the inconsistent fields corresponding to the candidate abnormal nodes according to the source system identifier, and count the number of records and node time corresponding to each group.

[0060] Grouping is used to distinguish the source of inconsistent fields, record count is used to determine whether inconsistent fields come from multiple types of records, and node time is used to determine whether inconsistent fields are located within the same node time window.

[0061] S602, when records corresponding to at least two source system identifiers point to the same inconsistent field within the same node time window, the candidate abnormal node is determined to meet the multi-source record consistency abnormality condition, and the corresponding candidate abnormal node is identified as an abnormal node.

[0062] Multi-source record consistency exception conditions are used to prevent field differences from a single record source from directly entering the exception node determination process.

[0063] S603, when the number of records does not meet the multi-source record consistency exception condition, retain the pending confirmation status of the corresponding candidate exception node.

[0064] The pending confirmation status is stored together with the candidate abnormal node, the corresponding business node, the inconsistency field, the source system identifier, and the node time, so that the subsequent handling results or node recovery status can continue to be written back to the corresponding link node data.

[0065] S7 generates an abnormal risk level based on the abnormal node type, abnormal duration, quantity of goods involved, quantity of related documents, and scope of impact of subsequent nodes, and generates early warning information or control instructions based on the abnormal risk level and abnormal node.

[0066] S701. Determine the node risk category according to the abnormal node type. The abnormal node type comes from the missing node marker, the inverted node marker, the timeout node marker, and the inconsistent field in the cross-source verification result. The node risk category is used to describe the risk source to which the abnormal node belongs.

[0067] S702, determine the persistent risk category according to the duration of the anomaly. The duration of the anomaly is determined by the time relationship between the node time corresponding to the anomaly node and the node time corresponding to the node recovery status in subsequent records. The persistent risk category is used to describe the state in which the anomaly node continues to exist.

[0068] S703 determines the risk category of an object based on the quantity of goods involved and the number of associated documents. The quantity of goods involved is obtained by counting the goods identifiers associated with the same abnormal node, and the number of associated documents is obtained by counting the customs declaration numbers or document fields associated with the same abnormal node. The object risk category is used to describe the scope of objects associated with the abnormal node.

[0069] S704. The link impact category is determined according to the impact range of subsequent nodes. The impact range of subsequent nodes is determined by the number of business nodes affected after the abnormal node and the node status. The link impact category is used to describe the impact relationship of the abnormal node on subsequent links.

[0070] S705 compares the priority coverage of node risk categories, persistent risk categories, object risk categories, and link impact categories to obtain candidate risk levels.

[0071] When the number of subsequent nodes corresponding to the link impact category reaches the preset number of impact nodes, the link impact category covers the node risk category, the ongoing risk category, and the object risk category. When the number of affected nodes in the link does not reach the preset number of affected nodes, and the number of goods or related documents involved in the object risk category reaches the preset number of objects, the object risk category covers the node risk category and the ongoing risk category. When neither the link impact category nor the object risk category meets the corresponding preset conditions, the candidate risk level is determined based on the category with the highest risk ranking among the node risk category and the continuous risk category.

[0072] S706 When the number of abnormal sources corresponding to the candidate risk level is consistent with the number of abnormal sources corresponding to the cross-source verification result, the candidate risk level shall be determined as the abnormal risk level.

[0073] When there is a discrepancy between the candidate risk level and the number of anomaly sources, retain the corresponding record between the candidate risk level and the number of anomaly sources, and put the anomaly node into a pending review state.

[0074] This process makes the anomaly risk level jointly constrained by node type, persistent state, object scope, link impact, and number of anomaly sources.

[0075] S707 generates early warning information or control instructions based on the level of abnormal risk and abnormal nodes. The early warning information includes the abnormal node, the type of abnormal node, the duration of the abnormality, the quantity of goods involved, the number of related documents, and the scope of impact of subsequent nodes. The control instructions include the business node corresponding to the abnormal node, the processing object, the warning level, and the handling status that needs to be written back. The warning information or control instructions are used to establish a correspondence between the abnormal node and the subsequent handling process.

[0076] S8 receives the handling result or node recovery status, and updates the abnormal status of the corresponding link node data based on the handling result or node recovery status.

[0077] S801: When the handling result indicates that the abnormality has been confirmed, the abnormal status of the corresponding abnormal node will be updated to the confirmed status, and the corresponding early warning information or control instructions will be retained.

[0078] The confirmed status, abnormal nodes, abnormal risk level, number of abnormal sources, and corresponding business nodes are all stored together for subsequent status tracing.

[0079] S802, when the node recovery status indicates that the service node has met the node time window and cross-source verification conditions again, the abnormal status of the corresponding link node data will be updated to the recovered status.

[0080] Node recovery status includes the business node's node time falling back into the node time window, the node position matching the corresponding business node, the node status being consistent with the corresponding record, the document fields matching the corresponding goods identifier, the delivery status being consistent with the previous node status, and the cost occurrence status being consistent with the business node status.

[0081] S803: When the handling result indicates an abnormality that has not been confirmed and the node recovery status does not meet the recovery conditions, the abnormal status of the corresponding link node data will be updated to a pending review status.

[0082] The pending verification status, together with the candidate abnormal node, abnormal node, cross-source verification result, handling result, and node recovery status, form a feedback relationship, so that when the same cargo identifier or the same transportation batch of data is obtained again, the previous abnormal status can be referenced.

[0083] S804: After the abnormal state update is completed, the updated link node data will be reintegrated into the corresponding full-link data sequence.

[0084] When re-integrating, the original cargo identifier, business node, node time, node location, node status, and source system identifier remain unchanged. Only the abnormal status, handling result, node recovery status, and status fields related to the abnormal risk level are updated. Thus, the process of confirming, controlling, and recovering abnormal nodes maintains the same data thread as the entire data sequence.

[0085] Through the above S1 to S8, this embodiment converts eight types of cross-border supply chain records into link node data and organizes them into a full-link data sequence through multi-identifier association index. On this basis, this embodiment first obtains time-series deviation nodes by comparing missing, inverted, and timeout, generates candidate abnormal nodes through cross-source verification results, and determines abnormal nodes after confirming the consistency anomaly of multi-source records. This embodiment forms an abnormal risk level based on the abnormal node type, abnormal duration, quantity of goods involved, quantity of associated documents, and scope of impact of subsequent nodes, and writes back and associates the early warning information or control instructions with the handling results, node recovery status, and abnormal status of link node data.

[0086] The specific processes, field states, processing order, judgment conditions, status markers, risk categories, feedback update methods, and process relationships presented in the accompanying drawings described in this embodiment are used to explain the feasible implementation methods of the technical solutions described in this invention in cross-border supply chain digital management, and should not limit this invention to the specific embodiments listed. Without departing from the technical solutions described in this invention and the scope supported by the original disclosure, any equivalent substitutions, equivalent modifications, equivalent combinations, order adjustments, equivalent changes in field names, equivalent acceptance of the executing entity, or equivalent changes in the carrier form that can be conceived by those skilled in the art should fall within the scope of protection of this patent; however, equivalent changes must not extend to unclaimed topics, must not alter the substantive correspondence of technical objects through name changes, and must not replace, delete, or weaken the method features, data relationships, temporal relationships, judgment relationships, feedback relationships, or parameter sources necessary to achieve the technical effect.

Claims

1. A method for cross-border supply chain end-to-end data fusion and intelligent anomaly early warning and control, characterized in that, Includes the following steps: Obtain cross-border orders, customs documents, warehouse inventory, transportation tracking, port status, customs clearance status records, delivery and fee records; The acquired records are converted into link node data containing cargo identifier, business node, node time, node location, node status and source system identifier; A multi-identifier association index is established based on the order number, waybill number, customs declaration number, container number, batch number, and cargo identifier. The link node data pointing to the same cargo or transportation batch are merged according to the multi-identifier association index to generate a full-link data sequence. Based on the order of business nodes and the node time window, the business nodes in the full-link data sequence are compared for missing, inverted and timed out to obtain the time sequence deviation nodes. The time-series deviation nodes are cross-validated with cross-border orders, customs documents, warehouse inventory, transportation trajectory, port status, customs clearance status, delivery and fee records to obtain cross-source verification results. When the cross-source verification results indicate inconsistencies between node location, node status, source system identifier, document fields, delivery status or fee occurrence status, the corresponding time-series deviation node is identified as a candidate abnormal node. The candidate abnormal nodes are confirmed to have multi-source record consistency anomalies, and the candidate abnormal nodes that meet the multi-source record consistency anomaly conditions are determined as abnormal nodes. An abnormal risk level is generated based on the abnormal node type, abnormal duration, quantity of goods involved, quantity of related documents, and scope of impact of subsequent nodes. Early warning information or control instructions are generated based on the abnormal risk level and the abnormal node. Receive the handling result or node recovery status, and update the abnormal status of the corresponding link node data based on the handling result or node recovery status.

2. The cross-border supply chain end-to-end data fusion and intelligent anomaly early warning and control method according to claim 1, characterized in that: The step of converting the acquired records into link node data containing cargo identifier, business node, node time, node location, node status, and source system identifier includes: The following steps are performed: extracting the order number, planned node, and planned time from the cross-border order record; extracting the customs declaration number, document fields, and declaration status from the customs declaration record; extracting the inventory status and warehouse time from the warehousing inventory record; extracting the trajectory location and trajectory time from the transportation trajectory record; and extracting the corresponding node status and node occurrence time from the port status record, the customs clearance status record, the delivery record, and the cost record. The extraction results are written into the link node data according to the cargo identifier, the business node, the node time, the node location, the node status, and the source system identifier; When any record lacks the cargo identifier, the corresponding cargo identifier is supplemented based on the order number, waybill number, customs declaration number, box number, or batch number in the same record, and a field integrity flag is written to the corresponding link node data.

3. The cross-border supply chain end-to-end data fusion and intelligent anomaly early warning and control method according to claim 2, characterized in that: The process of generating the field integrity marker includes: Set the cargo identifier, business node, node time, and node status in the link node data as basic fields, and set the node location and source system identifier as verification fields; When all the basic fields exist and the verification field is missing, an integrity tag to be verified is generated. When any of the basic fields is missing, a completeness tag to be supplemented is generated; When both the base field and the verification field exist, a valid integrity flag is generated, and the field integrity flag is used as the filtering basis when the multi-identifier association index is subsequently established.

4. The cross-border supply chain end-to-end data fusion and intelligent anomaly early warning and control method according to claim 3, characterized in that: The steps for establishing the multi-identifier association index include: Filter the link node data with the aforementioned valid integrity marker, and use the order number, waybill number, customs declaration number, container number, batch number, and cargo identifier in the link node data as associated fields; When the same cargo identifier corresponds to at least two of the aforementioned associated fields being identical, and the node time is within the time range of the same transportation batch, the corresponding link node data will be grouped into the same index group; When the associated field is missing, the link node data with the integrity tag to be verified is called to supplement the matching, and the matching result is written into the multi-identifier association index.

5. The cross-border supply chain end-to-end data fusion and intelligent anomaly early warning and control method according to claim 1, characterized in that: The steps for comparing missing, inverted, and timeout business nodes in the entire data sequence according to the business node order and node time window include: The order of the business nodes is determined according to the planned nodes in the cross-border order records, and adjacent business nodes in the full-link data sequence are sorted by node time. When a pre-defined business node is missing between adjacent business nodes, a missing node marker is generated; When the node time of the subsequent business node is earlier than the node time of the previous business node, an inverted node marker is generated. When the time difference between adjacent business nodes exceeds the corresponding node time window, a timeout node marker is generated, and the timing deviation node is obtained based on the missing node marker, the inverted node marker, or the timeout node marker.

6. The cross-border supply chain end-to-end data fusion and intelligent anomaly early warning and control method according to claim 1, characterized in that: The steps for cross-validating the time-series deviation node with the cross-border order record, the customs declaration record, the warehousing and inventory record, the transportation trajectory record, the port status record, the customs clearance status record, the delivery record, and the fee record include: Using the cargo identifier and business node of the time-series deviation node as search conditions, the corresponding node location, node status, source system identifier, document field, delivery status and cost occurrence status are extracted from various records. The results are compared and extracted based on the corresponding business nodes to obtain the cross-source verification results. When the cross-source verification result indicates that there is a field inconsistency between at least two types of records, and the inconsistent field corresponds to the same business node, the corresponding time-series deviation node is determined as a candidate abnormal node.

7. The cross-border supply chain end-to-end data fusion and intelligent anomaly early warning and control method according to claim 6, characterized in that: The steps for identifying the candidate abnormal nodes as abnormal nodes include: The inconsistent fields corresponding to the candidate abnormal nodes are grouped according to the source system identifier, and the number of records and node time corresponding to each group are counted. When records corresponding to at least two source system identifiers point to the same inconsistent field within the same node time window, the candidate abnormal node is determined to meet the multi-source record consistency abnormality condition, and the corresponding candidate abnormal node is identified as an abnormal node. When the number of records does not meet the multi-source record consistency exception condition, the pending confirmation status of the corresponding candidate exception node is retained.

8. The cross-border supply chain end-to-end data fusion and intelligent anomaly early warning and control method according to claim 6, characterized in that: The steps for generating an anomaly risk level based on the anomaly node type, anomaly duration, quantity of goods involved, number of related documents, and impact scope of subsequent nodes include: The risk category of a node is determined by the type of abnormal node, the risk category of a continuous node is determined by the duration of the abnormality, the risk category of an object is determined by the quantity of goods involved and the number of related documents, and the impact category of the link is determined by the scope of impact of subsequent nodes. The node risk category, the persistent risk category, the object risk category, and the link impact category are compared by priority coverage to obtain candidate risk levels; When the number of abnormal sources corresponding to the candidate risk level is consistent with the number of abnormal sources corresponding to the cross-source verification result, the candidate risk level is determined as the abnormal risk level.

9. The cross-border supply chain end-to-end data fusion and intelligent anomaly early warning and control method according to claim 8, characterized in that: The priority coverage comparison process includes: When the number of subsequent nodes corresponding to the link impact category reaches the preset number of impact nodes, the link impact category covers the node risk category, the ongoing risk category, and the object risk category. When the number of affected nodes in the link does not reach the preset number of affected nodes, and the number of goods or related documents involved in the object risk category reaches the preset number of objects, the object risk category covers the node risk category and the continuous risk category. When neither the link impact category nor the object risk category meets the corresponding preset conditions, the candidate risk level is determined based on the risk category with the highest risk ranking among the node risk category and the continuous risk category.

10. The cross-border supply chain end-to-end data fusion and intelligent anomaly early warning and control method according to claim 6, characterized in that: The steps of receiving the handling result or node recovery status, and updating the abnormal status of the corresponding link node data based on the handling result or node recovery status, include: When the handling result indicates that the anomaly has been confirmed, the anomaly status of the corresponding anomaly node will be updated to the confirmed status, and the corresponding early warning information or control instructions will be retained. When the node recovery status indicates that the business node has once again met the node time window and cross-source verification conditions, the abnormal status of the corresponding link node data will be updated to the recovered status. When the processing result indicates an abnormality that has not been confirmed and the node recovery status does not meet the recovery conditions, the abnormal status of the corresponding link node data will be updated to a pending review status.

Images