Method and system for full-level penetration supervision of supply chain based on digital ticket

Abstract

The application discloses a kind of supply chain all levels penetration supervision method and system based on digital electricity ticket, comprising: first determine the supply chain penetration check matter needing to be completed by the regulatory side node and the business side node cooperation, both sides system pre-configured target full chain supervision association data including same transaction link but different dimension information, and with completely consistent digital electricity ticket number as transaction anchor point.The regulatory side is according to the rule constraint of check matter, and the ticket number anchor point of specific number transaction link is screened out from local data.Subsequently, according to the pre-set coding scale, these ticket number anchor points are coded and processed to generate compact digital electricity ticket number anchor point coding data.Finally, the coding data is sent to the business side node, and the specific ticket number anchor point is restored by the business side decoding, and the penetration check is executed accordingly.The application realizes the accurate, safe, efficient cooperation of cross-system data, and improves the penetration ability and efficiency of supply chain supervision.

Description

Technical Field

[0001] This invention relates to the field of supply chain management, and more specifically, to a method and system for full-level, transparent supervision of the supply chain based on electronic invoices. Background Technology

[0002] In the fields of supply chain finance and tax supervision, achieving thorough verification of the authenticity and compliance of transactions across the entire chain is a core requirement. Currently, regulatory agencies and market entities each hold fragmented information about the same transaction chain; for example, regulators hold information on invoice deduction status, while operators hold purchase contracts and logistics vouchers. Due to system heterogeneity and inconsistent data standards, both parties face challenges in collaborative verification, including difficulties in data alignment, high transmission load, easy leakage of sensitive information, and difficulty in accurately locating related transaction chains. Existing methods mostly rely on self-reporting by enterprises or regulatory sampling, which suffers from low efficiency, limited coverage, and inability to guarantee real-time data consistency across systems, making it difficult to support efficient and accurate full-level penetration supervision of complex supply chain networks. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for full-level, transparent supply chain supervision based on digital electronic tickets.

[0004] In a first aspect, embodiments of the present invention provide a method for full-level, transparent supply chain supervision based on digital invoices, including:

[0005] Identify the supply chain penetration verification items to be regulated; the supply chain penetration verification items need to be completed collaboratively by regulatory and operational nodes, and the regulatory and operational nodes are configured with the target full-chain regulatory correlation data required to execute the supply chain penetration verification items; the target full-chain regulatory correlation data of the regulatory and operational nodes includes different transaction dimension information of the same transaction link, and the same transaction link has a completely consistent electronic ticket number anchor point in the regulatory and operational nodes;

[0006] Based on the rules and constraints of the supply chain penetration verification matters, select the first number of digital invoice number anchor points from the target full-chain regulatory related data of the regulatory side node;

[0007] Based on the coding scale corresponding to the target full-chain regulatory association data of the regulatory side node, the digital ticket number anchor points of the first number of transaction links are encoded to obtain digital ticket number anchor point coding data.

[0008] The digital invoice number anchor data is sent to the operation-side node, so that the operation-side node can decode the digital invoice number anchor data to obtain the digital invoice number anchors of the first number of transaction links, and perform the supply chain penetration verification based on the digital invoice number anchors of the first number of transaction links.

[0009] In a second aspect, embodiments of the present invention provide a server system, including a server, the server being used to execute the method described in the first aspect.

[0010] Compared to existing technologies, the beneficial effects of this invention include: By employing the supply chain full-level penetration supervision method and system based on digital invoices disclosed in this invention, the supply chain penetration verification items requiring collaboration between regulatory and operational nodes are identified. Both systems pre-configure target full-chain regulatory association data containing the same transaction links but different dimensional information, using identical digital invoice numbers as transaction anchors. The regulatory side, based on the rules constraining the verification items, selects a specific number of transaction link invoice number anchors from local data. Subsequently, according to a pre-set encoding scale, these invoice number anchors are encoded to generate compact digital invoice number anchor code data. Finally, this code data is sent to the operational node, where it decodes and reconstructs the specific invoice number anchor, and performs penetration verification accordingly. This invention achieves accurate, secure, and efficient cross-system data collaboration, improving the penetration capability and efficiency of supply chain supervision. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as limiting the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A flowchart illustrating the steps of the supply chain full-level penetration supervision method based on digital invoices provided in this embodiment of the invention;

[0013] Figure 2 A schematic block diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0015] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0016] In order to solve the technical problems mentioned in the background art Figure 1 This is a flowchart illustrating the supply chain full-level penetration supervision method based on digital electronic invoices provided in this embodiment of the disclosure. The following is a detailed description of the supply chain full-level penetration supervision method based on digital electronic invoices.

[0017] Step S201: Determine the supply chain penetration verification items to be regulated; the supply chain penetration verification items need to be completed jointly by the regulatory side node and the operational side node, and the regulatory side node and the operational side node are configured with the target full-chain regulatory association data required to execute the supply chain penetration verification items; the target full-chain regulatory association data of the regulatory side node and the target full-chain regulatory association data of the operational side node include different transaction dimension information of the same transaction link, and the same transaction link has a completely consistent electronic ticket number anchor point in the regulatory side node and the operational side node;

[0018] Step S202: Based on the rule constraints of the supply chain penetration verification items, select the first number of digital invoice number anchor points of the transaction links from the target full-chain regulatory association data of the regulatory side node;

[0019] Step S203: Based on the encoding scale corresponding to the target full-chain regulatory association data of the regulatory side node, the digital ticket number anchor points of the first number of transaction links are encoded to obtain digital ticket number anchor point encoding data.

[0020] Step S204: The digital invoice number anchor point encoding data is sent to the operation side node, so that the operation side node can deencode the digital invoice number anchor point encoding data to obtain the digital invoice number anchor points of the first number of transaction links, and perform the supply chain penetration verification based on the digital invoice number anchor points of the first number of transaction links.

[0021] In this embodiment of the invention, for example, a tax supervision server is used as the execution subject to conduct full-level penetration supervision of the input tax deduction authenticity of the core component procurement supply chain of a large equipment manufacturing group in the first quarter of 2024. The execution process of the supply chain full-level penetration supervision method based on digital invoices is described in detail:

[0022] First, the tax supervision server identifies the supply chain penetration verification items to be supervised. This verification item is the "Verification of the Authenticity of Input Tax Deductions at All Levels of the Core Component Procurement Supply Chain of Large Equipment Manufacturing Groups in the First Quarter of 2024." This item requires collaboration between the regulatory and operational nodes. The regulatory side needs to locate potential risky transaction links through digital invoice anchors, while the operational side needs to provide physical transaction vouchers for the corresponding links to verify the compliance of the deduction behavior. The pre-configured target full-chain regulatory data in the regulatory-side server includes regulatory dimension information such as the invoice number anchor, seller taxpayer identification number, deduction status, transaction amount, and tax period for all input VAT invoices of the group in the first quarter of 2024. The target full-chain regulatory data configured in the group's operational-side supply chain management server includes operational dimension information such as standardized purchase vouchers, warehousing verification records, cross-regional logistics vouchers, original VAT invoices, and supplier qualification documents for the corresponding transaction link. Moreover, the same transaction link shares a completely consistent VAT invoice number anchor in both systems. For example, the transaction of purchasing engine spindles from an upstream core component supplier on January 12, 2024, has the VAT invoice number "2024000011223344556" in both the regulatory-side VAT deduction ledger and the operational-side procurement management ledger, serving as a unique anchor to achieve precise alignment of transaction links across nodes. It is worth noting that in this embodiment, the digital invoice number anchor point is the original invoice number authoritatively generated by the tax authority's digital invoice system. Its generation, distribution, and synchronization mechanism can be as follows: when the tax authority issues a digital invoice, the nationally unified digital invoice number generation engine generates a unique invoice number according to the rule of "administrative division code + date of issuance + 12-digit random sequence code". This invoice number is synchronously pushed to the financial / supply chain management system of the operational node along with the digital invoice PDF / OFD file; the regulatory node directly collects the invoice number data from the core database of the digital invoice system without the need for additional custom synchronization mechanisms, thus naturally ensuring the complete consistency between the regulatory and operational anchor points and eliminating the risk of synchronization conflicts in the distributed system.

[0023] Secondly, based on the rules governing supply chain penetration verification, the tax supervision server selects the first number of electronic invoice number anchors for transaction links from its stored target full-chain supervision-related data. The rule constraint for this verification is "screening transaction links corresponding to input electronic invoices with a deduction amount exceeding 100,000 yuan in the first quarter of 2024 and where the seller is a company registered outside the province." The supervision server first retrieves its stored target full-chain supervision-related data, which includes the electronic invoice number anchors for 1500 input transaction links for the group in the first quarter of 2024. These anchors are generated from basic anchors through anchor status remapping: the supervision server first obtains the basic anchors for the 1500 transactions (i.e., the company's original purchase transaction serial numbers), and then, based on the anchors, generates a rule priority relationship of "60% weight for transaction amount and 40% weight for seller region" in the rule set (in this embodiment, the configuration and acquisition method of the rule priority relationship is: the rule priority relationship is generated by...). The regulatory side sets up configurations based on regulatory needs using a visual configuration tool. Weights can be dynamically adjusted, and the configuration results are stored in the regulatory side's database and synchronized to the operational side via a direct tax-enterprise connection. The operational side can query currently valid rules through the configuration interface, but can only read and not modify them, ensuring consistency between the rules on both sides. 1500 basic anchor points are arranged in descending order of transaction amount and from outside the province to within the province for the seller's region. Then, a progressively associated discrete status identifier value is used to remap the status of the arranged basic anchor points. For example, basic anchor points with transaction amounts ≥ 500,000 yuan are marked with status code S1, 200,000-500,000 yuan with S2, and 100,000-200,000 yuan with S3, generating digital invoice number anchor points that facilitate rule matching and rapid filtering. After anchor point preprocessing, the regulatory server traverses the 1500 digital invoice number anchor points according to rule constraints, ultimately selecting 92 digital invoice number anchor points corresponding to eligible transaction links. These 92 are the first number to be verified in this case.

[0024] Next, in this embodiment of the invention, the "encoding scale" refers to a unified set of encoding input dimensions, output formats, and mapping rules, which includes three core elements: 1. Anchor core extraction rule: uniformly extract the last 10 digits of the digital invoice number as the core dimension of the encoding input (if the length of the digital invoice number is less than 10 digits, pad with 0s to 10 digits on the left); 2. Encoding bit length: preset to 512-bit binary format, which is a fixed encoding length pre-agreed upon by the regulatory side and the operating side; 3. Status identification rule: the first status identification value is "1" (representing associated anchor points), the second status identification value is "0" (representing no associated anchor points), the encoding status identification value is "1" (representing encoded data), and the non-encoding status identification value is "0" (representing reference anchor point baseline data). The encoding scale is pre-configured by the regulatory side node and synchronized to the operating side node through the tax-enterprise direct connection channel to ensure that the encoding / decoding rules on both sides are completely consistent. Based on the encoding scale corresponding to its own target full-chain regulatory associated data, the tax regulatory server encodes 92 digital invoice number anchor points to obtain digital invoice number anchor point encoded data. The coding standard set by the regulatory side is "using the last 10 digits of the electronic ticket number as the coding dimension, and adopting a 512-bit binary coding rule". The regulatory server first divides the 92 electronic ticket number anchors sorted by transaction time into reference anchors and related sequences: it initializes the dynamic anchor carrying unit for temporarily storing related sequence anchors. Since the carrying unit does not carry a valid anchor in the initial state, the regulatory server marks the first anchor after sorting, "2024000011223344556" (corresponding to the transaction of purchasing the main shaft from the core component supplier outside the province), as the reference electronic ticket number anchor. Then, it performs related sequence determination on the next anchor, "2024000011223344557". By comparing the transaction dimension information shared by the nodes on both sides, it is determined that the transaction corresponding to this anchor belongs to the "power system core component procurement supply chain branch" with the reference anchor. Therefore, it is added to the dynamic anchor carrying unit. When processing the third anchor point "2024000011223344558", it was determined that the corresponding transaction belonged to the "Vehicle Control System Procurement Supply Chain Branch", which is an adjacent related sequence with the "Power System Core Components Branch" in the current carrier unit (both belong to the first-level branch of core component procurement). Therefore, the regulatory server determined that the current related sequence determination was completed, reset the dynamic anchor point carrier unit, and added the anchor point corresponding to this control system to the carrier unit as the first related anchor point of the new sequence. When processing the 14th anchor point "2024000011223344569" (corresponding to the transaction of purchasing special steel from raw material suppliers outside the province), it was determined that the corresponding transaction belonged to the "Basic Raw Material Procurement Supply Chain Branch", which has a cross-sequence related relationship with the "Vehicle Control System Branch" in the current carrier unit. Therefore, the regulatory server determined that the current related sequence determination was completed, reset the dynamic anchor point carrier unit, and marked the anchor point corresponding to this steel transaction as the updated reference number electronic ticket number anchor point.

[0025] After completing the division of reference anchors and related sequences, the regulatory server encodes each reference anchor and its related sequences. First, it generates reference anchor encoded data for each anchor based on the encoding scale. For example, the last 10 bits of the reference anchor "2024000011223344556" are "1122334456", which the regulatory server converts into a corresponding 512-bit binary reference code. For the related sequences of each reference anchor, the regulatory server uses a multi-head attention encoding model for encoding processing. The reference anchor is used as the query vector, and each anchor in the related sequence is used as the key and value vectors input into the model. The model calculates the cosine similarity between the query vector and each key vector, and combines dimensions such as transaction amount proportion and transaction frequency to obtain the attention weight of each related anchor relative to the reference anchor. For example, among the 6 related anchors in the core component branch of the power system, 3 spindle procurement transactions account for 70% of the total branch amount, with a corresponding attention weight of 0.75; 2 connecting rod procurement transactions account for 20%, with a weight of 0.2. One sealing component procurement transaction accounts for 10%, with a weight set to 0.05. Based on this attention weight, the value vector is weighted and fused to generate a context-aware anchor feature vector that incorporates the importance of relevant anchor points. This feature vector is then mapped to 512-bit bitmap data according to the encoding scale. Anchor points with higher attention weights are marked with a first state identifier value "1", while those with lower or irrelevant weights are marked with a second state identifier value "0". Simultaneously, encoded anchor bits are added to the bitmap (setting the first bit of the bitmap to an encoded state identifier value "1", indicating that the data is encoded anchor data), ultimately yielding the relevant anchor code data. The regulatory server integrates the reference codes and related sequence codes corresponding to all 92 anchor points to generate a complete electronic ticket number anchor code data package containing information on the 92 transaction link anchor points.

[0026] Finally, the tax supervision server sends the encoded data packet of the digital invoice number anchor to the supply chain management server on the operational side of the large equipment manufacturing group through the encrypted data transmission channel between the tax system and the enterprise's operational side nodes. Upon receiving the data packet, the operational side server first performs decoding according to preset decoding rules: it first identifies the status flag value of the encoded anchor bit to confirm that the data is the encoded anchor data, then performs binary-to-decimal decoding on the reference anchor code to restore the reference digital invoice number anchor; for the bitmap encoding of related sequences, the operational side server reverse maps according to the encoding scale to restore the digital invoice number anchor corresponding to each related sequence, ultimately fully restoring the digital invoice number anchors of 92 transaction links. After decoding, the operations-side server performs supply chain penetration verification based on the 92 digital invoice number anchors: it automatically retrieves standardized purchase vouchers, warehousing verification records, cross-regional logistics vouchers, original digital invoices, and other operational data corresponding to each anchor. For example, for anchor "2024000011223344556", it retrieves the purchase contract template from January 8, 2024, the warehousing quality inspection record from January 16, 2024, the road freight voucher from outside the province to the factory, and the original digital invoice to verify the authenticity of the transaction and the compliance of the deduction. At the same time, the operations-side server packages and encrypts the verification results (including electronic archiving of vouchers, transaction authenticity verification report, and compliance judgment conclusion) and feeds them back to the regulatory-side server. The regulatory server completes full-level penetration supervision of the group's input tax deduction behavior based on the feedback results. If abnormal data is found, the subsequent compliance review process is initiated. It is worth noting that the core technical specifications for cross-node collaboration in this embodiment are as follows: 1. Interface Specification: The standardized API interface for direct connection between tax authorities and enterprises is adopted, supporting the RESTful protocol; 2. Security Mechanism: The SM2 asymmetric algorithm is used to sign and verify data packets, and the SM4 symmetric algorithm is used to encrypt data content. The key is synchronized monthly by both tax authorities and enterprises through the tax digital certificate system; 3. Transaction Consistency and Error Handling: Distributed transactions are implemented based on the RabbitMQ message queue. If the operating node does not return confirmation information within 5 minutes, the regulatory side automatically triggers a retry (up to 3 times); if decoding or verification fails, the operating node synchronizes an exception log containing error codes and error descriptions to the regulatory side. The regulatory side triggers manual intervention or data correction processes according to the exception type (such as decoding failure or data missing).

[0027] It is worth noting that the generation and synchronization of the aforementioned "completely consistent digital invoice number anchor" relies on a core system uniformly maintained by the national tax authorities. The specific mechanism is as follows: When issuing digital invoices, the tax authorities generate a globally unique invoice number through a nationally unified number generation system, according to predetermined coding rules (usually including elements such as administrative divisions, timestamps, and random sequences). This invoice number, as a data element, is simultaneously written to two channels: first, it is embedded in the electronic file of the digital invoice and sent to the invoice recipient (operating side) through a standard interface; second, it is synchronized in real time to the tax data warehouse for collection by the regulatory system. This "source generation, two-way distribution" model ensures that the invoice number anchors held by the regulatory side and the operating side for the same transaction are naturally consistent, requiring no additional collaborative verification.

[0028] The "coding scale" is a set of data conversion rules pre-agreed between the regulatory and operational sides, used to convert the anchor sequence of digital invoice numbers into compact coded data. It mainly consists of the following three configurable components:

[0029] Feature extraction rules: These rules specify which parts of the original ticket number should be extracted as the basic features for encoding. For example, it can be agreed that the last few digits of the ticket number should be extracted, or that the result of a specific hash operation on the complete ticket number should be used as the feature value.

[0030] Encoding format and length: The final form of the encoded output data (such as a binary bitmap, a string hash value of a specific length) and its fixed length (such as 256 bits, 512 bits) are agreed upon. This length is preset according to the balance requirements of data security and efficiency.

[0031] State mapping dictionary: Defines the specific numerical markers corresponding to different logical states (such as "related anchors", "unrelated anchors", "encoding identifiers", and "baseline identifiers"). For example, in a binary bitmap, it can be agreed that "1" represents a related anchor mapping bit and "0" represents an unmapped bit.

[0032] The "related sequence determination" relies on an initializeable and updatable supply chain category knowledge graph. This knowledge graph is organized in a tree structure, with the root node representing the industry and each level of child nodes representing product categories from broad categories to specific subcategories. Each node has a unique code. The determination process is as follows:

[0033] (1) Product name recognition: Extract core product names from the "Name of Goods or Taxable Services" field of the electronic invoice through keyword matching or simple natural language processing model.

[0034] (2) Graph matching: The extracted product names are matched with the nodes in the knowledge graph to find the corresponding finest classification node.

[0035] (3) Identifier generation: trace back from the matched node to the root node, combine the codes of each node on the path in hierarchical order to generate the "sequence hierarchical identifier" corresponding to the ticket number. For example, the identifier "L1-03-L2-08" may represent "secondary category (08) under primary category (03)".

[0036] (4) Relationship determination: The relationship is determined by comparing the similarity of the "sequence level identifiers" of two ticket numbers. The specific rule is: calculate the length of the continuous identical part of the two identifier coding sequences from the beginning. If the length reaches the preset threshold (e.g., level 2), it is determined to be the same related sequence; if the length is 1 level less than the threshold, it is determined to be an adjacent sequence; otherwise, it is determined to be a cross sequence.

[0037] In this embodiment of the invention, the encoding scale corresponding to the target full-chain regulatory association data of the regulatory side node is used to encode the digital ticket number anchor points of the first number of transaction links to obtain digital ticket number anchor point encoding data. This can be implemented through the following example.

[0038] From the first number of electronic ticket number anchor points, determine a reference number of electronic ticket number anchor points and a related number of electronic ticket number anchor point sequence of the reference number of electronic ticket number anchor points, wherein the related number of electronic ticket number anchor point sequence of the reference number of electronic ticket number anchor points includes the related number of electronic ticket number anchor points of the reference number of electronic ticket number anchor points.

[0039] Based on the encoding scale, determine the reference anchor point encoding data corresponding to the reference number electronic ticket number anchor point;

[0040] Based on the encoding scale, and taking the reference electronic ticket number anchor point as a reference, the relevant electronic ticket number anchor point sequence of the reference electronic ticket number anchor point is encoded to obtain the relevant anchor point encoded data corresponding to the relevant electronic ticket number anchor point sequence of the reference electronic ticket number anchor point.

[0041] The digital ticket number anchor point encoding data includes reference anchor point encoding data corresponding to the reference digital ticket number anchor point, and related anchor point encoding data corresponding to the related digital ticket number anchor point sequence of the reference digital ticket number anchor point.

[0042] In this embodiment of the invention, for example, the tax supervision server initiates an encoding process for the 92 selected electronic invoice number anchors: First, the server initializes the dynamic anchor carrying unit and processes the anchors sorted by transaction time sequentially. Initially, the carrying unit has no valid anchors, so the server marks the first anchor "2024000011223344556" (corresponding to the procurement of the main shaft of the power system) as a reference electronic invoice number anchor. Then, it verifies the next anchor "2024000011223344557," comparing the commodity name on the invoice and the upstream and downstream enterprise relationships to determine that it belongs to the same "power system core component procurement branch" as the reference anchor, adding it to the carrying unit. The subsequent three anchors in the same branch are also added sequentially, forming a sequence of related electronic invoice number anchors for the reference anchor. When processing the 6th anchor point "2024000011223344561", it was determined that it belonged to the "Vehicle Control System Procurement Branch". The server reset the bearer unit and marked this anchor point as a new reference anchor point. This logic was repeated to complete the grouping of 92 anchor points, ultimately resulting in 7 groups of reference anchor points and their corresponding sequences.

[0043] Next, the server generates reference anchor code data based on the preset 512-bit binary encoding scale: extract the last 10 bits "11223344556" of the reference anchor "2024000011223344556", convert each digit into an 8-bit binary value and concatenate them to generate 512-bit binary reference anchor code data, ensuring that the code uniquely corresponds to the digital ticket number anchor.

[0044] Subsequently, the server encodes the relevant sequence based on the reference anchor point: according to the encoding scale, it calculates the numerical difference between the last 10 bits of each relevant anchor point and the reference anchor point, and determines its mapping position in the 512-bit encoding bitmap. For example, the last 10 bits of the relevant anchor point "2024000011223344557" are 1 greater than the reference value, corresponding to the 2nd bit in the bitmap; in the bitmap, the corresponding position of the relevant anchor point is marked as the first state value "1", and the remaining positions are marked as the second state value "0", generating basic anchor point encoding data; then the first bit of the bitmap (encoding anchor point bit) is marked as the encoding state value "1", indicating that the data is the encoded content, and finally the relevant anchor point encoding data of the reference anchor point is obtained.

[0045] After the server completes the encoding of all 7 sets of reference anchors and their corresponding sequences, it integrates all the reference anchor encoding data with the relevant anchor encoding data to form a complete digital ticket number anchor encoding data packet containing 92 transaction link information.

[0046] In this embodiment of the invention, a dynamic anchor point carrying unit is configured in the process of determining the reference electronic ticket number anchor point and the related electronic ticket number anchor point sequence of the reference electronic ticket number anchor point. The dynamic anchor point carrying unit is used to temporarily store the related electronic ticket number anchor points in the related electronic ticket number anchor point sequence of the reference electronic ticket number anchor point. The process of determining the reference electronic ticket number anchor point and the related electronic ticket number anchor point sequence of the reference electronic ticket number anchor point from a first number of electronic ticket number anchor points can be implemented through the following example.

[0047] The first number of electronic ticket number anchor points are arranged in order, and the first number of electronic ticket number anchor points after the order arrangement are processed one by one; among the first number of electronic ticket number anchor points after the order arrangement, the electronic ticket number anchor point currently being processed is the target number of electronic ticket number anchor points;

[0048] If the dynamic anchor point carrying unit does not carry a valid anchor point, then the target number electronic ticket number anchor point is marked as the reference number electronic ticket number anchor point, and the subsequent number electronic ticket number anchor points are re-sequenced based on the target number electronic ticket number anchor point to obtain the re-sequenced result of the subsequent number electronic ticket number anchor point; the subsequent number electronic ticket number anchor point is the next number electronic ticket number anchor point of the target number electronic ticket number anchor point;

[0049] If the relevant sequence determination result of the subsequent number electronic ticket number anchor point indicates that the subsequent number electronic ticket number anchor point belongs to the relevant number electronic ticket number anchor point sequence of the target number electronic ticket number anchor point, then the subsequent number electronic ticket number anchor point is added to the dynamic anchor point carrying unit.

[0050] If the relevant sequence determination result of the subsequent number electronic ticket number anchor point indicates that the subsequent number electronic ticket number anchor point does not belong to the relevant number electronic ticket number anchor point sequence of the target number electronic ticket number anchor point, then the subsequent number electronic ticket number anchor point is determined as the updated reference number electronic ticket number anchor point.

[0051] In an embodiment of the present invention, for example, the tax supervision server initiates a process of collecting reference anchors and related sequences for the 92 selected electronic invoice number anchors. First, it initializes a dynamic anchor carrying unit (a dedicated cache area in the server's memory) for temporarily storing anchors of the same sequence. Then, it arranges the 92 anchors in ascending order according to the transaction time from January 1, 2024 to March 31, 2024 to generate an ordered anchor list. The anchors in the list are processed one by one, and the anchors currently being processed are uniformly marked as target electronic invoice number anchors.

[0052] Initially, the dynamic anchor point carrying unit does not carry any valid anchor points. The server processes the first target anchor point "2024000011223344556" (corresponding to the power system spindle procurement transaction on January 5th) and directly marks it as the reference electronic invoice number anchor point. Subsequently, the server retrieves the next subsequent electronic invoice number anchor point "2024000011223344557" (corresponding to the linkage procurement transaction in the same supply chain branch on January 8th) and initiates relevant sequence determination: based on the preset supply chain level mapping rules, the "Goods or Taxable Services Name" field of the electronic invoices corresponding to the two anchor points is parsed. The level of the reference anchor point is "Power System Core Components > Spindle", and the level of the subsequent anchor point is "Power System Core Components > Linkage". Both belong to the first-level supply chain branch of "Power System Core Components". Therefore, it is determined that the subsequent anchor point belongs to the relevant electronic invoice number anchor point sequence of the reference anchor point. The server writes the subsequent anchor point into the cache area of ​​the dynamic anchor point carrying unit for temporary storage.

[0053] The server continues processing the next target anchor point, "2024000011223344561" (corresponding to the vehicle control system chip procurement transaction on January 12th). Based on rule analysis, the anchor point belongs to the hierarchy of "Vehicle Control System > Core Chip," which is a cross-level supply chain branch with the current reference anchor point's "Powertrain Core Components." The determination result indicates that it does not belong to the relevant data invoice number anchor point sequence of the current reference anchor point. The server then clears the cached data of the dynamic anchor point carrying unit, determines the anchor point corresponding to the vehicle control system chip procurement as the updated reference data invoice number anchor point, and starts a new set of related sequence aggregation processes.

[0054] In this embodiment of the invention, the step of determining a reference electronic ticket number anchor point and a related electronic ticket number anchor point sequence from a first number of electronic ticket number anchor points is further provided with the following implementation method.

[0055] If the dynamic anchor point carrying unit carries a valid anchor point, then the relevant sequence determination is performed on the target number of electronic ticket number anchor points to obtain the relevant sequence determination result of the target number of electronic ticket number anchor points; when the currently processed number of electronic ticket number anchor points is the target number of electronic ticket number anchor points, the dynamic anchor point carrying unit temporarily stores the relevant number of electronic ticket number anchor points of the target relevant number of electronic ticket number anchor point sequence;

[0056] If the determination result of the relevant sequence of the target number electronic ticket number anchor point indicates that the target number electronic ticket number anchor point belongs to the target relevant number electronic ticket number anchor point sequence, then the target number electronic ticket number anchor point is added to the dynamic anchor point carrying unit;

[0057] If the determination result of the relevant sequence of the target number electronic ticket number anchor point indicates that the target number electronic ticket number anchor point belongs to the adjacent relevant number electronic ticket number anchor point sequence of the target relevant number electronic ticket number anchor point sequence, then it is determined that the determination of the relevant number electronic ticket number anchor point of the target relevant number electronic ticket number anchor point sequence is completed, the dynamic anchor point carrying unit is reset, and the target number electronic ticket number anchor point is added to the dynamic anchor point carrying unit as a relevant number electronic ticket number anchor point in the updated relevant number electronic ticket number anchor point sequence;

[0058] If the determination result of the relevant sequence of the target number electronic ticket number anchor point indicates that the target number electronic ticket number anchor point belongs to the cross-sequence relevant number electronic ticket number anchor point sequence of the target relevant number electronic ticket number anchor point sequence, then the determination of the relevant number electronic ticket number anchor point of the target relevant number electronic ticket number anchor point sequence is completed, the dynamic anchor point carrying unit is reset, and the target number electronic ticket number anchor point is used as the updated reference number electronic ticket number anchor point.

[0059] In an embodiment of the present invention, for example, the tax supervision server is processing 92 electronic invoice number anchors sorted by transaction time. At this time, the dynamic anchor carrying unit temporarily stores three target-related electronic invoice number anchor sequence anchors belonging to "procurement of core components of power system", which are in the state of carrying valid anchors. The server performs the following processing on subsequent target electronic invoice number anchors in sequence:

[0060] First, the server processes the target anchor point "2024000011223344558" and initiates the relevant sequence determination process: it calls the supply chain level mapping rules, parses the "goods attribute tag" field of the electronic invoice corresponding to the anchor point, and determines that its supply chain level is "core component procurement > power system core components > sealing components," which completely matches the first-level level "core component procurement > power system core components" of the target-related sequence in the dynamic carrying unit. The determination result clearly indicates that the target anchor point belongs to the target-related electronic invoice number anchor point sequence. The server then performs a write operation, appending the anchor point to the buffer area of ​​the dynamic anchor point carrying unit, completing the aggregation of anchor points in the same sequence.

[0061] Next, the server processes the next target anchor point "2024000011223344561" and initiates the relevant sequence determination process: it parses the goods attribute tag of the electronic ticket corresponding to this anchor point, determines that its supply chain level is "core component procurement > vehicle control system > core chip", and that it belongs to the same parallel branch under the first-level supply chain of "core component procurement" as the "power system core components" in the target-related sequence in the dynamic carrier unit. The determination result indicates that it belongs to the adjacent related electronic ticket number anchor point sequence of the target-related electronic ticket number anchor point sequence. The server immediately generates a sequence end marker, marking the completion of the anchor point determination for the current "power system core component procurement" sequence, calls the cache clearing instruction to reset the dynamic anchor point carrier unit, and then writes the vehicle control system chip procurement anchor point as the first anchor point in the updated relevant sequence into the reset dynamic anchor point carrier unit.

[0062] Finally, the server processes the target anchor point "2024000011223344572" (corresponding to a special steel raw material procurement transaction), initiating the relevant sequence determination process: parsing the goods attribute tag of the electronic invoice corresponding to this anchor point, determining its supply chain level as "basic raw material procurement > special steel," and confirming that it does not belong to the same first-level supply chain branch as the current "vehicle control system" sequence in the dynamic carrier unit. The determination result indicates that it belongs to a cross-sequence related electronic invoice number anchor point sequence of the target related electronic invoice number sequence. The server generates a sequence end marker, marking the completion of the anchor point determination for the current "vehicle control system procurement" sequence, executes the operation of resetting the dynamic anchor point carrier unit, and then directly marks the special steel procurement anchor point as the updated reference electronic invoice number anchor point, initiating a new set of supply chain sequence anchor point aggregation process.

[0063] In this embodiment of the invention, the step of performing a correlation sequence determination on the target number of electronic ticket number anchor points to obtain the correlation sequence determination result of the target number of electronic ticket number anchor points can be implemented through the following example.

[0064] Obtain the sequence level identifier of the target-related electronic ticket number anchor sequence;

[0065] The target number of electronic ticket number anchors, the current reference number of electronic ticket number anchors, and the coding scale are subjected to rule parsing processing to determine the sequence level identifier to which the target number of electronic ticket number anchors belong;

[0066] If the sequence level identifier of the target related number electronic ticket number anchor point sequence is the same as the sequence level identifier of the target number electronic ticket number anchor point, then a related sequence determination result of the target number electronic ticket number anchor point is generated. The related sequence determination result of the target number electronic ticket number anchor point indicates that the target number electronic ticket number anchor point belongs to the target related number electronic ticket number anchor point sequence.

[0067] If the hierarchical spacing between the sequence level identifier of the target related number electronic ticket number anchor point sequence and the sequence level identifier to which the target number electronic ticket number anchor point belongs is equal to the preset hierarchical discrimination boundary, then the related sequence determination result of the target number electronic ticket number anchor point is generated. The related sequence determination result of the target number electronic ticket number anchor point indicates that the target number electronic ticket number anchor point belongs to the adjacent related number electronic ticket number anchor point sequence of the target related number electronic ticket number anchor point sequence.

[0068] If the hierarchical distance between the sequence level identifier of the target related number electronic ticket number anchor sequence and the sequence level identifier to which the target number electronic ticket number anchor belongs is greater than a preset hierarchical discrimination boundary, then a related sequence determination result for the target number electronic ticket number anchor is generated. The related sequence determination result for the target number electronic ticket number anchor indicates that the target number electronic ticket number anchor belongs to a cross-sequence related number electronic ticket number anchor sequence of the target related number electronic ticket number anchor sequence.

[0069] In this embodiment of the invention, the lightweight graph neural network model, for example, adopts a simplified version of the GraphSAGE model. The specific implementation details are as follows: 1. Network structure: 2 layers of graph convolutional layers, each with an output dimension of 16, and the activation function is ReLU; 2. Training method: Using historical supply chain transaction relationship graphs as training data, the cross-entropy loss function is used for 300 training rounds until the loss value is below 0.02; 3. Message passing mechanism: An average aggregation method is used to aggregate the features of the target node's neighboring nodes and edge weights (edge ​​weights are calculated based on ticket information similarity (40%), upstream and downstream correlation (30%), and transaction time interval (30%), with a weight range of 0-1); 4. Attribution probability determination threshold: When the attribution probability of a certain class is ≥80%, the target node is directly determined to belong to that class sequence.

[0070] The technical details of the sequence hierarchy identification and judgment rules are as follows: 1. Hierarchical identification assignment rules: The format of "L + number of levels + 2-digit branch code" is adopted, where L1 is the first-level supply chain branch (such as "core component procurement" and "basic raw material procurement"), and L2 is the second-level branch (such as "core components of the power system" and "vehicle control system"). The branch code is generated by matching the core category words of "name of goods or taxable services" on the electronic invoice with the supply chain branch mapping dictionary pre-configured by the regulatory side (for example, "main shaft" is mapped to 01, "connecting rod" is mapped to 02, and "core chip" is mapped to 03); 2. Hierarchical spacing calculation method: If two identifiers belong to different L2 levels under the same L1 level, the hierarchical spacing is 1; if they belong to different L1 levels, the hierarchical spacing is 2; if they belong to the same L2 level, the hierarchical spacing is 0; 3. Preset hierarchical discrimination boundary: The specific value is 1, which is pre-configured by the regulatory side according to the supply chain supervision requirements and synchronized to the operation side. The tax supervision server initiates the relevant sequence determination process for the target number of electronic invoice number anchor points. The first step is to directly read the sequence level identifier L2-01 of the currently temporarily stored target-related number of electronic invoice number anchor point sequence from the metadata cache area of ​​the dynamic anchor point carrying unit. This identifier corresponds to the second-level supply chain branch of "core component procurement > power system core components" and is a fixed identifier that the server synchronously writes to the metadata cache when collecting this sequence anchor point.

[0071] Subsequently, the server performs rule parsing to determine the sequence level identifier of the target electronic invoice number anchor point: it retrieves the electronic invoice data corresponding to the target anchor point "2024000011223344561" and parses the "goods attribute label" field to obtain "vehicle control system core chip"; at the same time, it combines the coding scale rules preset by the regulatory side, which clearly defines the mapping relationship between the last two digits of the electronic invoice number and the supply chain level, where 01 corresponds to the power system branch, 02 corresponds to the vehicle control system branch, and 03 corresponds to the basic raw materials branch. The last two digits of this target anchor point are 61, corresponding to the mapping code 02. Therefore, the sequence level identifier of this target anchor point is determined to be L2-02, corresponding to the second-level supply chain branch of "core component procurement > vehicle control system".

[0072] Next, the server calculates the spacing relationship between the two hierarchical identifiers: the hierarchical identifier L2-01 of the target related sequence and the hierarchical identifier L2-02 of the target anchor point both belong to the L1 level "core component procurement" supply chain branch, with a branch spacing of 1. The hierarchical discrimination boundary preset by the regulatory side is 1, and the two values ​​are equal. Therefore, the server directly generates a judgment result, clearly indicating that the target number electronic ticket number anchor point belongs to the adjacent related number electronic ticket number anchor point sequence of the target related number electronic ticket number anchor point sequence.

[0073] If the target anchor point processed by the server is "2024000011223344572", after rule parsing, its level identifier is obtained as L2-03, corresponding to the second-level branch of "basic raw material procurement > special steel". The L2-01 of the target-related sequence belongs to the L1 level of "core component procurement" and "basic raw material procurement". The branch spacing is 2, which is greater than the preset level discrimination boundary of 1. The server then generates a judgment result, which clearly indicates that the target anchor point belongs to the cross-sequence related number anchor point sequence of the target-related number anchor point sequence.

[0074] If the target anchor point processed by the server is "2024000011223344558", its hierarchical identifier is obtained after rule parsing as L2-01, which is completely consistent with the hierarchical identifier of the target-related sequence. The server directly generates a judgment result, clearly indicating that the target anchor point belongs to the target-related electronic ticket number anchor point sequence.

[0075] In this embodiment of the invention, the step of encoding the relevant electronic ticket number anchor sequence of the reference electronic ticket number anchor based on the encoding scale and with the reference electronic ticket number anchor as a reference, to obtain the relevant anchor code data corresponding to the relevant electronic ticket number anchor sequence of the reference electronic ticket number anchor, can be implemented through the following example.

[0076] Obtain the encoded bitmap with the scale of the encoding scale;

[0077] Based on the reference electronic ticket number anchor point, the sequence level identifier of the related electronic ticket number anchor point sequence of the reference electronic ticket number anchor point, and each related electronic ticket number anchor point in the related electronic ticket number anchor point sequence of the reference electronic ticket number anchor point, determine the regular mapping position of the encoding bitmap corresponding to each related electronic ticket number anchor point;

[0078] In the coded bitmap with the scale of the coded scale, the coded bitmap corresponding to each relevant digital ticket number anchor point is marked as the first state identifier value, and the remaining coded bitmaps other than the coded bitmaps corresponding to each relevant digital ticket number anchor point are marked as the second state identifier value, so as to obtain the basic anchor point coded data of the relevant digital ticket number anchor point sequence of the reference digital ticket number anchor point;

[0079] Add coded anchor bits to the basic anchor bit encoded data of the relevant electronic ticket number anchor sequence of the reference electronic ticket number anchor point to obtain the relevant anchor bit encoded data corresponding to the relevant electronic ticket number anchor sequence of the reference electronic ticket number anchor point; the coded anchor bit is marked as an encoding status identifier value, and the encoding status identifier value indicates that the relevant anchor bit encoded data is encoded data.

[0080] In an embodiment of the present invention, for example, the tax supervision server performs encoding processing on the relevant electronic invoice number anchor sequence of the reference anchor point "procurement of core components of power system". First, it directly retrieves the standardized encoding bitmap with a scale of 512 bits preset by the supervision side. This bitmap is a fixed mapping carrier built based on the last 10 bits of the electronic invoice number and is the unified encoding scale corresponding to the full-chain supervision related data of the supervision side.

[0081] Subsequently, the server determines the rule mapping positions of each relevant anchor point: First, it reads the last 10 core values ​​of the reference number electronic ticket number anchor point "2024000011223344556" "1122334456", and combines them with the fixed offset of 0 corresponding to the hierarchical identifier L2-01 of the sequence. The mapping relationship is calculated one by one for the three relevant number electronic ticket number anchor points in the sequence: The last 10 digits of the first relevant anchor point are "1122334457", which has a difference of 1 from the reference value and corresponds to the 2nd bit of the encoded bitmap; the second relevant anchor point has a difference of 2 from the reference value and corresponds to the 3rd bit; the third relevant anchor point has a difference of 3 from the reference value and corresponds to the 4th bit. All positions are determined strictly according to the "difference + 1" mapping rule of the encoding scale.

[0082] Next, the server completes the generation of basic anchor point encoding data: in the 512-bit encoding bitmap, the 2nd, 3rd, and 4th bits are marked as the first state identifier value "1", and the remaining 509 irrelevant bits are marked as the second state identifier value "0", clearly distinguishing the mapping position of relevant anchor points from irrelevant areas.

[0083] Finally, the server adds encoding anchor bits to the basic encoded data: the 0th bit of the encoding bitmap (the preset encoding identifier bit) is marked with the encoding status identifier value "1", which clearly indicates that the data is the encoded related anchor sequence information, forming a clear distinction from the separately encoded reference anchor data, and finally generating the complete related anchor encoded data corresponding to the related data of the reference anchor number anchor sequence.

[0084] In this embodiment of the invention, the step of determining the reference anchor point encoding data corresponding to the reference number electronic ticket number anchor point according to the encoding scale can be implemented through the following example.

[0085] Based on the encoding scale, the reference number electronic ticket number anchor point is mapped to obtain the basic anchor point encoding data of the reference number electronic ticket number anchor point;

[0086] Add coded anchor bits to the basic anchor bit coded data of the reference number electronic ticket number anchor point to obtain the reference anchor bit coded data corresponding to the reference number electronic ticket number anchor point; the coded anchor bit is marked as a non-coded state identifier value, the non-coded state identifier value indicates that the reference anchor bit coded data is uncoded data.

[0087] In an embodiment of the present invention, for example, the tax supervision server generates reference anchor code data for the reference number electronic invoice number anchor "2024000011223344556" in the "Procurement of Core Components of Power System" sequence: First, according to the 512-bit binary encoding scale preset by the supervision side, a mapping process is performed to extract the last 10 core values ​​of the reference anchor "1122334456". Each digit is converted into an 8-bit binary value according to the encoding scale rules (e.g., digit 1 is converted into 00000001, digit 2 is converted into 00000010). All the converted binary values ​​are concatenated in sequence to generate basic anchor code data with a length of 512 bits. Subsequently, the server adds a preset encoding anchor bit (bit 0 of the encoding bitmap) to the basic anchor bit encoded data and marks the encoding anchor bit with a non-encoded state identifier value "0", which clearly indicates that the data is an unencoded reference anchor bit data, so that the operator can quickly identify its attribute as a full sequence mapping reference when decoding, and finally obtain the complete reference anchor bit encoded data corresponding to the reference number of the electronic ticket anchor bit.

[0088] In this embodiment of the invention, the method also provides the following implementation methods.

[0089] Based on the reference electronic ticket number anchor point and the related electronic ticket number anchor point sequence of the reference electronic ticket number anchor point, the first number of electronic ticket number anchor points are encoded and determined.

[0090] If the first number of electronic ticket number anchor points meet the encoding conditions, then the following steps are triggered: determining the reference anchor point encoding data corresponding to the reference number electronic ticket number anchor point according to the encoding scale; and according to the encoding scale, using the reference number electronic ticket number anchor point as a reference, encoding the relevant number electronic ticket number anchor point sequence of the reference number electronic ticket number anchor point to obtain the relevant anchor point encoding data corresponding to the relevant number electronic ticket number anchor point sequence of the reference number electronic ticket number anchor point.

[0091] If the first number of electronic ticket number anchor points does not meet the encoding conditions, then according to the encoding scale, the first number of electronic ticket number anchor points are mapped to obtain the encoded data of the first number of electronic ticket number anchor points.

[0092] The digital ticket number anchor point encoding data includes the encoding data of the first number of digital ticket number anchor points.

[0093] In this embodiment of the invention, for example, the tax supervision server executes an encoding determination process for 92 electronic invoice number anchors that have been divided into reference anchors and related sequences: It calls preset encoding condition rules, requiring that the total number of anchors in a single group of reference anchors and related sequences be ≥3, and verifies each group one by one. Specifically, the "Power System Core Component Procurement" group contains 1 reference anchor and 3 related anchors, totaling 4, which meets the encoding conditions; the "Precision Fastener Procurement" group contains 1 reference anchor and 1 related anchor, totaling 2, which does not meet the encoding conditions.

[0094] For the "Power System Core Component Procurement" group that meets the coding conditions, the server triggers the standard coding process: First, based on the 512-bit binary coding scale, the reference anchor point "2024000011223344556" is converted into the corresponding basic code, and a coding anchor point marked with a non-coding status identifier value "0" is added to it to generate reference anchor point coding data; then, based on this reference anchor point, the three related anchor points are mapped, the position is marked, and coding anchor points are added (marked with a coding status identifier value "1") to generate related anchor point coding data.

[0095] For the "Precision Fastener Procurement" group that does not meet the coding conditions, the server directly executes a separate mapping process: extract the last 10 core values ​​of the two anchor points in the group, convert each digit into an 8-bit binary value according to the coding scale and concatenate them to generate a 512-bit basic code, add a coding anchor point marked with the mixed state identifier value "2" to each code, and clarify that it is a separately coded anchor point data.

[0096] Finally, the server integrates the encoded data of all groups to form a complete digital ticket number anchor encoded data packet containing reference anchor encoded data, related anchor encoded data, and individual encoded data, ensuring that no information on the 92 transaction link anchor points is omitted.

[0097] In this embodiment of the invention, the step of encoding and determining the first number of electronic ticket number anchors based on the reference electronic ticket number anchor and the related electronic ticket number anchor sequence of the reference electronic ticket number anchor includes any one of the following:

[0098] If none of the relevant electronic ticket number anchor points exist in the relevant electronic ticket number anchor point sequence of each of the aforementioned reference electronic ticket number anchor points, then it is determined that the first number of electronic ticket number anchor points do not meet the encoding conditions.

[0099] If there is a related electronic ticket number anchor in the sequence of related electronic ticket number anchors for each of the aforementioned reference electronic ticket number anchors, then it is determined that the first number of electronic ticket number anchors do not meet the encoding conditions.

[0100] In an exemplary embodiment of the present invention, the tax supervision server performs encoding determination on the anchor points of the two batches of selected electronic invoice numbers respectively:

[0101] For the first batch of 12 anchor points, after the server completed sequence aggregation, it was found that there were no related anchor points in the relevant electronic invoice number anchor point sequence for each reference electronic invoice number anchor point. For example, the reference anchor point "2024000011223344560" corresponds to the purchase of special bearings outside the province. All subsequent anchor points belong to isolated transactions of different raw material categories and have no anchor points related to the same supply chain branch. All 12 anchor points are independent reference anchor points with no matching related sequences. The server directly determined that these 12 electronic invoice number anchor points do not meet the coding conditions.

[0102] For the second batch of 18 anchor points, after the server completed sequence aggregation, it was found that each reference number of electronic ticket number anchor point had only one related anchor point in its related sequence. For example, the reference anchor point "2024000011223344581" corresponds to the purchase of motor stators from outside the province, and the related sequence only contains one anchor point for the purchase of motor rotors from the same branch. The related sequences of the 9 sets of reference anchor points all contain only one anchor point, which meets the judgment rules. The server then determined that the 18 number of electronic ticket number anchor points did not meet the coding conditions and triggered a separate mapping coding process to complete independent coding for each anchor point according to the coding scale.

[0103] In this embodiment of the invention, the target full-chain regulatory association data of the regulatory side node includes a second number of digital invoice number anchors for transaction links. The second number of digital invoice number anchors for transaction links are obtained by remapping the anchor states of the basic anchors of the second number of transaction links. The second number is not less than the first number. The method also provides the following implementation methods.

[0104] Obtain the base anchor points of the second number of transaction links, which are generated based on the anchor point generation elements in the anchor point generation rule set;

[0105] Based on the rule priority relationship in the anchor point generation rule set, the basic anchor points of the second number of transaction links are arranged in order to obtain the second number of transaction links in order.

[0106] Based on the anchor point mapping order of the second number of transaction links' basic anchor points arranged in sequence, the anchor point states of the second number of transaction links are remapped using progressively associated discrete state identifier values ​​to obtain the digital ticket number anchor points of the second number of transaction links.

[0107] In an embodiment of the present invention, for example, the tax supervision server executes a process for generating digital invoice number anchors for the input transaction data of a large equipment manufacturing group in the first quarter of 2024:

[0108] First, the server retrieves the second batch (1500) of basic anchor points for the transaction links. These basic anchor points are generated based on a set of anchor point generation rules preset by the regulatory side. The anchor point generation elements in the rule set include the enterprise's original purchase transaction serial number, the real-time generation time of the electronic invoice, and the transaction amount including tax. The server directly retrieves the basic anchor points corresponding to the group's 1500 input transactions from the tax administration database. Each anchor point is in the format "CGYYYYMMDDNNN" (e.g., CG20240105012), which perfectly matches the format requirements of the anchor point generation elements.

[0109] Secondly, the server sorts the 1500 basic anchor points in order according to the rule priority relationship in the anchor point generation rule set. The rule priority relationship is clearly defined as "transaction amount including tax (weight 60%) > electronic invoice generation time (weight 30%) > seller region (weight 10%)". The server first sorts the basic anchor points by transaction amount from high to low, and if the amounts are the same, it sorts them by electronic invoice generation time from early to late. If the first two are completely consistent, it sorts them by seller region (outside the province first, inside the province second), and finally obtains a list of 1500 basic anchor points sorted by priority.

[0110] Finally, the server uses progressively associated discrete state identifier values ​​for anchor state remapping. The anchor point mapping order strictly follows the sorted sequence. The discrete state identifier values ​​are divided into S1 (transaction amount ≥ 500,000 yuan), S2 (200,000-500,000 yuan), S3 (100,000-200,000 yuan), and S4 (< 100,000 yuan). The progressive association is reflected in the progressive switching of the identifier value according to the sorting position: the server traverses the sorted basic anchor points, marks the first 212 basic anchor points with an amount ≥ 500,000 yuan as S1, the middle 426 basic anchor points with an amount ≥ 500,000 yuan as S2, the next 612 basic anchor points with an amount ≥ 200,000 yuan as S3, and the last 50 basic anchor points with an amount < 100,000 yuan as S4; at the same time, each basic anchor point is associated with the last 10 digits of the corresponding digital bill number to generate a digital bill number anchor point in the format of "state identifier-basic anchor point-last 10 digits of digital bill". Finally, 1,500 digital bill number anchor points that have completed state remapping are obtained. Among them, the 92 anchor points to be verified in the subsequent screening (the first number) all come from this set, which meets the requirement that the second number is not less than the first number.

[0111] In this embodiment of the invention, in the regulatory side node and the operational side node, the full-chain regulatory related data required to perform the supply chain penetration verification is collected into different regulatory data ranges. The full-chain regulatory related data corresponding to the regulatory data range includes different transaction dimension information of the same transaction link, and there is a completely consistent verification range anchor point in the corresponding regulatory data range. The target full-chain regulatory related data of the regulatory side node is the full-chain regulatory related data included in the target regulatory data range of the regulatory side node. The method also provides the following implementation methods.

[0112] Send the verification scope anchor point of the target regulatory data range to the operating side node; so that the operating side node can obtain the target regulatory data range of the operating side node based on the verification scope anchor point of the target regulatory data range, and obtain the transaction dimension information of the first number of transaction links from the target full-chain regulatory related data of the operating side node included in the target regulatory data range of the operating side node according to the digital invoice number anchor point of the first number of transaction links, and perform supply chain penetration verification operation on the transaction dimension information of the first number of transaction links according to the rule constraints of the supply chain penetration verification items, so as to execute the supply chain penetration verification items.

[0113] In an embodiment of the present invention, for example, after the tax supervision server completes the encoding processing of 92 electronic invoice number anchor points for transaction links, it initiates the regulatory data scope alignment process: First, it calls the regulatory data scope metadata stored in its own database to extract the target regulatory data scope verification scope anchor point "CF2024Q1-CL01" corresponding to the "2024 Q1 Core Component Input Tax Deduction Authenticity Penetration Verification". This anchor point is a unified identifier pre-agreed upon by the regulatory side and the operating side, corresponding to two consistent regulatory data scopes: the regulatory side's "2024Q1 Core Component Input Tax Deduction Ledger" and the operating side's "2024Q1 Core Component Procurement Full-Link Data". Both sides' data scopes contain different dimensions of information for the same transaction link. The regulatory side focuses on regulatory dimensions such as tax deduction and seller qualifications, while the operating side focuses on operational dimensions such as purchase contracts, warehousing verification, and logistics vouchers.

[0114] Subsequently, the server, through a direct encrypted channel between the tax authorities and enterprises, sent the verification scope anchor point "CF2024Q1-CL01" to the supply chain management server on the operational side of the large equipment manufacturing group. Upon receiving the anchor point, the operational server immediately matched it with its own regulatory data scope mapping table, locating the corresponding target regulatory data scope "2024Q1 Core Component Procurement Full-Link Data". Then, based on the 92 electronic invoice number anchor points subsequently sent by the regulatory side, it batch-retrieved the corresponding transaction dimension information from this data scope, including purchase contract numbers, warehousing and acceptance records, cross-regional logistics waybills, and supplier qualification documents. Finally, according to the verification rules preset by the regulatory side, it conducted a penetrating verification of the operational dimension data of the 92 transaction links: verifying the consistency between the electronic invoice amount and the purchase contract amount, the warehousing order value, matching the origin of the logistics waybill with the seller's registered location, verifying the correlation between supplier qualifications and transaction content, and finally generating a verification result report to be fed back to the regulatory side, completing the execution of the supply chain penetration verification.

[0115] In this embodiment of the invention, the step of encoding the relevant electronic ticket number anchor sequence of the reference electronic ticket number anchor based on the encoding scale and with the reference electronic ticket number anchor as a reference, to obtain the relevant anchor code data corresponding to the relevant electronic ticket number anchor sequence of the reference electronic ticket number anchor, can be implemented through the following example.

[0116] The reference electronic ticket number anchor point is used as the query vector, and each related electronic ticket number anchor point in the related electronic ticket number anchor point sequence of the reference electronic ticket number anchor point is used as the key vector and value vector, and input into the multi-head attention encoding model.

[0117] The attention weights between the query vector and each of the key vectors are calculated using the multi-head attention encoding model. The attention weights are used to characterize the importance of each of the relevant electronic ticket number anchors relative to the reference electronic ticket number anchors during the encoding process.

[0118] Based on the attention weights, the value vectors are weighted and fused to generate a fused context-aware anchor feature vector.

[0119] Based on the encoding scale, the context-aware anchor feature vector is mapped to bitmap data or hash-encoded data with a scale matching the encoding scale, and used as the relevant anchor encoding data;

[0120] In the mapping process, the information corresponding to the relevant electronic ticket number anchor point with higher attention weight is assigned a higher retention priority or encoding precision in the bitmap data or hash-encoded data.

[0121] In an embodiment of the present invention, for example, the tax supervision server initiates a multi-head attention encoding process for the reference number of electronic invoices and three related anchors in the "procurement of core components of power system" sequence:

[0122] First, the server converts the reference digital ticket number anchor "2024000011223344556" into a 10-dimensional query vector. The vector elements are composed of each digit of the last 10 digits of the digital ticket "1122334456". At the same time, the three related digital ticket number anchors "2024000011223344557, 2024000011223344558, and 2024000011223344559" are converted into 10-dimensional key vectors and value vectors, respectively. The vector elements correspond to the last 10 digits of the digital ticket. Then, the query vector, all key vectors, and value vectors are input into a pre-trained multi-head attention encoding model. In this embodiment, the specific implementation details of the multi-head attention encoding model are as follows: 1. Network structure: A 3-head attention mechanism is adopted, with an input dimension of 10 dimensions (corresponding to the numerical vector of the last 10 digits of the electronic ticket number), a hidden layer dimension of 64, and an output dimension of 10 dimensions; 2. Training method: Using the historical 3-year supply chain electronic ticket anchor sequence as training data, the cross-entropy loss function is adopted, and iterative training is performed for 500 rounds until the loss value is lower than 0.01; 3. Bitmap mapping rule: Each element of the context-aware anchor feature vector is mapped to 51 consecutive bits of a 512-bit bitmap (10-dimensional features correspond to 510 bits, and the remaining 2 bits are used for encoding identification). Elements with feature values ​​greater than 0.5 are marked as "1", otherwise marked as "0", ensuring that the features of high-weight anchors are preferentially retained.

[0123] Next, the model calculates attention weights based on the cosine similarity between vectors and the proportion of transaction amount: the reference anchor point corresponds to the main shaft procurement, and the three relevant anchor points correspond to the procurement of connecting rods, seals, and bearings, respectively. The transaction amounts of the three account for 60%, 25%, and 15% of the total amount of the sequence. The model uses the proportion of transaction amount as a weight correction factor and calculates the attention weights of the three relevant anchor points to be 0.6, 0.25, and 0.15, respectively, which clearly characterize the importance of each relevant anchor point relative to the reference anchor point.

[0124] Subsequently, the server performs weighted fusion of the value vectors based on attention weights: the value vector of the first relevant anchor is multiplied by 0.6, the second by 0.25, and the third by 0.15. The three weighted vectors are then summed according to element position to generate a 10-dimensional context-aware anchor feature vector that integrates the importance of each anchor. This vector retains the baseline attributes of the reference anchor while highlighting the features of high-weight relevant anchors.

[0125] Finally, the server maps the context-aware anchor feature vectors into bitmap data based on the 512-bit binary encoding scale: according to the attention weight, the encoding precision and retention priority are allocated, the link procurement anchor with the highest weight is mapped to the first 256 bits of the bitmap and marked as the first state identifier value "1"; the seal procurement anchor with the second highest weight is mapped to the middle 128 bits and marked as "1"; the bearing procurement anchor with the lowest weight is mapped to the next 64 bits and marked as "1"; the remaining 64 bits are all marked as the second state identifier value "0", and finally the relevant anchor encoding data that conforms to the encoding scale is generated, ensuring that the highly important transaction link information is more fully preserved in the encoding.

[0126] In this embodiment of the invention, the step of calculating the attention weights between the query vector and each of the key vectors using the multi-head attention encoding model can be implemented through the following example.

[0127] Obtain the sequence hierarchy identifier of the relevant electronic ticket number anchor point sequence of the reference electronic ticket number anchor point; determine the corresponding relative position offset based on the hierarchy distance between the reference electronic ticket number anchor point and each of the relevant electronic ticket number anchor points;

[0128] When calculating the attention weight, the relative position bias is introduced for weighted correction, so that the relevant electronic ticket number anchor points with closer hierarchical distances receive higher base attention weight values.

[0129] In an embodiment of the present invention, for example, when the tax supervision server calculates the attention weight of the multi-head attention encoding model, it first reads the hierarchical identifier L2-01 of the "power system core component procurement" sequence from the metadata cache of the dynamic anchor carrying unit. This identifier corresponds to the secondary supply chain branch and is fixed metadata synchronously written during sequence aggregation.

[0130] Subsequently, the server parses the third-level identifier L3-01 (core component of the power system > main shaft) of the reference electronic ticket number anchor point "2024000011223344556", as well as the third-level identifiers of three related anchor points: connecting rod corresponds to L3-02, seal corresponds to L3-03, and bearing corresponds to L3-04. The server calculates that the hierarchical distances of the three points from the reference anchor point are 1, 2, and 3 respectively, and then matches the preset mapping rules to determine the relative position offsets as 0.3, 0.15, and 0.05.

[0131] Finally, the server incorporates this bias weighting correction when calculating attention weights: using the proportion of transaction amount as the base weight (0.6, 0.25, 0.15), multiplied by a correction coefficient of (1 + relative position bias) to obtain corrected attention weights of 0.78, 0.2875, and 0.1575, ensuring that related anchors with closer hierarchical distances receive higher base attention weight values, highlighting the correlation of transaction links at the same level.

[0132] It is worth noting that the multi-head attention encoding model is an adaptive modification of the encoder part in the standard Transformer architecture, and its implementation and training involve the following repeatable steps:

[0133] Model input preparation: The feature vector of the reference anchor point is used as the "query" input, and the feature vector of the relevant anchor point sequence is used as both the "key" and "value" input. All feature vectors must first pass through a linear transformation layer to adapt to the internal dimensionality of the model.

[0134] Attention Calculation Process: The core of the model is the parallel execution of multiple sets of attention calculations. In each set (called a "head"), the system calculates the relevance score between the query vector and each key vector (usually obtained by vector dot product and scaling). These scores are then converted into weights (using the Softmax function), and finally, these weights are used to perform a weighted summation of the corresponding "value" vectors to obtain a composite vector. The output vectors of multiple "heads" are concatenated and then subjected to a linear transformation to form the final "context-aware anchor feature vector".

[0135] Model Training Method: To train the model, a training dataset needs to be constructed. From historical electronic invoice data, based on real supply chain relationships, a large number of correct "reference anchor-related anchor sequence" combinations are sampled as positive samples; simultaneously, unrelated anchors are randomly combined to form negative samples. During training, the model aims to learn a feature representation that makes the context vectors generated by the reference anchor and related anchors in the positive samples as similar as possible, while making them as different as possible from the context vectors of the negative samples. By iteratively optimizing this objective, the model can learn to capture the deep correlation features between supply chain transactions. The trained model parameters are fixed and deployed on the regulatory side node.

[0136] In this embodiment of the invention, the determination of the relevant sequence of the subsequent electronic ticket number anchor point, or the determination of the relevant sequence of the target number electronic ticket number anchor point, can be performed through the following examples.

[0137] A dynamic transaction relationship graph is constructed based on the preset logical association rules between the digital and electronic ticket number anchor points; wherein, the first number of digital and electronic ticket number anchor points are used as nodes, and edges are constructed between the associated nodes according to at least one of the ticket information association, upstream and downstream enterprise association, or transaction time window association.

[0138] The current electronic ticket number anchor point to be determined is taken as the target node, and the relevant electronic ticket number anchor points and their corresponding edge relationships that have been temporarily stored in the dynamic anchor point carrying unit are taken as the current subgraph context and input into the lightweight graph neural network model.

[0139] Through the message passing mechanism of the lightweight graph neural network model, the neighbor node information of the target node in the current subgraph context is aggregated, and the probability of the target node belonging to the current related number anchor sequence, the adjacent related number anchor sequence, or the cross-sequence related number anchor sequence is output.

[0140] The relevant sequence determination result is generated based on the attribution probability.

[0141] In an embodiment of the present invention, for example, the tax supervision server initiates a relevant sequence determination process for the 92 selected electronic invoice number anchors: First, a dynamic transaction relationship graph is constructed based on preset logical association rules. The preset rules cover three dimensions: "similarity of goods category on the invoice ≥ 80%, upstream and downstream enterprises are at the same supply chain level, and transaction time interval ≤ 7 days". The server converts the 92 electronic invoice number anchors one by one into graph nodes, and then traverses the association information between nodes: undirected edges are constructed between nodes whose goods category belongs to "core components of the power system", are supplied by the same supplier outside the province, and have a transaction time interval ≤ 3 days, and finally a dynamic transaction relationship graph containing 92 nodes and 38 edges is formed.

[0142] Subsequently, the server processes the target number of the electronic ticket anchor "2024000011223344561" to be judged. At this time, the dynamic anchor carrying unit temporarily stores 3 nodes of the "core components of the power system" sequence and their corresponding 4 edge relationships. The server sets the target node as the center node, extracts the temporarily stored nodes and edges as the current subgraph context, and inputs them into the pre-trained lightweight graph convolutional neural network model.

[0143] The model initiates a message passing mechanism: using the target node's "vehicle control system core chip" category attribute, out-of-province independent supplier attribute, and transaction time attribute as initial features, it aggregates the neighboring nodes' "power system components" category, same fixed supplier attribute, and dense transaction time attribute through edge relationships, and calculates the feature similarity and association weight between nodes; after two rounds of message passing, the model outputs the belonging probability: the probability that the target node belongs to the current related sequence is 5%, the probability that it belongs to the adjacent related sequence is 90%, and the probability that it belongs to the cross-sequence related sequence is 5%.

[0144] Finally, based on the attribution probability threshold rule (the highest probability ≥ 80% is sufficient to determine the corresponding category), the server directly generates the relevant sequence determination result, clarifying that the target number of electronic ticket number anchor points belongs to the adjacent relevant number of electronic ticket number anchor point sequences of the current target's relevant number of electronic ticket number anchor points sequence.

[0145] It is worth noting that the lightweight graph neural network adopts a graph convolutional network architecture, and its construction and operation process is as follows:

[0146] Transaction relationship graph construction: Multiple electronic ticket number anchors to be judged are considered as nodes in the graph. The node features are the feature vectors of that anchor. Based on preset association rules (e.g., ticket name similarity exceeds a certain percentage, upstream and downstream relationships exist between the transacting parties, and the transaction time falls within a specific window), edges are established between these nodes. The weights of the edges can be quantified and assigned according to the degree to which the association rules are satisfied.

[0147] Model Inference Process: The model operates through a "message passing" mechanism. In each layer of graph convolution, each node aggregates the feature information of its directly connected neighboring nodes (usually using a weighted average based on edge weights), then combines the aggregated information with its original features and transforms it to form the node's new features. After two layers of such aggregation and transformation, each node contains information about its local graph structure. Finally, the final features of the target node are input into a classifier, which outputs the probability that it belongs to the "current sequence," "neighboring sequence," or "cross sequence."

[0148] Model Training Method: Training this model requires labeled graph data samples. Using historical data, multiple small-scale transaction relationship graphs with varying numbers of nodes and edges are simulated and constructed. Each node in the graph is manually or according to business rules labeled with the correct sequence category. The model is then trained under supervised supervision using this labeled graph data. The model parameters are adjusted to ensure that the output category probabilities are as consistent as possible with the true labels. Once trained, the model can be used for online dynamic decision-making.

[0149] In this embodiment of the invention, the step of remapping the anchor point status of the basic anchor points of the second number of transaction links using progressively associated discrete state identifier values ​​to obtain the digital ticket number anchor points of the second number of transaction links can be implemented through the following example.

[0150] The sequence of base anchor points of the second number of transaction links, arranged in order, is input into the state remapping autoencoder.

[0151] The encoder network of the state remapping autoencoder is used to extract and compress features from the basic anchor sequence, and outputs a discrete logical representation of a preset dimension at the bottleneck layer.

[0152] The discrete logic representation is discretized using Gumbel-Softmax to obtain the discrete state identifier value corresponding to each basic anchor point, which is used as the digital ticket number anchor point of the corresponding transaction link.

[0153] The decoder network of the state remapping autoencoder is used to reconstruct features based on the discrete state identifier value sequence, and is trained with the goal of minimizing the sum of reconstruction loss and contrastive learning loss; wherein, the contrastive learning loss is used to constrain the base anchors with similar transaction dimension information to be mapped to the same or adjacent discrete state identifier values.

[0154] In an embodiment of the present invention, for example, the tax supervision server initiates an anchor state remapping process based on a state remapping autoencoder for 1500 sequentially arranged basic anchor point sequences of a large equipment manufacturing group in the first quarter of 2024. In this embodiment, the specific parameters and implementation logic of the state remapping autoencoder are as follows: 1. Encoder network: 3 fully connected layers, input layer dimension 20 (corresponding to the length of the basic anchor point character vector), intermediate layer dimensions 128 and 64 respectively, bottleneck layer dimension 4 (corresponding to 4 discrete state identifier values ​​S1-S4), and the activation function is ReLU; 2. Decoder network: symmetrical 3 fully connected layers, bottleneck layer dimension 4, intermediate layers dimensions 64 and 128, output layer dimension 20, activation function is Sigmoid; 3. Gumbel-Softmax parameters: temperature coefficient is set to 1.0, batch size is 32 to ensure the stability of discretization results; 4. Loss function: reconstruction loss (mean squared error) is the main component, and contrastive learning loss (InfoNCE loss, weight 0.3) is the auxiliary component. The basic anchor points in the same transaction amount range and the same sales region are regarded as positive sample pairs, and similar anchor points are constrained to map to adjacent discrete states.

[0155] First, the server imports the pre-trained state remapping autoencoder in batches, sorting the basic anchor point sequence by "transaction amount > electronic invoice generation time > seller region". The encoder network of this autoencoder adopts a 3-layer fully connected structure. The input layer dimension matches the 20-bit character vector length of the basic anchor points, and the bottleneck layer dimension is set to 4, corresponding to the four preset discrete state label values ​​S1 to S4.

[0156] The encoder network performs feature extraction and compression on the input basic anchor sequence: the character vector of each basic anchor is converted into a numerical vector, key features such as transaction amount, generation time, and seller region are extracted through convolutional layers, and then compressed to a 4-dimensional bottleneck layer through a fully connected layer, outputting a discrete logical representation containing state classification logic, where the larger the value of the representation, the higher the transaction amount and the higher the priority of the anchor.

[0157] Subsequently, the server applies Gumbel-Softmax discretization to the discrete logic representation output by the bottleneck layer: noise conforming to the Gumbel distribution is added to each representation, and the continuous representation is converted into a probability distribution through the Softmax function. The category corresponding to the highest probability is taken as the discrete state identifier value of the basic anchor point. The first 212 basic anchor points with transaction amounts ≥ 500,000 yuan are mapped to S1, the middle 426 transactions with amounts between 200,000 and 500,000 yuan are mapped to S2, the next 612 transactions with amounts between 100,000 and 200,000 yuan are mapped to S3, and the last 50 transactions with amounts < 100,000 yuan are mapped to S4, generating the digital ticket number anchor points for the corresponding transaction links.

[0158] Finally, the decoder network of the autoencoder performs feature reconstruction based on the discrete state identifier value sequence: the 4-dimensional discrete identifier values ​​are restored to a reconstructed vector with the same dimension as the base anchor points, and the mean square error between the reconstructed vector and the original base anchor points is calculated as the reconstruction loss. Simultaneously, a contrastive learning loss is introduced. For base anchor points with similar transaction dimensional information (such as the same amount range or the same seller's region), the distance difference between their discrete identifier values ​​is calculated and backpropagated for optimization, constraining similar anchor points to be mapped to the same or adjacent discrete state identifier values. The server iteratively trains the autoencoder with the goal of minimizing the sum of the reconstruction loss and the contrastive learning loss until the loss value drops below a preset threshold, ensuring the mapping accuracy and relevance of the discrete state identifier values.

[0159] In this embodiment of the invention, for example, the training of the state remapping autoencoder simultaneously pursues two objectives: first, to accurately reconstruct the input features (reconstruction loss), and second, to make the features of similar transactions close together in the encoding space (contrastive learning loss).

[0160] The training process is briefly described as follows: The system prepares a large number of basic anchor point features from transaction chains as input. In each training iteration, the encoder compresses the input features into a low-dimensional code, and the decoder attempts to reconstruct the input from this code. The reconstruction loss measures the difference between the reconstructed result and the original input. Simultaneously, the system identifies anchor point pairs with similar transaction attributes (such as amount range or seller region) in the batch data as "positive sample pairs." By contrastively learning the loss function, the system encourages the low-dimensional codes of these positive sample pairs to be close to each other, while keeping them away from the codes of other random anchor points. The entire model training involves iteratively adjusting the network parameters to continuously reduce the sum of the two losses until the model stabilizes. After training, the low-dimensional code output by the encoder is mapped to discrete state identifier values ​​according to set rules (such as taking the maximum value index).

[0161] This invention provides a computer device 100, which includes a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device 100 executes the aforementioned supply chain full-level penetration supervision method based on electronic invoices. Figure 2 As shown, Figure 2 This is a structural block diagram of a computer device 100 provided in an embodiment of the present invention. The computer device 100 includes a memory 111, a processor 112, and a communication unit 113. To enable data transmission or interaction, the memory 111, processor 112, and communication unit 113 are electrically connected to each other directly or indirectly. For example, these components can be electrically connected to each other through one or more communication buses or signal lines.

[0162] For illustrative purposes, the foregoing description has been made with reference to specific embodiments. However, the foregoing illustrative discussions are not intended to be exhaustive or to limit the present disclosure to the precise forms disclosed. Numerous modifications and variations are possible in accordance with the foregoing teachings. These embodiments were chosen and described in order to best illustrate the principles of the present disclosure and its practical application, thereby enabling those skilled in the art to best utilize the disclosure and to employ various embodiments with different modifications to suit a particular intended application.

Claims

1. A supply chain full-level penetration supervision method based on electronic invoices, characterized in that, include: Identify the supply chain penetration verification items to be regulated; the supply chain penetration verification items need to be completed collaboratively by regulatory and operational nodes, and the regulatory and operational nodes are configured with the target full-chain regulatory correlation data required to execute the supply chain penetration verification items; the target full-chain regulatory correlation data of the regulatory and operational nodes includes different transaction dimension information of the same transaction link, and the same transaction link has a completely consistent electronic ticket number anchor point in the regulatory and operational nodes; Based on the rules and constraints of the supply chain penetration verification matters, select the first number of digital invoice number anchor points from the target full-chain regulatory related data of the regulatory side node; Based on the coding scale corresponding to the target full-chain regulatory association data of the regulatory side node, the digital ticket number anchor points of the first number of transaction links are encoded to obtain digital ticket number anchor point coding data. The digital ticket number anchor encoding data is sent to the operation side node, so that the operation side node decodes the digital ticket number anchor encoding data to obtain the digital ticket number anchors of the first number of transaction links, and performs the supply chain penetration verification based on the digital ticket number anchors of the first number of transaction links. The encoding scale corresponding to the target full-chain regulatory association data based on the regulatory side node is used to encode the digital invoice number anchor points of the first number of transaction links to obtain digital invoice number anchor point encoding data, including: From the first number of electronic ticket number anchor points, determine a reference number of electronic ticket number anchor points and a related number of electronic ticket number anchor point sequence of the reference number of electronic ticket number anchor points, wherein the related number of electronic ticket number anchor point sequence of the reference number of electronic ticket number anchor points includes the related number of electronic ticket number anchor points of the reference number of electronic ticket number anchor points. Based on the encoding scale, determine the reference anchor point encoding data corresponding to the reference number electronic ticket number anchor point; Based on the encoding scale, and taking the reference electronic ticket number anchor point as a reference, the relevant electronic ticket number anchor point sequence of the reference electronic ticket number anchor point is encoded to obtain the relevant anchor point encoded data corresponding to the relevant electronic ticket number anchor point sequence of the reference electronic ticket number anchor point. The digital ticket number anchor point encoding data includes reference anchor point encoding data corresponding to the reference digital ticket number anchor point, and related anchor point encoding data corresponding to the related digital ticket number anchor point sequence of the reference digital ticket number anchor point.

2. The method according to claim 1, characterized in that, In the process of determining the reference electronic ticket number anchor point and the related electronic ticket number anchor point sequence of the reference electronic ticket number anchor point, a dynamic anchor point carrying unit is configured. The dynamic anchor point carrying unit is used to temporarily store the related electronic ticket number anchor points in the related electronic ticket number anchor point sequence of the reference electronic ticket number anchor point. The step of determining the reference electronic ticket number anchor point and the related electronic ticket number anchor point sequence of the reference electronic ticket number anchor point from a first number of electronic ticket number anchor points includes: The first number of electronic ticket number anchor points are arranged in order, and the first number of electronic ticket number anchor points after the order arrangement are processed one by one; among the first number of electronic ticket number anchor points after the order arrangement, the electronic ticket number anchor point currently being processed is the target number of electronic ticket number anchor points; If the dynamic anchor point carrying unit does not carry a valid anchor point, then the target number electronic ticket number anchor point is marked as the reference number electronic ticket number anchor point, and the subsequent number electronic ticket number anchor points are re-sequenced based on the target number electronic ticket number anchor point to obtain the re-sequenced result of the subsequent number electronic ticket number anchor point; the subsequent number electronic ticket number anchor point is the next number electronic ticket number anchor point of the target number electronic ticket number anchor point; If the relevant sequence determination result of the subsequent number electronic ticket number anchor point indicates that the subsequent number electronic ticket number anchor point belongs to the relevant number electronic ticket number anchor point sequence of the target number electronic ticket number anchor point, then the subsequent number electronic ticket number anchor point is added to the dynamic anchor point carrying unit. If the relevant sequence determination result of the subsequent number electronic ticket number anchor point indicates that the subsequent number electronic ticket number anchor point does not belong to the relevant number electronic ticket number anchor point sequence of the target number electronic ticket number anchor point, then the subsequent number electronic ticket number anchor point is determined as the updated reference number electronic ticket number anchor point.

3. The method according to claim 2, characterized in that, The step of determining the reference number of electronic ticket number anchor points and the related number of electronic ticket number anchor points sequence of the reference number of electronic ticket number anchor points from the first number of electronic ticket number anchor points further includes: If the dynamic anchor point carrying unit carries a valid anchor point, then the relevant sequence determination is performed on the target number of electronic ticket number anchor points to obtain the relevant sequence determination result of the target number of electronic ticket number anchor points; when the currently processed number of electronic ticket number anchor points is the target number of electronic ticket number anchor points, the dynamic anchor point carrying unit temporarily stores the relevant number of electronic ticket number anchor points of the target relevant number of electronic ticket number anchor point sequence; If the determination result of the relevant sequence of the target number electronic ticket number anchor point indicates that the target number electronic ticket number anchor point belongs to the target relevant number electronic ticket number anchor point sequence, then the target number electronic ticket number anchor point is added to the dynamic anchor point carrying unit; If the determination result of the relevant sequence of the target number electronic ticket number anchor point indicates that the target number electronic ticket number anchor point belongs to the adjacent relevant number electronic ticket number anchor point sequence of the target relevant number electronic ticket number anchor point sequence, then it is determined that the determination of the relevant number electronic ticket number anchor point of the target relevant number electronic ticket number anchor point sequence is completed, the dynamic anchor point carrying unit is reset, and the target number electronic ticket number anchor point is added to the dynamic anchor point carrying unit as a relevant number electronic ticket number anchor point in the updated relevant number electronic ticket number anchor point sequence; If the determination result of the relevant sequence of the target number electronic ticket number anchor point indicates that the target number electronic ticket number anchor point belongs to the cross-sequence relevant number electronic ticket number anchor point sequence of the target relevant number electronic ticket number anchor point sequence, then the determination of the relevant number electronic ticket number anchor point of the target relevant number electronic ticket number anchor point sequence is completed, the dynamic anchor point carrying unit is reset, and the target number electronic ticket number anchor point is used as the updated reference number electronic ticket number anchor point.

4. The method according to claim 3, characterized in that, The step of determining the correlation sequence of the target number of electronic ticket number anchors to obtain the correlation sequence determination result of the target number of electronic ticket number anchors includes: Obtain the sequence level identifier of the target-related electronic ticket number anchor sequence; The target number of electronic ticket number anchors, the current reference number of electronic ticket number anchors, and the coding scale are subjected to rule parsing processing to determine the sequence level identifier to which the target number of electronic ticket number anchors belong; If the sequence level identifier of the target related number electronic ticket number anchor point sequence is the same as the sequence level identifier of the target number electronic ticket number anchor point, then a related sequence determination result of the target number electronic ticket number anchor point is generated. The related sequence determination result of the target number electronic ticket number anchor point indicates that the target number electronic ticket number anchor point belongs to the target related number electronic ticket number anchor point sequence. If the hierarchical spacing between the sequence level identifier of the target related number electronic ticket number anchor point sequence and the sequence level identifier to which the target number electronic ticket number anchor point belongs is equal to the preset hierarchical discrimination boundary, then the related sequence determination result of the target number electronic ticket number anchor point is generated. The related sequence determination result of the target number electronic ticket number anchor point indicates that the target number electronic ticket number anchor point belongs to the adjacent related number electronic ticket number anchor point sequence of the target related number electronic ticket number anchor point sequence. If the hierarchical distance between the sequence level identifier of the target related number electronic ticket number anchor sequence and the sequence level identifier to which the target number electronic ticket number anchor belongs is greater than a preset hierarchical discrimination boundary, then a related sequence determination result for the target number electronic ticket number anchor is generated. The related sequence determination result for the target number electronic ticket number anchor indicates that the target number electronic ticket number anchor belongs to a cross-sequence related number electronic ticket number anchor sequence of the target related number electronic ticket number anchor sequence.

5. The method according to claim 1, characterized in that, The step involves encoding the relevant electronic ticket number anchor sequence of the reference electronic ticket number anchor based on the encoding scale and with the reference electronic ticket number anchor as a reference, to obtain the relevant anchor code data corresponding to the relevant electronic ticket number anchor sequence of the reference electronic ticket number anchor, including: Obtain the encoded bitmap with the scale of the encoding scale; Based on the reference electronic ticket number anchor point, the sequence level identifier of the related electronic ticket number anchor point sequence of the reference electronic ticket number anchor point, and each related electronic ticket number anchor point in the related electronic ticket number anchor point sequence of the reference electronic ticket number anchor point, determine the regular mapping position of the encoding bitmap corresponding to each related electronic ticket number anchor point; In the encoded bitmap with the scale of the encoding scale, the regular mapping site corresponding to each relevant digital ticket number anchor point is marked as a first state identifier value, and the remaining sites other than the regular mapping site corresponding to each relevant digital ticket number anchor point are marked as a second state identifier value, so as to obtain the basic anchor point encoding data of the relevant digital ticket number anchor point sequence of the reference digital ticket number anchor point; Add coded anchor bits to the basic anchor bit encoded data of the relevant electronic ticket number anchor sequence of the reference electronic ticket number anchor point to obtain the relevant anchor bit encoded data corresponding to the relevant electronic ticket number anchor sequence of the reference electronic ticket number anchor point; the coded anchor bit is marked as an encoding status identifier value, and the encoding status identifier value indicates that the relevant anchor bit encoded data is encoded data.

6. The method according to claim 1, characterized in that, The target full-chain regulatory association data of the regulatory side node includes digital invoice number anchors for a second number of transaction links. These second number of digital invoice number anchors are obtained by remapping the anchor states of the basic anchors for the second number of transaction links, and the second number is not less than the first number. The method further includes: Obtain the base anchor points of the second number of transaction links, which are generated based on the anchor point generation elements in the anchor point generation rule set; Based on the rule priority relationship in the anchor point generation rule set, the basic anchor points of the second number of transaction links are arranged in order to obtain the second number of transaction links in order. Based on the anchor point mapping order of the second number of transaction links' basic anchor points arranged in sequence, the anchor point states of the second number of transaction links are remapped using progressively associated discrete state identifier values ​​to obtain the digital ticket number anchor points of the second number of transaction links.

7. The method according to claim 1, characterized in that, The step involves encoding the relevant electronic ticket number anchor sequence of the reference electronic ticket number anchor based on the encoding scale and with the reference electronic ticket number anchor as a reference, to obtain the relevant anchor code data corresponding to the relevant electronic ticket number anchor sequence of the reference electronic ticket number anchor, including: The reference electronic ticket number anchor point is used as the query vector, and each related electronic ticket number anchor point in the related electronic ticket number anchor point sequence of the reference electronic ticket number anchor point is used as the key vector and value vector, and input into the multi-head attention encoding model. The attention weights between the query vector and each of the key vectors are calculated using the multi-head attention encoding model. The attention weights are used to characterize the importance of each of the relevant electronic ticket number anchors relative to the reference electronic ticket number anchors during the encoding process. Based on the attention weights, the value vectors are weighted and fused to generate a fused context-aware anchor feature vector. Based on the encoding scale, the context-aware anchor feature vector is mapped to bitmap data or hash-encoded data with a scale matching the encoding scale, and used as the relevant anchor encoding data; In the mapping process, the information corresponding to the relevant electronic ticket number anchor point with higher attention weight is assigned a higher retention priority or encoding precision in the bitmap data or hash-encoded data.

8. The method according to claim 2 or 3, characterized in that, The step of determining the relevant sequence for the subsequent electronic ticket number anchor points, or the step of determining the relevant sequence for the target number of electronic ticket number anchor points, includes: A dynamic transaction relationship graph is constructed based on the preset logical association rules between the digital and electronic ticket number anchor points; wherein, the first number of digital and electronic ticket number anchor points are used as nodes, and edges are constructed between the associated nodes according to at least one of the ticket information association, upstream and downstream enterprise association, or transaction time window association. The current electronic ticket number anchor point to be determined is taken as the target node, and the relevant electronic ticket number anchor points and their corresponding edge relationships that have been temporarily stored in the dynamic anchor point carrying unit are taken as the current subgraph context and input into the lightweight graph neural network model. Through the message passing mechanism of the lightweight graph neural network model, the neighbor node information of the target node in the current subgraph context is aggregated, and the probability of the target node belonging to the current related number anchor sequence, the adjacent related number anchor sequence, or the cross-sequence related number anchor sequence is output. The relevant sequence determination result is generated based on the attribution probability.

9. A server system, characterized in that, Includes a server, the server being used to perform the method according to any one of claims 1-8.

Images