A steel industry supply chain financial risk management method based on big data

Abstract

The application discloses a steel industry supply chain financial risk management method based on big data, which comprises the following steps: generating a steel supply chain financial standardization data set; performing event traceability hash chain reconstruction to generate a steel supply chain financial business link data set; mapping transaction flow, goods flow and fund flow under the same financing business into a multi-flow super-edge to construct a steel supply chain financial business correlation graph; generating a steel supply chain financial risk feature set; performing risk potential gradient tracing to generate a steel supply chain financial risk transmission path set; performing comprehensive evaluation by using an improved EGATCL model to generate a steel supply chain financial risk evaluation result; performing risk strategy sand table deduction to generate a steel supply chain financial risk management result, and the completeness of steel industry supply chain financial business link restoration, the accuracy of cross-flow risk identification and the continuity of risk transmission analysis are improved.

Description

Technical Field

[0001] This invention relates to the field of supply chain finance management, and in particular to a big data-based method for supply chain finance risk management in the steel industry. Background Technology

[0002] Supply chain finance in the steel industry typically revolves around steel procurement, order fulfillment, warehousing supervision, goods circulation, bill settlement, fund collection, and credit management. This involves a long business chain, numerous participants, large transaction amounts, and strong temporal connections between transaction flows, goods flows, and fund flows. In practice, financing applicants, core enterprises, warehousing providers, logistics providers, funding providers, and regulators each possess different business data. The original data sources are scattered, data formats are inconsistent, and business field definitions are inconsistent. This forces financial institutions to simultaneously process contract data, order data, warehousing data, logistics data, invoice data, fund collection data, and credit data when conducting risk management. Existing supply chain finance risk management methods largely rely on business personnel reviewing individual documents or using rule models to statically assess credit limits, overdue records, enterprise qualifications, and transaction amounts. While these methods can identify some risks, they fail to reflect the true connections between multiple entities, documents, and flows in steel industry financing transactions.

[0003] Existing technologies for processing supply chain finance data in the steel industry typically treat transaction contracts, title deeds, logistics documents, and payment records as independent data objects for verification. This lacks a reconstruction of the temporal sequence between business entities, transactions, goods movement, and fund receipts within the same financing transaction. Due to the absence of an event tracing mechanism, existing methods struggle to determine whether a transaction truly follows a previous business event, and also fail to identify gaps, misalignments, or inversions between contract signing, goods release, transit, warehousing, and payment receipt within the same financing transaction. This can easily lead to problems with the business chain not being closed. Particularly in steel trading scenarios, there are multiple changes between goods batches, warehouse receipt status, transportation nodes, and payment nodes. Relying solely on single-point data verification makes it difficult to accurately reconstruct the complete financing chain, thus affecting the accuracy of risk identification. Summary of the Invention

[0004] One objective of this invention is to propose a big data-based supply chain finance risk management method for the steel industry. This invention fully utilizes big data processing, event tracing hash chain reconstruction, multi-flow hyperedge graph modeling, risk potential gradient tracing, and an improved EGATCL model, and has the advantages of high risk identification accuracy, strong business adaptability, timely risk handling, and reliable management results.

[0005] A method for supply chain finance risk management in the steel industry based on big data, according to an embodiment of the present invention, includes the following steps: Collect and preprocess the raw datasets from the steel industry supply chain finance business to generate a standardized dataset for steel supply chain finance. Reconstruct the event tracing hash chain of the standardized dataset for steel supply chain finance to determine the temporal sequence relationship between business entities, transactions, goods flow and fund recovery, and generate a steel supply chain finance business link dataset. Using the temporal sequence relationship in the data set of steel supply chain finance business links as the main axis, the transaction flow, goods flow and capital flow under the same financing business are mapped as multi-flow hyperedges to construct a steel supply chain finance business association graph; The multi-flow hyperedge in the steel supply chain finance business association graph is determined as the verification skeleton, the cross-flow closed loop breakpoints corresponding to the same financing business are identified, the multi-dimensional risk characteristics of steel supply chain finance are extracted, and the steel supply chain finance risk characteristic set is generated. By identifying risk triggers through the set of risk characteristics of steel supply chain finance, and tracing the risk potential gradient along the graph edges of the steel supply chain finance business association graph, a set of risk transmission paths for steel supply chain finance is generated. The set of risk transmission paths in the steel supply chain finance is input into the improved EGATCL model for comprehensive evaluation, generating the risk assessment results of the steel supply chain finance. Based on the risk assessment results of steel supply chain finance, a risk strategy simulation was conducted to simulate strategies for reducing financing limits, verifying ownership of goods, locking in payments, and verifying transactions, thereby generating risk management results for steel supply chain finance.

[0006] Optionally, the original dataset includes order data, contract data, invoice data, logistics data, warehousing data, accounts receivable data, cash flow data, enterprise credit data, and steel price data. The preprocessing includes field normalization, time alignment, and identifier encoding.

[0007] Optionally, the generation of the steel supply chain finance business link dataset specifically includes: Based on the standardized dataset of steel supply chain finance, events are grouped according to the business entity, transaction, goods flow and fund recovery corresponding to the same financing business to generate a financing business event sequence. Set up time slots according to the business sequence of business entity, transaction, goods flow and fund recovery, and write the financing business event sequence into the corresponding time slot to generate the financing business time slot sequence. Perform joint hash anchoring on the time-series slot sequence of financing business to generate an event tracing hash chain; Perform chain-based verification on the event tracing hash chain to determine the corresponding temporal sequence relationship; Based on the temporal sequence relationship, the business entities, transactions, goods transfer and fund recovery corresponding to the same financing business are encapsulated into a chain to generate a steel supply chain finance business chain dataset.

[0008] Optionally, the construction of the steel supply chain finance business association map specifically includes: Taking the time sequence relationship of the steel supply chain finance business link data as the main axis, the business entities, transaction items, goods flow and fund recovery are divided into link units to generate a set of financing business link units. Based on the set of financing business link units, transactions are grouped into transaction flows, goods transfers are grouped into goods flows, and fund collections are grouped into fund flows, generating a multi-flow data set for financing business. Map the transaction flow, goods flow and capital flow in the multi-flow data set of financing business to multi-flow hyperedges to generate a multi-flow hyperedge set of financing business. Based on the multi-flow super-edge set of financing business, a correlation graph of steel supply chain finance business is constructed according to the direction of connection between transaction flow, goods flow and capital flow under the same financing business.

[0009] Optionally, the generation of the steel supply chain financial risk characteristic set specifically includes: Based on the business relationship graph of steel supply chain finance, the corresponding multi-flow hyperedges are read according to the same financing business to determine the verification skeleton; Based on the verification skeleton, cross-flow closed-loop topology stitching is performed on the transaction flow, goods flow and capital flow under the same financing business to generate cross-flow closed-loop topology links; Perform residual inversion processing on the cross-flow closed-loop topology link to determine the cross-flow closed-loop breakpoint; Risk vectorization is performed on the cross-flow closed-loop breakpoints to generate multi-dimensional risk characteristics of steel supply chain finance, and these characteristics are aggregated to form a set of risk characteristics of steel supply chain finance.

[0010] Optionally, the generation of the set of financial risk transmission paths in the steel supply chain specifically includes: Based on the risk characteristics set of steel supply chain finance, the transaction flow, goods flow or capital flow where the cross-flow closed loop breakpoints corresponding to the multi-dimensional risk characteristics of steel supply chain finance are located are identified as risk triggering sources. Based on the risk trigger source, in the steel supply chain finance business association graph, starting from the risk trigger source, the same source edge domain is pruned along the graph edge bearing direction corresponding to the same financing business to generate a risk tracing graph edge set. Based on the risk tracing graph edge set, determine the risk potential gradient corresponding to each graph edge, and generate the graph edge risk potential gradient set. Based on the graph edge risk potential gradient set, risk potential gradient tracing is performed along the graph edges in the risk tracing graph edge set to generate a set of risk propagation candidate paths. Perform path potential energy continuity compression on the candidate path set of risk transmission to generate a set of risk transmission paths for the steel supply chain finance.

[0011] Optionally, the generation of the steel supply chain financial risk assessment results specifically includes: The improved EGATCL model inputs the set of risk transmission paths in steel supply chain finance. This improved model includes a path potential energy spectrum folding module, a breakpoint source-flow coupling inversion module, a same-source-different-path comparison and calibration module, and a risk magnitude tomographic assessment module. The improvements are as follows: the traditional EGATCL model enhances the representation of node and edge relationships in the graph structure and distinguishes representation differences between different graph samples through graph comparison learning. The improved EGATCL model takes the set of risk transmission paths in steel supply chain finance as input, introduces the risk potential energy gradient into the path-level expression of the graph edges, forms the set of potential energy spectra of steel supply chain finance paths through potential energy ridge rearrangement and folding, determines the set of breakpoint source-flow coupling relationships through coupled phase rollback, and introduces a risk potential energy tomographic projection mechanism to comprehensively evaluate the main risk transmission path, accompanying risk transmission path, risk potential energy concentration, risk trigger source diffusion range, and cross-flow closed-loop breakpoint aggregation location in the same financing business, generating a steel supply chain finance risk assessment result. In the path potential energy spectrum folding module, the risk transmission path in the set of steel supply chain finance risk transmission paths is used as the processing unit. The risk potential energy gradient corresponding to the graph edge is rearranged and folded according to the graph edge acceptance order to generate the set of steel supply chain finance path potential energy spectrum. In the breakpoint source-flow coupling inversion module, based on the potential energy spectrum set of the steel supply chain finance path, the transaction flow, cargo flow and capital flow associated with the cross-flow closed loop breakpoint are coupled and phase-wounded to determine the source-flow coupling relationship between the risk triggering source, the cross-flow closed loop breakpoint, the transaction flow, the cargo flow and the capital flow, and generate a breakpoint source-flow coupling relationship set. In the same source but different path comparison and calibration module, based on the set of breakpoint source-flow coupling relationships, the same source but different path comparison and calibration is performed on different risk transmission paths according to the same financing business, and the same source but different path comparison and calibration results are generated. In the risk level analysis and assessment module, a risk potential energy layering projection mechanism is introduced. Taking the same financing business as the projection object, the risk transmission path in the same source but different path comparison calibration results is projected onto the corresponding business sections of the transaction flow, goods flow, and capital flow according to the graph edge connection order. The risk potential energy gradient, risk triggering source, and cross-flow closed loop breakpoint in the risk transmission path are mapped by cross-sectional potential energy to generate risk layering projection results. The degree of risk potential energy concentration, risk triggering source diffusion range, and cross-flow closed loop breakpoint clustering position of the same financing business in the transaction flow, goods flow, and capital flow are comprehensively evaluated to generate the steel supply chain finance risk assessment results.

[0012] Optionally, the generation of the steel supply chain financial risk management results specifically includes: Read the risk level, risk concentration location and risk transmission intensity corresponding to the risk assessment results of steel supply chain finance, and generate a set of risk strategy trigger parameters; Based on the set of risk strategy trigger parameters, risk strategy sandbox simulation is conducted on financing business strategies such as limit reduction, title verification, repayment lock-in and transaction verification, and risk strategy sandbox simulation results are generated. Based on the results of the risk strategy simulation, the target risk management strategy combination is determined, and the steel supply chain financial risk management results are generated.

[0013] The beneficial effects of this invention are: This invention preprocesses the raw dataset from steel industry supply chain finance to generate a standardized dataset for steel supply chain finance. Further, it reconstructs the event tracing hash chain, unifying the temporal relationships between business entities, transactions, goods flow, and fund collection. This avoids the business chain disruptions caused by the fragmented processing of contract, warehousing, logistics, and fund collection data in existing technologies. By mapping transaction flows, goods flows, and fund flows under the same financing business to multi-flow hyperedges and constructing a steel supply chain finance business association graph, this invention can express the relationships between multiple entities, multiple documents, and multiple business flows within a unified graph structure. This improves the completeness of steel supply chain finance business chain reconstruction and the accuracy of cross-flow data verification.

[0014] This invention uses multi-flow superedges as the verification framework to identify cross-flow closed-loop breakpoints corresponding to the same financing business, extracts multi-dimensional risk characteristics of steel supply chain finance, and generates a set of risk transmission paths for steel supply chain finance through risk potential gradient tracing. This overcomes the problems of existing technologies that only identify isolated risk points and are difficult to analyze the direction and intensity of risk transmission. By inputting the set of risk transmission paths for steel supply chain finance into an improved EGATCL model for comprehensive evaluation, and conducting risk strategy simulation based on the risk assessment results, including credit limit reduction, title verification, payment lock-in, and transaction verification strategies, this invention enables effective integration of risk identification results and risk management strategies, improving the reliability of risk assessment results, the pertinence of risk disposal strategies, and the overall efficiency of supply chain finance risk management in the steel industry. Attached Figure Description

[0015] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is an overall flowchart of a big data-based supply chain finance risk management method for the steel industry proposed in this invention. Figure 2 This is a schematic diagram of the structure of the improved EGATCL model, which is a big data-based method for supply chain finance risk management in the steel industry proposed in this invention. Detailed Implementation

[0016] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0017] refer to Figures 1-2 A big data-based approach to supply chain finance risk management in the steel industry includes the following steps: Collect and preprocess the raw datasets from the steel industry supply chain finance business to generate a standardized dataset for steel supply chain finance. Reconstruct the event tracing hash chain of the standardized dataset for steel supply chain finance to determine the temporal sequence relationship between business entities, transactions, goods flow and fund recovery, and generate a steel supply chain finance business link dataset. Using the temporal sequence relationship in the data set of steel supply chain finance business links as the main axis, the transaction flow, goods flow and capital flow under the same financing business are mapped as multi-flow hyperedges to construct a steel supply chain finance business association graph; The multi-flow hyperedge in the steel supply chain finance business association graph is determined as the verification skeleton, the cross-flow closed loop breakpoints corresponding to the same financing business are identified, the multi-dimensional risk characteristics of steel supply chain finance are extracted, and the steel supply chain finance risk characteristic set is generated. By identifying risk triggers through the set of risk characteristics of steel supply chain finance, and tracing the risk potential gradient along the graph edges of the steel supply chain finance business association graph, a set of risk transmission paths for steel supply chain finance is generated. The set of risk transmission paths in the steel supply chain finance is input into the improved EGATCL model for comprehensive evaluation, generating the risk assessment results of the steel supply chain finance. Based on the risk assessment results of steel supply chain finance, a risk strategy simulation was conducted to simulate strategies for reducing financing limits, verifying ownership of goods, locking in payments, and verifying transactions, thereby generating risk management results for steel supply chain finance.

[0018] In this embodiment, the original dataset includes order data, contract data, invoice data, logistics data, warehousing data, accounts receivable data, cash flow data, enterprise credit data, and steel price data. Preprocessing includes field normalization, time alignment, and identifier encoding.

[0019] In this embodiment, the generation of the steel supply chain finance business link dataset specifically includes: Based on the standardized dataset of steel supply chain finance, events are grouped according to the business entity, transaction, goods flow and fund recovery corresponding to the same financing business to generate a financing business event sequence. The specific process of generating the financing business event sequence is as follows: Based on the standardized dataset of steel supply chain finance, data records are collected according to the same financing business. The business entities, transactions, goods transfers and fund returns belonging to the same financing business are treated as the same group of business events. The events are arranged in order according to the occurrence time of each business event. Data records that are missing business entities, transactions, goods transfers or fund returns are marked as abnormal, thus generating the financing business event sequence. Set up time slots according to the business sequence of business entity, transaction, goods flow and fund recovery, and write the financing business event sequence into the corresponding time slots to generate the financing business time slot sequence. The specific process of generating the financing business time-series slot sequence is as follows: Four types of time-series slots are set up for the same financing business according to the business order of business entity, transaction, goods flow, and fund recovery. Business entity records in the financing business event sequence are written into the time-series slot corresponding to the business entity, transaction records into the time-series slot corresponding to the transaction, goods flow records into the time-series slot corresponding to the goods flow, and fund recovery records into the time-series slot corresponding to the fund recovery. When the same financing business contains one business entity identifier, one transaction identifier, three goods flow records, and two fund recovery records, the records in the same time-series slot are arranged sequentially according to the minute-level business occurrence time of each record. This allows the business entity, transaction, goods flow, and fund recovery to form a slotted event arrangement relationship with a fixed business order and internal time order under the same financing business, generating the financing business time-series slot sequence. Perform joint hash anchoring on the time-series slot sequence of financing business to generate an event tracing hash chain; The process of generating the event tracing hash chain is as follows: Based on the time-series slot sequence of financing business, the business records in each time-series slot are summarized and compressed according to the slot order of business entity, transaction, goods flow, and fund recovery to generate the corresponding slot summary; the slot summary of the current time-series slot, the hash anchor of the previous time-series slot, the financing business identifier, and the business occurrence time are used as joint anchoring elements for hash calculation to generate the hash anchor of the current time-series slot; the hash calculation involves concatenating the slot summary of the current time-series slot, the hash anchor of the previous time-series slot, the financing business identifier, and the business occurrence time according to a fixed field order to form a string to be calculated, and performing unified character encoding, group filling, and compression iterative operations on the string to be calculated to output a hash anchor of fixed length; the generation of hash anchors for the corresponding time-series slots of business entity, transaction, goods flow, and fund recovery is completed in sequence, so that the hash anchor of the next time-series slot inherits the hash anchor of the previous time-series slot, generating the event tracing hash chain; Perform chain-based verification on the event tracing hash chain to determine the corresponding temporal sequence relationship; The process of determining the temporal sequence relationship is as follows: Based on the event tracing hash chain, according to the slot order of business entity, transaction, goods flow and fund return, the hash anchor connection relationship between adjacent temporal slots, the consistency of financing business identifiers and the chronological order of business occurrence time are verified in turn; when the hash anchor of the previous temporal slot is continuously connected by the hash anchor of the next temporal slot, and the adjacent temporal slots belong to the same financing business, the chronological connection relationship between business entity and transaction, transaction and goods flow, and goods flow and fund return is determined as the temporal sequence relationship; Based on the temporal sequence relationship, the business entities, transactions, goods flow and fund recovery corresponding to the same financing business are encapsulated in the chain to generate a steel supply chain finance business chain dataset. The specific process of generating the steel supply chain finance business link dataset is as follows: Based on the temporal sequence relationship, taking the same financing business as the encapsulation object, the business entity, transaction, goods flow and fund return corresponding to the financing business are sequentially bound according to the temporal sequence relationship. The financing business identifier, business occurrence time, event traceability hash chain and corresponding temporal sequence relationship are written into the same data record. This makes the business entity, transaction, goods flow and fund return form traceable, verifiable and transferable link data under the same financing business, thus generating the steel supply chain finance business link dataset.

[0020] In this embodiment, the construction of the steel supply chain finance business relationship map specifically includes: Taking the time sequence relationship of the steel supply chain finance business link data as the main axis, the business entities, transaction items, goods flow and fund recovery are divided into link units to generate a set of financing business link units. The specific process of generating the financing business link unit set is as follows: taking the temporal succession relationship in the steel supply chain finance business link dataset as the main axis, and using the same financing business as the dividing boundary, the business entity, transaction, goods flow, and fund return are read from the steel supply chain finance business link dataset. Based on the sequential connection direction determined by the temporal succession relationship, the business entities, transaction, goods flow, and fund return that are mutually successful under the same financing business are divided into a financing business link unit. When there are multiple transaction, multiple goods flow, or multiple fund returns in the same financing business, corresponding financing business link units are formed according to the temporal succession relationship. All financing business link units are then summarized to generate a financing business link unit set. Based on the set of financing business link units, transactions are grouped into transaction flows, goods transfers are grouped into goods flows, and fund collections are grouped into fund flows, generating a multi-flow data set for financing business. The specific process of generating a multi-stream data set for financing business is as follows: Based on the set of financing business link units, taking the same financing business and its business entity as the merging boundary, the corresponding transaction items, goods flow and fund recovery are read according to the time sequence relationship. The transaction items are merged according to the transaction occurrence time, transaction amount and transaction object to form a transaction flow. The goods flow is merged according to the goods batch, flow time and flow status to form a goods flow. The fund recovery is merged according to the recovery time, recovery amount and recovery object to form a fund flow, thus generating a multi-stream data set for financing business. Map the transaction flow, goods flow and capital flow in the multi-flow data set of financing business to multi-flow hyperedges to generate a multi-flow hyperedge set of financing business. Based on the multi-flow super-edge set of financing business, a correlation graph of steel supply chain finance business is constructed according to the direction of connection between transaction flow, goods flow and capital flow under the same financing business. The construction process of the steel supply chain finance business association graph is as follows: Based on the multi-flow superedge set of financing business, the transaction flow, goods flow and capital flow under the same financing business are read one by one, and the transaction flow, goods flow and capital flow are encoded in a graph according to the receiving direction of the transaction flow to the goods flow and the goods flow to the capital flow; the transaction flow, goods flow and capital flow are written into the steel supply chain finance business association graph as graph nodes, the multi-flow superedges in the multi-flow superedge set of financing business are written into the cross-flow association relationship between the transaction flow, goods flow and capital flow, and graph edges are generated according to the receiving direction of the transaction flow, goods flow and capital flow under the same financing business, thus forming the steel supply chain finance business association graph.

[0021] In this embodiment, the generation of the steel supply chain financial risk characteristic set specifically includes: Based on the business relationship graph of steel supply chain finance, the corresponding multi-flow hyperedges are read according to the same financing business to determine the verification skeleton; The process of determining the verification skeleton is as follows: Based on the steel supply chain finance business association graph, the corresponding multi-flow hyperedges are read using the same financing business as the index, and the transaction flow, goods flow and capital flow connected by the multi-flow hyperedges are sequentially anchored according to the direction of the transaction flow, goods flow and capital flow in the financing business; when the multi-flow hyperedge covers the transaction flow, goods flow and capital flow under the same financing business, and the connection direction of the multi-flow hyperedge is consistent with the direction of the graph edge in the steel supply chain finance business association graph, the multi-flow hyperedge is determined as the verification skeleton; Based on the verification skeleton, cross-flow closed-loop topology stitching is performed on the transaction flow, goods flow and capital flow under the same financing business to generate cross-flow closed-loop topology links; The specific process of generating a cross-flow closed-loop topology link is as follows: Based on the verification skeleton, taking the same financing business as the processing unit, the transaction flow, goods flow, and capital flow under the same financing business are read, and according to the connection scope, acceptance direction, and business affiliation relationship retained by the verification skeleton, the transaction flow, goods flow, and capital flow are stitched together in a cross-flow closed-loop topology. The cross-flow closed-loop topology stitching takes the transaction flow as the starting business flow, the goods flow as the intermediate performance flow, and the capital flow as the repayment verification flow. According to the acceptance direction of the transaction flow pointing to the goods flow and the goods flow pointing to the capital flow, the transaction item identifier in the transaction flow, the goods batch identifier in the goods flow, and the repayment record identifier in the capital flow are written into the topology connection sequence of the same financing business. The topology connection sequence is then fixed according to the direction of the graph edge in the steel supply chain finance business association graph, so that the transaction flow, goods flow, and capital flow form a closed topology structure with a starting point, acceptance segment, and repayment endpoint under the same financing business, thus generating a cross-flow closed-loop topology link. Perform residual inversion processing on the cross-flow closed-loop topology link to determine the cross-flow closed-loop breakpoint; The process of determining cross-flow closed-loop breakpoints is as follows: Based on the cross-flow closed-loop topology, according to the topological connection sequence of transaction flow, goods flow, and capital flow under the same financing business, the transaction item identifier, goods batch identifier, repayment record identifier, business occurrence time, transaction amount, goods circulation quantity, and repayment amount are read sequentially; using the connection scope, acceptance direction, and business affiliation relationship recorded in the verification skeleton as the inversion benchmark, residual inversion processing is performed on the numerical correspondence between the transaction item identifier to the goods batch identifier between the transaction flow and the goods flow, the goods batch identifier to the repayment record identifier between the goods flow and the capital flow, the transaction amount, the goods circulation quantity, and the repayment amount; the residual inversion processing compares the actual acceptance results in the cross-flow closed-loop topology with the benchmark acceptance results in the verification skeleton item by item to obtain the acceptance sequence residual, the flow direction corresponding residual, and the numerical correspondence residual, and then reverses the location of the residual to the corresponding transaction flow, goods flow, or capital flow; the locations where there are acceptance interruptions, flow direction reversals, inconsistent business affiliations, or inconsistent numerical correspondences are determined as cross-flow closed-loop breakpoints. Risk vectorization is performed on the cross-flow closed-loop breakpoints to generate multi-dimensional risk characteristics of steel supply chain finance, and these characteristics are aggregated to form a set of risk characteristics of steel supply chain finance. The process of forming the risk feature set of steel supply chain finance is as follows: Based on the cross-flow closed-loop breakpoint, the transaction flow, goods flow, and capital flow corresponding to the cross-flow closed-loop breakpoint are read, and the graph edge position, multi-flow superedge, business affiliation relationship, acceptance sequence residual, flow direction residual, and numerical residual of the cross-flow closed-loop breakpoint in the cross-flow closed-loop topology are extracted; the risk vectorization labeling of the cross-flow closed-loop breakpoint is performed. The risk vectorization labeling is performed according to a fixed field order, labeling the graph edge position as the position risk field, the multi-flow superedge as the association risk field, the business affiliation relationship as the affiliation risk field, the acceptance sequence residual as the acceptance risk field, the flow direction residual as the flow direction risk field, and the numerical residual as the numerical risk field, generating multi-dimensional risk features of steel supply chain finance; the multi-dimensional risk features of steel supply chain finance are aggregated according to the same financing business to form the steel supply chain finance risk feature set.

[0022] In this embodiment, the generation of the set of financial risk transmission paths in the steel supply chain specifically includes: Based on the risk characteristics set of steel supply chain finance, the transaction flow, goods flow or capital flow where the cross-flow closed loop breakpoints corresponding to the multi-dimensional risk characteristics of steel supply chain finance are located are identified as risk triggering sources. The process for determining risk triggers is as follows: Based on the risk characteristic set of steel supply chain finance, the corresponding multi-dimensional risk characteristics of steel supply chain finance are read according to the same financing business. From these multi-dimensional risk characteristics, the cross-flow closed-loop breakpoint, the corresponding financing business, its position on the graph edge, and the transaction flow, goods flow, and capital flow corresponding to the cross-flow closed-loop breakpoint are extracted. When the cross-flow closed-loop breakpoint is located between the transaction flow and the goods flow, the corresponding transaction flow and goods flow are identified as risk triggers. When the cross-flow closed-loop breakpoint is located between the goods flow and the capital flow, the corresponding goods flow and capital flow are identified as risk triggers. When the cross-flow closed-loop breakpoint simultaneously causes an anomaly in the connection between the transaction flow, goods flow, and capital flow, the corresponding transaction flow, goods flow, and capital flow are identified as risk triggers. Based on the risk trigger source, in the steel supply chain finance business association graph, starting from the risk trigger source, the same source edge domain is pruned along the graph edge bearing direction corresponding to the same financing business to generate a risk tracing graph edge set. The process of generating the risk tracing graph edge set is as follows: Based on the risk trigger source, in the steel supply chain finance business association graph, the transaction flow, goods flow or capital flow where the risk trigger source is located is taken as the starting point. The graph edges directly connected to the risk trigger source are read, and the same-origin edge domains are pruned according to the business affiliation relationship and the direction of graph edge acceptance corresponding to the same financing business. Same-origin edge domain pruning is to retain graph edges that belong to the same financing business as the risk trigger source and extend along the direction of acceptance from transaction flow to goods flow and goods flow to capital flow, and delete graph edges that have no graph edge connection relationship with the risk trigger source or whose graph edge acceptance direction is opposite to that of the graph edge. The retained graph edges are summarized to generate the risk tracing graph edge set. Based on the risk tracing graph edge set, determine the risk potential gradient corresponding to each graph edge, and generate the graph edge risk potential gradient set. The process of generating the graph edge risk potential gradient set is as follows: Based on the risk tracing graph edge set, the graph edges are read one by one according to the direction of acceptance, and the multi-dimensional risk features of steel supply chain finance corresponding to the starting and ending points of the graph edges are extracted from the steel supply chain finance risk feature set; the acceptance risk field, flow risk field, and numerical risk field corresponding to the starting point of the graph edge are determined as the starting risk potential, and the acceptance risk field, flow risk field, and numerical risk field corresponding to the ending point of the graph edge are determined as the ending risk potential; according to the direction of acceptance, the difference between the ending risk potential and the starting risk potential is determined as the risk potential gradient corresponding to the graph edge; the risk potential gradients corresponding to each graph edge in the risk tracing graph edge set are summarized to generate the graph edge risk potential gradient set. Based on the graph edge risk potential gradient set, risk potential gradient tracing is performed along the graph edges in the risk tracing graph edge set to generate a set of risk propagation candidate paths. The process of generating the risk transmission candidate path set is as follows: Based on the graph edge risk potential energy gradient set, the graph edge connected to the risk trigger source in the risk tracing graph edge set is used as the tracing starting point. The corresponding risk potential energy gradient is read edge by edge according to the edge's direction of connection, and risk potential energy gradient tracing is performed. Risk potential energy gradient tracing uses the edge's direction of connection as the tracing direction and the graph edge risk potential energy gradient as the tracing driving force. Starting from the graph edge connected to the risk trigger source, the next graph edge that is continuously connected within the same financing business and has the same risk potential energy gradient direction as the current graph edge is selected sequentially. The risk potential energy gradients between adjacent edges are then recorded. The intensity change is written into the path extension judgment process; when the risk potential gradient direction of the next graph edge is consistent with the risk potential gradient direction of the current graph edge, the next graph edge is connected to the current path; when the next graph edge and the current graph edge do not belong to the same financing business, the direction of acceptance is interrupted, or the risk potential gradient direction is inconsistent, the current path extension is terminated; according to the above tracing process, the graph edges in the risk tracing graph edge set are recursively deduced, and the graph edge sequence formed from the risk trigger source and along the direction of graph edge acceptance is determined as the risk transmission candidate path, and all risk transmission candidate paths are summarized to generate a risk transmission candidate path set; Perform path potential energy continuity compression on the risk transmission candidate path set to generate a steel supply chain finance risk transmission path set; The specific process of generating the risk transmission path set for steel supply chain finance is as follows: Based on the risk transmission candidate path set, the graph edges arranged according to the edge acceptance direction in each risk transmission candidate path are read, and the risk potential energy gradients corresponding to each graph edge in the graph edge risk potential energy gradient set are called to perform path potential energy continuity compression on the risk transmission candidate path. Path potential energy continuity compression involves scanning each edge of the same risk transmission candidate path according to the edge acceptance order, compressing adjacent graph edges belonging to the same financing business, with continuous graph edge acceptance directions and consistent risk potential energy gradient directions into continuous potential energy segments, and recording the starting point of the continuous potential energy segment. The system identifies the starting edge, ending edge, cumulative risk potential gradient, and corresponding transaction flow, goods flow, and capital flow. When a change in financing business ownership occurs between adjacent edges, the edge connection direction is interrupted, or the risk potential gradient direction is reversed, the position is designated as a compression boundary, and the compression of the current continuous potential segment is terminated. Each continuous potential segment is reorganized according to the connection order in the original risk transmission candidate path, and duplicate intermediate edge records and duplicate risk potential gradient records are deleted. This generates a risk transmission path with continuous potential direction, clear start and end boundaries, and business ownership relationships, forming a set of risk transmission paths for steel supply chain finance.

[0023] In this embodiment, the generation of the steel supply chain finance risk assessment results specifically includes: The improved EGATCL model inputs the set of risk transmission paths in steel supply chain finance. This improved model includes a path potential energy spectrum folding module, a breakpoint source-flow coupling inversion module, a same-source-different-path comparison and calibration module, and a risk magnitude tomographic assessment module. The improvements are as follows: the traditional EGATCL model enhances the representation of node and edge relationships in the graph structure and distinguishes representation differences between different graph samples through graph comparison learning. The improved EGATCL model takes the set of risk transmission paths in steel supply chain finance as input, introduces the risk potential energy gradient into the path-level expression of the graph edges, forms the set of potential energy spectra of steel supply chain finance paths through potential energy ridge rearrangement and folding, determines the set of breakpoint source-flow coupling relationships through coupled phase rollback, and introduces a risk potential energy tomographic projection mechanism to comprehensively evaluate the main risk transmission path, accompanying risk transmission path, risk potential energy concentration, risk trigger source diffusion range, and cross-flow closed-loop breakpoint aggregation location in the same financing business, generating a steel supply chain finance risk assessment result. The improved EGATCL model has the following structure: The improved EGATCL model connects the path potential energy spectrum folding module, the breakpoint source-flow coupling inversion module, the same-source-different-path comparison and calibration module, and the risk magnitude tomography assessment module sequentially according to the data transmission order. After the set of risk transmission paths in the steel supply chain finance is input, it is first converted into a path edge potential energy tensor. The dimensions of the path edge potential energy tensor are the number of risk transmission paths, the graph edge connection number, and the risk potential energy gradient field. The risk potential energy gradient field includes the risk potential energy gradient direction and the risk potential energy gradient intensity. The path potential energy spectrum folding module converts the path edge potential energy tensor into a steel supply chain finance model containing potential energy ridge segments and potential energy inflection points. The system includes: a supply chain finance path potential energy spectrum set; a breakpoint source-flow coupling inversion module that receives the steel supply chain finance path potential energy spectrum set and outputs a set of breakpoint source-flow coupling relationships including risk triggering sources, cross-flow closed-loop breakpoints, business flow direction, phase offset position, and risk potential energy gradient transmission direction; a same-source-different-path comparison and calibration module that receives the breakpoint source-flow coupling relationship set and outputs same-source-different-path comparison and calibration results including the main risk transmission path, accompanying risk transmission path, and redundant path segments; and a risk magnitude tomographic assessment module that receives same-source-different-path comparison and calibration results and outputs steel supply chain finance risk assessment results including risk magnitude, risk concentration position, and risk transmission intensity. The training data for the improved EGATCL model comes from historical business samples of supply chain finance in the steel industry, including financing application records, transaction contracts, order data, invoice data, warehousing data, cargo flow data, fund recovery data, ownership review records, credit limit adjustment records, overdue records, and risk disposal records. The annotation method involves jointly annotating historical financing transactions according to actual disposal results and business loop status. The annotation content includes risk magnitude labels, risk trigger source labels, cross-flow loop breakpoint labels, main risk transmission path labels, and business section risk labels. The loss function adopts a multi-objective constraint design, consisting of risk magnitude assessment loss, path potential spectrum consistency loss, breakpoint source flow rewind loss, same-source-different-path calibration loss, and business section projection loss. The risk magnitude assessment loss constrains the final risk assessment result; the path potential spectrum consistency loss constrains the continuous expression of the risk potential gradient in the graph edge continuation order; the breakpoint source flow rewind loss constrains the backtracking consistency between the risk trigger source and the cross-flow loop breakpoint; and the same-source-different-path calibration loss... The calibration loss is used to suppress the interference of redundant path segments on the main risk transmission path, and the business section projection loss is used to constrain the stability of risk projection in transaction flow, goods flow, and capital flow. The training weights of the above five types of losses are set to 0.35, 0.20, 0.20, 0.15, and 0.10 respectively. The training parameters are set as follows: 120 training rounds, batch size of 32, initial learning rate of 0.008, linear warm-up training for the first 10 rounds, and then step decay according to the decrease rate of the validation set loss. The maximum path length is set to 16 graph edges, the potential energy ridge retention rate is set to 0.70, the breakpoint phase rollback depth is set to 4 levels of graph edges, the differential path perturbation rate is set to 0.25, the service section projection occlusion rate is set to 0.15, and the convergence condition is: when the overall loss of the validation set decreases by less than 0.005 for 8 consecutive rounds, and the accuracy of risk level identification, the hit rate of the main risk transmission path, and the stability rate of cross-flow closed loop breakpoint localization no longer improve, training is stopped and the parameters of the current improved EGATCL model are saved. In the path potential energy spectrum folding module, the risk transmission path in the set of steel supply chain finance risk transmission paths is used as the processing unit. The risk potential energy gradient corresponding to the graph edge is rearranged and folded according to the graph edge acceptance order to generate the set of steel supply chain finance path potential energy spectrum. The specific process for generating the potential energy spectrum set of steel supply chain finance paths is as follows: Each risk transmission path in the steel supply chain finance risk transmission path set is used as a processing unit. The risk potential energy gradient corresponding to each graph edge is read sequentially according to the order of edge connection in the risk transmission path, and the direction and intensity of the risk potential energy gradient between consecutive graph edges are sequentially calibrated. The calibrated risk potential energy gradients are then rearranged and folded using potential energy ridge lines. This rearrangement and folding involves merging graph edges with consistent risk potential energy gradient directions and continuous edge connections into potential energy ridge segments. Graph edges where the risk potential energy gradient intensity changes or the risk potential energy gradient direction reverses are identified as potential energy inflection points. These are then rearranged according to the connection positions of the potential energy ridge segments and potential energy inflection points in the risk transmission path, so that each risk transmission path forms a path potential energy spectrum composed of potential energy ridge segments, potential energy inflection points, and corresponding risk potential energy gradients. Finally, the path potential energy spectra corresponding to all risk transmission paths in the steel supply chain finance risk transmission path set are summarized to generate the steel supply chain finance path potential energy spectrum set. In the breakpoint source-flow coupling inversion module, based on the potential energy spectrum set of the steel supply chain finance path, the transaction flow, cargo flow and capital flow associated with the cross-flow closed loop breakpoint are coupled and phase-wounded to determine the source-flow coupling relationship between the risk triggering source, the cross-flow closed loop breakpoint, the transaction flow, the cargo flow and the capital flow, and generate a breakpoint source-flow coupling relationship set. The specific process of generating the set of source-flow coupling relationships at breakpoints is as follows: Based on the potential energy spectrum set of the steel supply chain finance path, the potential energy ridge segment, potential energy inflection point, graph edge connection order, and risk potential energy gradient folding result in the potential energy spectrum of each path are read, and the transaction flow, goods flow, and capital flow associated with the cross-flow closed-loop breakpoint are located; the transaction flow, goods flow, and capital flow associated with the cross-flow closed-loop breakpoint are coupled with phase rollback. The coupled phase rollback takes the cross-flow closed-loop breakpoint as the rollback endpoint and traces back the direction and intensity changes of the risk potential energy gradient in reverse according to the graph edge connection order in the path potential energy spectrum, and rolls back the abnormal transaction phase in the transaction flow, the abnormal ownership phase in the goods flow, and the abnormal repayment phase in the capital flow. Under the same risk triggering source, and based on the potential energy inflection point, determine the phase offset position between the transaction flow, the goods flow, and the capital flow. Based on the risk triggering source, the cross-flow closed loop breakpoint, the business flow direction, the phase offset position, and the transmission direction of the risk potential energy gradient, determine the source-flow coupling relationship between the risk triggering source, the cross-flow closed loop breakpoint, the transaction flow, the goods flow, and the capital flow. The source-flow coupling relationship includes the source position corresponding to the risk triggering source, the breakpoint position corresponding to the cross-flow closed loop breakpoint, the business flow direction, the phase offset position, and the transmission direction of the risk potential energy gradient between the transaction flow, the goods flow, and the capital flow. All source-flow coupling relationships are summarized to generate a breakpoint source-flow coupling relationship set. In the same source but different path comparison and calibration module, based on the set of breakpoint source-flow coupling relationships, the same source but different path comparison and calibration is performed on different risk transmission paths according to the same financing business, and the same source but different path comparison and calibration results are generated. The specific process for generating the calibration results of the same-source-different-path comparison is as follows: Based on the set of source-flow coupling relationships at breakpoints, the source-flow coupling relationships corresponding to risk triggering sources, cross-flow closed-loop breakpoints, transaction flows, goods flows, and capital flows under the same financing business are aggregated to form a same-source calibration domain; within the same-source calibration domain, the risk transmission path with complete source-flow coupling relationships, continuous graph edges, and coverage of transaction flows, goods flows, and capital flows is determined as the same-source benchmark path, and the remaining risk transmission paths under the same financing business are determined as the different-path paths to be calibrated; the risk triggering source positions between the different-path paths to be calibrated and the same-source benchmark paths are then compared. The location of the cross-flow closed loop breakpoint, business flow direction, graph edge coverage, and risk transmission direction are aligned item by item. Path segments that repeatedly cover the same business flow direction and the same cross-flow closed loop breakpoint are merged. Path segments that deviate from the same source baseline path but still point to the same cross-flow closed loop breakpoint are retained as different paths. Path segments with inconsistent financing business affiliation or inconsistent source-flow coupling relationship are pruned. Based on the merged path segments, the retained path segments, and the pruned path segments, the main risk transmission path, the accompanying risk transmission path, and the redundant path segments under the same financing business are determined, and the same-source different-path comparison calibration results are generated. In the risk level analysis and assessment module, a risk potential energy analysis and projection mechanism is introduced. Taking the same financing business as the projection object, the risk transmission path in the same source but different path comparison and calibration results is projected to the business sections corresponding to the transaction flow, goods flow and capital flow according to the graph edge connection order. The risk potential energy gradient, risk triggering source and cross-flow closed loop breakpoint in the risk transmission path are mapped by cross-sectional potential energy to generate risk analysis and projection results. The degree of risk potential energy concentration, risk triggering source diffusion range and cross-flow closed loop breakpoint clustering position of the same financing business in the transaction flow, goods flow and capital flow are comprehensively evaluated to generate the steel supply chain finance risk assessment results. The process of generating risk assessment results for steel supply chain finance is as follows: Based on the calibration results of the same source but different paths comparison, the corresponding main risk transmission path, accompanying risk transmission path, and redundant path segments are read according to the same financing business, and a risk potential energy layering projection mechanism is introduced; the risk potential energy layering projection mechanism takes the same financing business as the projection object, and takes the transaction flow, goods flow, and capital flow as three business sections. The main risk transmission path and accompanying risk transmission path are projected to the corresponding business sections according to the graph edge connection order, and the risk potential energy gradient, risk trigger source, and cross-flow closed loop breakpoint in the risk transmission path are mapped to the section potential energy distribution, section source point distribution, and section breakpoint distribution, respectively, generating risk layering projection results; based on the risk layering projection results, the transaction flow, goods flow, and capital flow are statistically analyzed. The degree of concentration of risk potential energy distribution across the breakpoints is determined to ascertain the degree of risk potential energy concentration. The coverage of the source points in the three business sections is statistically analyzed to determine the diffusion range of risk triggering sources. The clustering locations of the breakpoints in the three business sections are statistically analyzed to determine the clustering locations of cross-current closed-loop breakpoints. Based on the degree of risk potential energy concentration, the diffusion range of risk triggering sources, and the clustering locations of cross-current closed-loop breakpoints, a comprehensive assessment of the same financing business is conducted. The comprehensive assessment is based on the risk strength corresponding to the degree of risk potential energy concentration, the risk impact range corresponding to the diffusion range of risk triggering sources, and the risk concentration links corresponding to the clustering locations of cross-current closed-loop breakpoints. The risk status of the same financing business is analyzed and merged to form a steel supply chain finance risk assessment result that includes risk magnitude, risk concentration location, and risk transmission intensity.

[0024] In this implementation method, the generation of the steel supply chain finance risk management result specifically includes: Read the risk level, risk concentration location and risk transmission intensity corresponding to the risk assessment results of steel supply chain finance, and generate a set of risk strategy trigger parameters; The process of generating the risk strategy trigger parameter set is as follows: Based on the risk assessment results of steel supply chain finance, the corresponding risk level, risk concentration location, and risk transmission intensity are read according to the same financing business. The risk level is mapped to the credit limit contraction trigger parameter, the risk concentration location is mapped to the title review trigger parameter and the transaction review trigger parameter, and the risk transmission intensity is mapped to the repayment lock-in trigger parameter. The credit limit contraction trigger parameter, title review trigger parameter, repayment lock-in trigger parameter, and transaction review trigger parameter are aggregated for the same financing business to form the risk strategy trigger parameter set. Based on the set of risk strategy trigger parameters, risk strategy sandbox simulation is conducted on financing business strategies such as limit reduction, title verification, repayment lock-in and transaction verification, and risk strategy sandbox simulation results are generated. The process of generating risk strategy simulation results is as follows: Based on the risk strategy trigger parameter set, the trigger parameters for credit limit contraction, title verification, repayment lock-in, and transaction verification corresponding to the same financing business are read. The credit limit contraction strategy is applied to the financing limit, the title verification strategy is applied to the title status, the repayment lock-in strategy is applied to the repayment account, and the transaction verification strategy is applied to the transaction document. During the risk strategy simulation, the changes in risk magnitude, risk concentration location, risk transmission intensity, and cross-loop breakpoint after the execution of the credit limit contraction strategy, title verification strategy, repayment lock-in strategy, and transaction verification strategy are simulated respectively. The changes after the execution of each strategy are compared with the steel supply chain finance risk assessment results before the strategy execution, generating risk strategy simulation results that include the credit limit contraction simulation results, title verification simulation results, repayment lock-in simulation results, and transaction verification simulation results. Based on the results of the risk strategy simulation, the target risk management strategy combination is determined, and the steel supply chain financial risk management results are generated. The process of generating the risk management results for steel supply chain finance is as follows: Based on the results of risk strategy simulation, the results of credit limit contraction simulation, title verification simulation, payment lock-in simulation, and transaction verification simulation are read. The effects of the strategies are compared by analyzing the changes in risk magnitude, risk concentration location, risk transmission intensity, and cross-loop breakpoints corresponding to each simulation result. Based on the strategy effect comparison results, strategies that can reduce the risk magnitude, converge the risk concentration location, weaken the risk transmission intensity, and reduce the impact of cross-loop breakpoints are selected to form a target risk management strategy combination. The target risk management strategy combination is applied to the financing amount, title status, payment account, and transaction documents corresponding to the same financing business to generate credit limit contraction results, title verification results, payment lock-in results, and transaction verification results, which are then summarized to form the steel supply chain finance risk management results.

[0025] Example 1: Taking a steel trade supply chain finance platform in East China as an application scenario, this platform serves steel purchasers, steel mills, traders, warehouse supervisors, logistics carriers, and financial institutions. Its business includes steel order financing, warehouse receipt pledge financing, accounts receivable financing, and payment collection supervision. Before trial operation, the platform primarily relied on manual verification of contracts, invoices, warehouse receipts, logistics receipts, and bank statements for risk assessment. This resulted in problems such as fragmented data sources, asynchronous verification of transaction flow, goods flow, and capital flow, delayed detection of changes in ownership status, and difficulty in timely locating payment deviations. Particularly in transactions involving significant steel price fluctuations, multiple batches of goods entering and leaving the warehouse, and multiple resales of the same batch of steel, traditional rule-based risk control could only identify single abnormal fields, making it difficult to determine whether anomalies belonged to continuous risk transmission within the same financing business chain. This led to untimely handling of some financing transactions in terms of credit limit occupancy, ownership verification, and payment collection locking.

[0026] In this scenario, the platform uses three consecutive months of data, including financing application records, purchase contracts, sales orders, invoice records, warehousing and outbound records, logistics tracks, payment receipts, credit limit adjustment records, and risk disposal records, as the raw dataset. This method first cleans the raw dataset by removing fields, unifying the subject, standardizing the time format, and merging business numbers to generate a standardized dataset for steel supply chain finance. Then, it reconstructs the event-based hash chain for contract signing, goods warehousing, goods outbound, logistics transportation, buyer receipt, and fund recovery within the same financing transaction, forming a steel supply chain finance business link dataset. Based on this business link dataset, transaction flow, goods flow, and fund flow are mapped as multi-flow hyperedges, constructing a steel supply chain finance business association graph. This allows contracts, ownership of goods, logistics, and payments within a single financing transaction to be verified and integrated within the same graph structure.

[0027] In practical operation, when a steel procurement financing transaction encounters situations such as "normal contract amount, early warehouse release, delayed logistics receipt, and change of payment account," this method maps the corresponding transaction flow, goods flow, and capital flow to a steel supply chain finance business association graph. Multi-flow hyperedges are used as a verification skeleton to identify cross-flow closed-loop breakpoints. After identifying inconsistencies between the goods flow node and the capital flow node in this transaction, the system extracts multi-dimensional risk characteristics of steel supply chain finance and identifies the goods flow and capital flow where the breakpoint is located as risk trigger sources. Subsequently, the system performs risk potential gradient tracing along the graph edges in the steel supply chain finance business association graph, forming a set of steel supply chain finance risk transmission paths. These paths are then input into an improved EGATCL model for comprehensive evaluation, outputting a steel supply chain finance risk assessment result that includes risk magnitude, risk concentration location, and risk transmission intensity. Based on the risk assessment results, the system triggers a risk strategy simulation involving credit limit reduction, ownership verification, payment lock-in, and transaction review. Ultimately, it generates the steel supply chain finance risk management results, enabling financial institutions to implement credit limit reduction, warehouse receipt verification, payment account freezing, and transaction document review for the same financing business.

[0028] During the trial operation, this embodiment processed 1,860 financing transactions, involving steel transactions totaling approximately RMB 4.28 billion, covering 126 steel mills and trading entities, 38 supervised warehouses, 64 logistics carrier nodes, and 312 payment accounts. The system processed approximately 296,000 records related to contracts, orders, warehousing, logistics, invoices, payments, and credit limits. To verify the effectiveness, three months of data from before the trial operation, using traditional rule-based risk control, were selected as the control group, while data from the trial operation using this method was used as the experimental group. Risk control personnel also reviewed and confirmed the risk items identified by the system. The results show that this method significantly improves the ability to verify the closure of the business chain and identify cross-flow risks. It provides more timely identification of abnormal ownership of goods, payment deviations, inconsistent transaction documents, and multi-flow connection breakpoints, shifting risk handling from post-event manual review to dynamic management during the financing process.

[0029] Table 1 Comparison of the Trial Implementation Effects of Steel Supply Chain Financial Risk Management Methods

[0030] As shown in Table 1, this method has achieved significant improvements in business link verification, cross-flow closed-loop breakpoint identification, risk transmission path review, and risk strategy generation. Traditional rule-based risk control methods require manual verification of contract, warehousing, logistics, and payment data, with an average verification time of 42.5 minutes per business link closure. However, this method, through the standardized dataset of steel supply chain finance, event tracing hash chain reconstruction, and steel supply chain finance business association graph, achieves unified processing, reducing the average time to 8.7 minutes. This indicates that this method can effectively reduce the verification cost caused by the decentralized processing of multi-source data. At the same time, the accuracy of cross-flow closed-loop breakpoint identification has increased from 76.8% to 91.5%, indicating that multi-flow super-edge and risk potential gradient tracing can more accurately locate the breakpoints between transaction flow, goods flow, and capital flow, solving the problem that existing technologies are unable to identify cross-flow risks.

[0031] From the perspective of risk management results, this method integrates the risk assessment results of steel supply chain finance with risk strategy simulations involving credit limit reduction, title verification, payment lock-in, and transaction review strategies. This allows risk identification results to be directly translated into management actions. During the trial operation, the early warning rate for high-risk financing transactions increased from 68.4% to 86.7%, and the false alarm rate decreased from 18.9% to 9.6%. This indicates that this method not only improves risk detection capabilities but also reduces interference with normal financing transactions. The abnormal confirmation rate after title verification and the abnormal payment interception rate after payment lock-in both showed significant improvements. This demonstrates that the steel supply chain finance risk management results generated by this method can more accurately affect financing limits, title status, payment accounts, and transaction documents, thereby improving the timeliness, accuracy, and business adaptability of supply chain finance risk management in the steel industry.

[0032] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A big data-based method for supply chain finance risk management in the steel industry, characterized in that, Includes the following steps: Collect and preprocess the raw datasets from the steel industry supply chain finance business to generate a standardized dataset for steel supply chain finance. Reconstruct the event tracing hash chain of the standardized dataset for steel supply chain finance to determine the temporal sequence relationship between business entities, transactions, goods flow and fund recovery, and generate a steel supply chain finance business link dataset. Using the temporal sequence relationship in the data set of steel supply chain finance business links as the main axis, the transaction flow, goods flow and capital flow under the same financing business are mapped as multi-flow hyperedges to construct a steel supply chain finance business association graph; The multi-flow hyperedge in the steel supply chain finance business association graph is determined as the verification skeleton, the cross-flow closed loop breakpoints corresponding to the same financing business are identified, the multi-dimensional risk characteristics of steel supply chain finance are extracted, and the steel supply chain finance risk characteristic set is generated. By identifying risk triggers through the set of risk characteristics of steel supply chain finance, and tracing the risk potential gradient along the graph edges of the steel supply chain finance business association graph, a set of risk transmission paths for steel supply chain finance is generated. The set of risk transmission paths in the steel supply chain finance is input into the improved EGATCL model for comprehensive evaluation, generating the risk assessment results of the steel supply chain finance. Based on the risk assessment results of steel supply chain finance, a risk strategy simulation was conducted to simulate strategies for reducing financing limits, verifying ownership of goods, locking in payments, and verifying transactions, thereby generating risk management results for steel supply chain finance.

2. The method for supply chain finance risk management in the steel industry based on big data as described in claim 1, characterized in that, The original dataset includes order data, contract data, invoice data, logistics data, warehousing data, accounts receivable data, cash flow data, enterprise credit data, and steel price data. The preprocessing includes field normalization, time alignment, and identifier encoding.

3. The method for supply chain finance risk management in the steel industry based on big data as described in claim 1, characterized in that, The generation of the steel supply chain finance business link dataset specifically includes: Based on the standardized dataset of steel supply chain finance, events are grouped according to the business entity, transaction, goods flow and fund recovery corresponding to the same financing business to generate a financing business event sequence. Set up time slots according to the business sequence of business entity, transaction, goods flow and fund recovery, and write the financing business event sequence into the corresponding time slot to generate the financing business time slot sequence. Perform joint hash anchoring on the time-series slot sequence of financing business to generate an event tracing hash chain; Perform chain-based verification on the event tracing hash chain to determine the corresponding temporal sequence relationship; Based on the temporal sequence relationship, the business entities, transactions, goods transfer and fund recovery corresponding to the same financing business are encapsulated into a chain to generate a steel supply chain finance business chain dataset.

4. The method for supply chain finance risk management in the steel industry based on big data as described in claim 1, characterized in that, The construction of the steel supply chain finance business relationship map specifically includes: Taking the time sequence relationship of the steel supply chain finance business link data as the main axis, the business entities, transaction items, goods flow and fund recovery are divided into link units to generate a set of financing business link units. Based on the set of financing business link units, transactions are grouped into transaction flows, goods transfers are grouped into goods flows, and fund collections are grouped into fund flows, generating a multi-flow data set for financing business. Map the transaction flow, goods flow and capital flow in the multi-flow data set of financing business to multi-flow hyperedges to generate a multi-flow hyperedge set of financing business. Based on the multi-flow super-edge set of financing business, a correlation graph of steel supply chain finance business is constructed according to the direction of connection between transaction flow, goods flow and capital flow under the same financing business.

5. The method for supply chain finance risk management in the steel industry based on big data as described in claim 1, characterized in that, The generation of the aforementioned set of financial risk characteristics for the steel supply chain specifically includes: Based on the business relationship graph of steel supply chain finance, the corresponding multi-flow hyperedges are read according to the same financing business to determine the verification skeleton; Based on the verification skeleton, cross-flow closed-loop topology stitching is performed on the transaction flow, goods flow and capital flow under the same financing business to generate cross-flow closed-loop topology links; Perform residual inversion processing on the cross-flow closed-loop topology link to determine the cross-flow closed-loop breakpoint; Risk vectorization is performed on the cross-flow closed-loop breakpoints to generate multi-dimensional risk characteristics of steel supply chain finance, and these characteristics are aggregated to form a set of risk characteristics of steel supply chain finance.

6. The method for supply chain finance risk management in the steel industry based on big data as described in claim 1, characterized in that, The generation of the set of financial risk transmission paths in the steel supply chain specifically includes: Based on the risk characteristics set of steel supply chain finance, the transaction flow, goods flow or capital flow where the cross-flow closed loop breakpoints corresponding to the multi-dimensional risk characteristics of steel supply chain finance are located are identified as risk triggering sources. Based on the risk trigger source, in the steel supply chain finance business association graph, starting from the risk trigger source, the same source edge domain is pruned along the graph edge bearing direction corresponding to the same financing business to generate a risk tracing graph edge set. Based on the risk tracing graph edge set, determine the risk potential gradient corresponding to each graph edge, and generate the graph edge risk potential gradient set. Based on the graph edge risk potential gradient set, risk potential gradient tracing is performed along the graph edges in the risk tracing graph edge set to generate a set of risk propagation candidate paths. Perform path potential energy continuity compression on the candidate path set of risk transmission to generate a set of risk transmission paths for the steel supply chain finance.

7. The method for supply chain finance risk management in the steel industry based on big data as described in claim 1, characterized in that, The generation of the steel supply chain financial risk assessment results specifically includes: The improved EGATCL model inputs the set of risk transmission paths in steel supply chain finance. This improved model includes a path potential energy spectrum folding module, a breakpoint source-flow coupling inversion module, a same-source-different-path comparison and calibration module, and a risk magnitude tomographic assessment module. The improvements are as follows: the traditional EGATCL model enhances the representation of node and edge relationships in the graph structure and distinguishes representation differences between different graph samples through graph comparison learning. The improved EGATCL model takes the set of risk transmission paths in steel supply chain finance as input, introduces the risk potential energy gradient into the path-level expression of the graph edges, forms the set of potential energy spectra of steel supply chain finance paths through potential energy ridge rearrangement and folding, determines the set of breakpoint source-flow coupling relationships through coupled phase rollback, and introduces a risk potential energy tomographic projection mechanism to comprehensively evaluate the main risk transmission path, accompanying risk transmission path, risk potential energy concentration, risk trigger source diffusion range, and cross-flow closed-loop breakpoint aggregation location in the same financing business, generating a steel supply chain finance risk assessment result. In the path potential energy spectrum folding module, the risk transmission path in the set of steel supply chain finance risk transmission paths is used as the processing unit. The risk potential energy gradient corresponding to the graph edge is rearranged and folded according to the graph edge acceptance order to generate the set of steel supply chain finance path potential energy spectrum. In the breakpoint source-flow coupling inversion module, based on the potential energy spectrum set of the steel supply chain finance path, the transaction flow, cargo flow and capital flow associated with the cross-flow closed loop breakpoint are coupled and phase-wounded to determine the source-flow coupling relationship between the risk triggering source, the cross-flow closed loop breakpoint, the transaction flow, the cargo flow and the capital flow, and generate a breakpoint source-flow coupling relationship set. In the same source but different path comparison and calibration module, based on the set of breakpoint source-flow coupling relationships, the same source but different path comparison and calibration is performed on different risk transmission paths according to the same financing business, and the same source but different path comparison and calibration results are generated. In the risk level analysis and assessment module, a risk potential energy layering projection mechanism is introduced. Taking the same financing business as the projection object, the risk transmission path in the same source but different path comparison calibration results is projected onto the corresponding business sections of the transaction flow, goods flow, and capital flow according to the graph edge connection order. The risk potential energy gradient, risk triggering source, and cross-flow closed loop breakpoint in the risk transmission path are mapped by cross-sectional potential energy to generate risk layering projection results. The degree of risk potential energy concentration, risk triggering source diffusion range, and cross-flow closed loop breakpoint clustering position of the same financing business in the transaction flow, goods flow, and capital flow are comprehensively evaluated to generate the steel supply chain finance risk assessment results.

8. A method for supply chain finance risk management in the steel industry based on big data, as described in claim 1, is characterized in that... The generation of the steel supply chain financial risk management results specifically includes: Read the risk level, risk concentration location and risk transmission intensity corresponding to the risk assessment results of steel supply chain finance, and generate a set of risk strategy trigger parameters; Based on the set of risk strategy trigger parameters, risk strategy sandbox simulation is conducted on financing business strategies such as limit reduction, title verification, repayment lock-in and transaction verification, and risk strategy sandbox simulation results are generated. Based on the results of the risk strategy simulation, the target risk management strategy combination is determined, and the steel supply chain financial risk management results are generated.

Images