A comprehensive assessment and diagnostic system for corporate financial risk control

Abstract

This application provides a comprehensive assessment and evaluation system for enterprise financial risk control, applied in the field of computer data processing and risk analysis technology. The system's functional block division module divides capital functional blocks and constructs a set of directional flow relationships. A risk sequence generation module generates a risk generation sequence based on the capital functional blocks. A transmission path generation module establishes a risk transmission mapping relationship based on the set of directional flow relationships and the risk generation sequence, generating a set of risk transmission paths. A closed-loop identification module identifies risk closed-loop paths based on the risk generation sequence. A contribution assessment module determines the path risk intensity identifier and the block risk contribution identifier based on the risk closed-loop path. A result generation module generates the enterprise financial risk assessment result based on the path risk intensity identifier and the block risk contribution identifier. This reduces redundant path traversal and repetitive time-series calculations in risk transmission analysis, improving the efficiency of risk assessment processing in scenarios with massive amounts of capital transaction data.

Description

Technical Field

[0001] This application relates to the fields of computer data processing and risk analysis technology, and in particular to a comprehensive assessment and diagnostic system for corporate financial risk control. Background Technology

[0002] As enterprises deepen their digital transformation of finance, internal fund transfers, related-party transactions, and supply chain settlements are becoming increasingly frequent and exhibit complex network characteristics. In this context, enterprises urgently need a comprehensive financial risk control assessment and diagnostic system that can not only handle massive, high-frequency fund transactions but also penetrate the superficial isolation between accounts to accurately identify functional clusters of funds across multiple accounts, risk transmission paths, and closed-loop fund return behaviors, in order to prevent deep-seated financial risks such as idle funds and hidden debt cycles.

[0003] In existing corporate financial risk control technologies, conventional comprehensive assessment and check-up systems typically treat each transaction account as an independent processing node. After receiving transaction records in a log format, the system calls a preset rule engine and performs isolated traversal screening of each account's transaction data based on static thresholds such as single transaction amount limits and fixed time window transaction frequency. When it is necessary to analyze the relationship between accounts, the system performs a full exhaustive scan, calculating the Cartesian product of all account pairs to find potential fund flow correlation paths, and summarizing and displaying the risk assessment results of each account.

[0004] The above-mentioned solution still has some problems in practical application. When analyzing the transmission of financial risks, the computational complexity is too high, making it impossible to complete the global correlation analysis within an acceptable time. Because the existing technology treats each account as an independent node and uses a global exhaustive matching method to search for correlation paths, as the number of accounts increases, the number of node pairs that need to be traversed increases in a flat manner, and the amount of computation and memory consumption expands sharply. This high consumption of computing power forces the system to reduce the scope of analysis in real business scenarios, and can only perform shallow screening of a few single-step financial links. As a result, it loses the ability to deeply explore the financial circulation structure composed of multiple nodes and multiple hops, making it impossible for conventional systems to effectively identify the closed-loop risks of corporate financial funds under controllable computing power costs. Summary of the Invention

[0005] This application provides a comprehensive assessment and check-up system for enterprise financial risk control. It aims to use computer data processing technology to perform functional block aggregation and dynamic time-series analysis of risk transmission paths on massive amounts of financial transaction data. Under controllable computing power consumption, it solves the problems of system response delay and deep risk omission caused by exhaustive matching of the entire domain in existing technologies, and realizes efficient and accurate assessment of enterprise capital flow network.

[0006] To achieve the above objectives, this application adopts the following technical solution: This application provides a comprehensive assessment and check-up system for enterprise financial risk control. The system includes: a functional block division module, used to perform functional division processing on the target enterprise's fund transaction data based on the fund transfer relationships and fund usage between accounts, forming multiple fund functional blocks, and statistically analyzing the fund inflow and outflow between any two fund functional blocks to construct a set of directional flow relationships between fund functional blocks; a risk sequence generation module, used to extract the financial indicator sequence of each fund functional block from the fund transaction data, and input the financial indicator sequence into a pre-trained risk state transition model to perform risk evolution and obtain a risk generation sequence; and a transmission path generation module, used to time-align the set of directional flow relationships with the risk generation sequence of the corresponding fund functional block, establishing a transmission mapping relationship between different fund functional blocks. The system generates a set of risk transmission paths between fund functional blocks based on the transmission mapping relationship; a closed-loop identification module is used to perform risk transmission processing along the risk transmission paths in the risk transmission path set according to the state change information of the risk generation sequence, and to filter risk closed-loop paths according to the processing results; a contribution assessment module is used to determine the path risk intensity identifier based on the path characteristics of the risk closed-loop path, determine the risk increment propagation value of each fund functional block in the corresponding risk closed-loop path based on the transmission mapping relationship, and generate a block risk contribution identifier for each fund functional block according to the correlation between the risk increment propagation value and the path risk intensity identifier of the corresponding risk closed-loop path; a result generation module is used to construct a risk heatmap for the fund transaction data based on the block risk contribution identifier and the path risk intensity identifier, and generate enterprise financial risk assessment results.

[0007] In some possible implementations, the transmission path generation module is specifically used to: construct a directed graph of fund function blocks, using each fund function block as a node and the directed flow relationships in the set of directed flow relationships as edges; for each directed edge in the directed graph of fund function blocks, determine the upstream and downstream fund function blocks of the directed edge, and perform feature extraction based on the state change information of the risk generation sequence by the upstream and downstream fund function blocks to obtain the risk state label and the first time step of the risk state label transition of the upstream fund function block, and the risk state label and the second time step of the risk state label transition of the downstream fund function block; calculate the time step spacing value between each upstream and downstream fund function block based on the first and second time steps, and take the time step spacing value with the highest frequency as the risk transmission delay value of the directed edge; and perform time translation processing on each risk state label transition time step of the downstream fund function block based on the risk transmission delay value. Subtract the risk transmission delay value from the second time step to obtain the corresponding alignment time step; match the alignment time step with the first time step of the upstream capital function block in chronological order to determine the alignment time step corresponding to the first time step, forming corresponding time step pairs; for each corresponding time step pair, extract the transition direction of the risk state label corresponding to the upstream capital function block in the first time step, and the transition direction of the risk state label corresponding to the downstream capital function block in the alignment time step; count the number of corresponding time steps with the same transition direction, and calculate the proportion of the number to the total number of risk state label transition time steps of the upstream capital function block, and determine the proportion as the risk state transition unidirectional rate of the corresponding directed edge; mark the directed edges with the risk state transition unidirectional rate exceeding a preset threshold as effective transmission edges, and encapsulate the effective transmission edges and the corresponding risk transmission delay value into a transmission mapping relationship; identify effective transmission paths from the directed graph of the capital function block as risk transmission paths based on the transmission mapping relationship, and summarize the risk transmission paths to obtain a risk transmission path set.

[0008] In some possible implementations, the closed-loop identification module is specifically used to: read risk state labels step-by-step from the state change information of the risk generation sequence, and progressively compare the changes in risk state labels between adjacent time steps to generate a risk state change sequence; identify the time step where a risk state transition occurs based on the risk state change sequence, determine the corresponding funding function block where the transition occurs as the risk activation block, and determine the time step where the first transition occurs as the activation timestamp; using the risk activation block as the starting node, traverse each risk transmission path in the risk transmission path set path by path, and sequentially follow the directed edges in the risk transmission path. The system first determines the subsequent funding functional blocks; based on the risk transmission delay value corresponding to each directed edge, it calculates the time step of the risk state transition of the subsequent funding functional blocks to obtain the corresponding target time step; at the target time step, it reads the risk state label of the corresponding funding functional block and compares it with the risk state label at the activation timestamp to determine whether a risk state transition in the same direction has occurred, and marks the risk transmission path corresponding to the risk state transition in the same direction as a candidate risk closed loop path; when the termination node of the candidate risk closed loop path is the same funding functional block as the risk activation block, the candidate risk closed loop path is determined as a risk closed loop path.

[0009] In some possible implementations, the contribution assessment module is specifically used for: numbering the positions of each funding functional block in the risk closed-loop path according to the transmission order to obtain a path block set; extracting the state change values ​​corresponding to adjacent funding functional blocks from the state change information of the risk generation sequence based on the path block set, and calculating the state change amount between adjacent funding functional blocks to obtain a path state change sequence; counting the number of funding functional blocks from the initial funding functional block to the return to the initial funding functional block based on the path block set, and determining the closed-loop transmission frequency value by combining the timestamp difference between the first and last funding functional blocks in the path block set; calculating the basic path risk value of the corresponding risk closed-loop path based on the path state change sequence and the closed-loop transmission frequency value; calculating the proportional relationship between the amount of return funds and the amount of initial outflow funds in the risk closed-loop path, and correcting the basic path risk value to obtain a path behavior risk value; and determining the path risk intensity identifier of the corresponding risk closed-loop path based on the path behavior risk value and a preset risk propagation association rule.

[0010] In some possible implementations, the contribution assessment module is specifically used to: along the corresponding risk closed-loop path, according to the transmission order of the funding functional blocks in the risk closed-loop path, read the state change values ​​corresponding to adjacent funding functional blocks, and calculate the state offset of the state change value corresponding to the later funding functional block relative to the state change value corresponding to the previous funding functional block; based on the transmission mapping relationship, determine the funding transmission time interval between adjacent funding functional blocks, and perform time delay correction processing on the state offset according to the funding transmission time interval to obtain the path state increment value; extract the funding inflow and funding outflow between adjacent funding functional blocks in the corresponding risk closed-loop path, and perform propagation offset correction processing on the corresponding path state increment value according to the offset relationship between the funding inflow and funding outflow to obtain the risk increment propagation value of the corresponding funding functional block.

[0011] In some possible implementations, the contribution assessment module is specifically used for: obtaining a set of path blocks corresponding to the risk closed-loop path, and reading the position number of each funding functional block in the risk closed-loop path from the set of path blocks; reading the path risk intensity identifier corresponding to the risk closed-loop path, and associating the path risk intensity identifier as a path risk benchmark value with each funding functional block in the funding functional block sequence; for each funding functional block in the set of path blocks, extracting the corresponding risk increment propagation value, and calculating the total propagation of the risk closed-loop path based on the risk increment propagation value corresponding to each funding functional block; calculating the propagation proportion of each funding functional block based on the total propagation, and allocating the path risk benchmark value to the corresponding funding functional block according to the propagation proportion to obtain the path allocation risk value of each funding functional block; performing calculations on the path allocation risk value of each funding functional block and the risk increment propagation value of the corresponding funding functional block to obtain the risk contribution value of each funding functional block; and generating a block risk contribution identifier for each funding functional block based on the size ranking relationship of the risk contribution values ​​of each funding functional block in the same risk closed-loop path.

[0012] In some possible implementations, the risk sequence generation module is specifically used to: extract net cash flow, cash turnover rate, and short-term debt coverage ratio of each financial functional block in each statistical period from the fund transaction data according to the division results of each financial functional block, and arrange the financial indicators in the order of time steps to form a financial indicator sequence for the corresponding financial functional block; input the financial indicator sequence into a pre-trained risk state transition model, the risk state transition model including a time-series feature extraction layer, a hidden state transition layer, and a risk state output layer; wherein, the time-series feature extraction layer is used to perform a linear transformation on the financial indicator sequence of each time step to generate a local financial feature vector for the corresponding time step; the hidden state transition layer is used to receive... The system collects the local financial feature vector at the current time step and performs gated fusion calculation with the hidden state vector of the previous time step to generate the hidden state vector at the current time step. The risk state output layer performs mapping calculation on the hidden state vectors of each time step, outputs the risk state probability distribution of the corresponding time step, and determines the risk state level corresponding to the maximum probability as the risk state label of the time step. Based on the time step sequence, the risk state labels corresponding to each time step are calculated step by step, and the difference between the risk state labels of adjacent time steps is used as the state change value. The state change value and the risk state label are encapsulated as state change information. The state change information is combined with the corresponding time step and recorded to generate the risk generation sequence of the corresponding funding function block.

[0013] In some possible implementations, the functional block partitioning module is specifically used for: extracting transaction account identifiers, transaction times, transaction amounts, transaction directions, and transaction purpose information between transaction accounts from the target enterprise's fund transaction data; and, based on the transaction account identifiers, performing fund flow analysis on the transaction times, transaction amounts, and transaction directions to obtain fund transfer relationships, which at least include the number of fund transfers, the amount of fund transfers, and the time interval between fund transfers; for any two transaction accounts, determining the degree of fund transfer correlation between the transaction accounts based on the number of fund transfers, the amount of fund transfers, and the time interval between fund transfers; and, based on the transaction purpose information corresponding to each transaction account, further... The system categorizes trading accounts by their fund usage type and calculates the fund inflows and outflows for each account under different fund usage types. Based on the correlation between fund transfers between trading accounts and the corresponding fund usage types, trading accounts with the same fund usage type and whose fund transfer correlation meets preset conditions are grouped into multiple fund function blocks. The system reads the fund transaction data between trading accounts belonging to different fund function blocks and determines the directed flow relationship between fund function blocks based on the fund outflow and inflow directions in the fund transaction data. The system calculates the fund inflows and outflows corresponding to each directed flow relationship and generates a set of directional flow relationships between fund function blocks.

[0014] In some possible implementations, the result generation module is specifically used to: read the block risk contribution identifier corresponding to each capital function block and the path risk intensity identifier corresponding to each risk closed-loop path; associate the block risk contribution identifier corresponding to each capital function block with the capital inflow and capital outflow in the corresponding capital transaction data to determine the block display area corresponding to each capital function block; associate the path risk intensity identifier corresponding to each risk closed-loop path with the capital function block transmission order in the risk closed-loop path to determine the path connection area corresponding to each risk closed-loop path; and, based on the block risk contribution identifier corresponding to each capital function block and the path risk intensity identifier corresponding to each risk closed-loop path, draw a risk heat map of the block display area and the path connection area to obtain a visualized corporate financial risk assessment result, which is used to display the risk assessment status of the corporate financial risk control.

[0015] In some possible implementations, the result generation module is further configured to: perform color layering processing on the corresponding block display area based on the block risk contribution identifier corresponding to each funding function block, wherein the higher the risk contribution value corresponding to the block risk contribution identifier, the higher the color saturation of the corresponding block display area; perform color gradient processing on the corresponding path connection area based on the path risk intensity identifier corresponding to each risk closed-loop path, wherein the color brightness of the subsequent path connection area along the transmission direction of the risk closed-loop path increases progressively; when the same block display area corresponds to multiple path connection areas, perform color overlay processing on the color gradient areas corresponding to the multiple path connection areas, and generate a risk aggregation identifier for the corresponding funding function block based on the coverage of the color overlay area.

[0016] As can be seen from the above technical solution, this application has the following beneficial effects: 1. This application utilizes a functional block partitioning module to perform functional partitioning processing on massive transaction data based on the fund transfer relationships and fund usage between accounts. This aggregates discrete accounts into multiple fund functional blocks and constructs a set of directional flow relationships between blocks. This fundamentally changes the traditional flat, discrete topology network structure. While retaining the characteristics of real business fund collection and settlement, it greatly compresses the data scale of subsequent graph calculations. This allows the system to avoid the combinatorial explosion problem caused by exhaustive scanning of the entire domain based on the directional flow relationships at the functional block level. Therefore, this application can easily handle massive, high-concurrency fund transaction data with extremely low computational complexity and memory overhead. This not only significantly improves the computational efficiency of the risk control system but also enables the system to release computing power to the mining of deep multi-hop associations, thereby achieving a comprehensive and accurate assessment and check of corporate financial risks at a controllable cost.

[0017] 2. This application introduces a risk transmission delay value through a transmission path generation module. It performs time-shift alignment on the risk generation sequences of upstream and downstream blocks to establish a precise transmission mapping relationship. Furthermore, a closed-loop identification module, combined with the delay value, calculates the target time step hop-by-hop along the transmission path, rather than blindly scanning the entire window. This mechanism aligns with the asynchronous characteristics of real-world financial transactions, such as approval cycles and ongoing settlements. In actual operation, the system does not need to perform redundant exhaustive merge sorting at each time step, thus significantly reducing the frequency of random data reads and I / O throughput pressure between distributed storage nodes. By accurately locking the target time step for state comparison, the system effectively avoids the memory congestion caused by a large number of invalid intermediate results from traditional rigid matching, significantly reducing disk swapping overhead caused by memory overflow. Therefore, this application achieves precise capture of cross-cycle, multi-node hidden fund loop return behavior while significantly reducing storage I / O latency and memory load, balancing system performance with enterprise risk control effectiveness. Attached Figure Description

[0018] Figure 1 This is an example diagram illustrating the composition of a comprehensive assessment and check-up system for corporate financial risk control, as described in this application. Figure 2 This is a flowchart illustrating the closed-loop recognition process of a closed-loop recognition module according to this application. Figure 3 The evaluation flowchart for the contribution assessment module of this application. Detailed Implementation

[0019] The terms "first," "second," and "third," etc., used in this application specification, claims, and drawings are used to distinguish different objects, not to limit a specific order.

[0020] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0021] Research has revealed that existing technologies, when modeling risks in corporate financial networks, rely solely on the existence of physical transaction records between accounts to construct a static topology, simplifying complex fund flows into undifferentiated binary connections. When the system initiates risk transmission tracking, it assumes that risk states can instantly and synchronously transition between nodes, forcibly classifying fluctuations in the status of upstream and downstream accounts within the same fixed time window as valid risk transmission. When retrieving closed-loop fund flows, the system lacks an understanding of the inherent patterns in fund flows, relying solely on mechanical node identifier matching and strict numerical equivalence verification to determine path closure. This rigid matching logic ignores the transmission time lags commonly found in real business scenarios, such as approval cycles and bank clearing transit times, as well as the natural losses and tax depreciation effects of funds in multi-layered flows. Consequently, the system is highly prone to misjudging normal inter-period procurement settlements and periodic salary payments as high-risk closed-loop fund risks. To reduce this extremely high false alarm rate, the system had to introduce extremely stringent filtering thresholds. This not only caused a large number of real hidden return paths to be missed due to timing misalignment or amount deviation, but also caused conventional risk control systems to exhaust computing resources when dealing with high-concurrency transaction data due to getting bogged down in checking massive invalid closed loops. As a result, the system lost its core ability to accurately quantify and identify real fund closed loop risks in complex business scenarios.

[0022] To address the aforementioned issues, this application provides a comprehensive assessment and evaluation system for enterprise financial risk control. The system includes: a functional block partitioning module, used to perform functional partitioning processing on the target enterprise's fund transaction data based on the fund transfer relationships and fund usage between accounts, forming multiple fund functional blocks, and statistically analyzing the fund inflow and outflow between any two fund functional blocks to construct a set of directional flow relationships between fund functional blocks; a risk sequence generation module, used to extract the financial indicator sequence of each fund functional block from the fund transaction data, and input the financial indicator sequence into a pre-trained risk state transition model to perform risk evolution and obtain a risk generation sequence; and a transmission path generation module, used to temporally align the set of directional flow relationships with the risk generation sequence of the corresponding fund functional block, establishing a transmission path mapping between different fund functional blocks. The system establishes a risk transmission relationship and generates a set of risk transmission paths between fund functional blocks based on the transmission mapping relationship. A closed-loop identification module performs risk transmission processing along the risk transmission paths in the risk transmission path set based on the state change information of the risk generation sequence, and filters risk closed-loop paths based on the processing results. A contribution assessment module determines the path risk intensity identifier based on the path characteristics of the risk closed-loop path, determines the risk increment propagation value of each fund functional block in the corresponding risk closed-loop path based on the transmission mapping relationship, and generates a block risk contribution identifier for each fund functional block based on the correlation between the risk increment propagation value and the path risk intensity identifier of the corresponding risk closed-loop path. A result generation module constructs a risk heatmap of the fund transaction data based on the block risk contribution identifier and the path risk intensity identifier, generating a corporate financial risk assessment result.

[0023] Example 1, such as Figure 1 As shown, this application provides a comprehensive assessment and check-up system for enterprise financial risk control. The system includes functional block division modules, as detailed below: The functional block division module is used to perform functional division processing on the target company's fund transaction data based on the fund transfer relationship between accounts and the purpose of funds, forming multiple fund functional blocks, and statistically analyzing the fund inflow and outflow between any two fund functional blocks to construct a set of directional flow relationships between fund functional blocks.

[0024] In this embodiment, a brief introduction to the following related terms is given: Account: refers to a uniquely identified fund storage unit opened by the target enterprise in a financial institution or used for internal accounting, including bank accounts, virtual accounts, internal settlement accounts, etc.

[0025] Funds transfer relationship: refers to the two-way fund flow record between any two accounts within a certain time range, including funds flowing from account A to account B and funds flowing from account B to account A, used to characterize the closeness of fund interaction between accounts.

[0026] Purpose of Funds: This refers to the business description, summary, or classification label accompanying each fund transaction, used to determine what kind of economic activity the funds belong to (such as purchase payment, sales revenue, salary payment, tax payment, internal transfer, etc.).

[0027] Target company: The corporate entity to be subject to financial risk control assessment and health check. Its fund transaction data consists of the bank statements and internal accounting records of the company and its related parties, including the fund receipt and payment details of all accounts of the target company within a certain historical period. Each record must include at least the transaction account identifier, transaction time, transaction amount, transaction direction and transaction purpose information.

[0028] Funds Function Block: A collection of accounts with the same type of fund use and close fund-related relationships, representing the aggregation of a company's fund activities in a specific business function area (such as procurement settlement, sales collection, investment and financing management).

[0029] Fund inflow: The total amount of funds flowing into this fund function block from fund accounts in other fund function blocks.

[0030] Outflow of funds: The total amount of funds flowing from the fund account of this fund functional block to other fund functional blocks.

[0031] Targeted flow relationship set: A dataset that records the inflow and outflow of funds for each directed edge, with the fund function blocks as nodes and the directed fund flows between blocks as edges.

[0032] In one feasible approach, the functional block partitioning module is specifically used for: S101, extract the transaction account identifier, transaction time, transaction amount, transaction direction and transaction purpose information between transaction accounts from the capital transaction data of the target enterprise, and perform capital flow analysis on the transaction time, transaction amount and transaction direction based on the transaction account identifier to obtain the capital transaction relationship. The capital transaction relationship includes at least the number of capital round trips, the amount of capital round trips and the time interval between capital round trips. Among them, the transaction account identifier is a code or name used to uniquely identify the account from which funds enter or leave.

[0033] Transaction time: The actual date and specific time when the transaction is completed.

[0034] Transaction amount: The amount of funds transferred in a single transaction, measured in currency.

[0035] Transaction direction: Indicates whether funds are transferred out of this account (outflow) or into this account (inflow).

[0036] Transaction Purpose Information: A description of the business background of the fund transfer, such as contract number, invoice number, summary description, etc.

[0037] Fund transfer relationship: A feature vector used to characterize the bidirectional fund interaction between any two accounts within a statistical period. It includes at least the number of fund transfers, the amount of fund transfers, and the time interval between fund transfers. The number of fund transfers refers to the cumulative frequency of fund transactions between the two accounts in a specified time window. The amount of fund transfers refers to the total amount of bidirectional fund transactions between the two accounts within the specified time window. The time interval between fund transfers refers to the time difference between two transactions calculated for each fund transfer, and the average, minimum, or distribution is statistically analyzed.

[0038] The method for conducting fund flow analysis is as follows: all transaction records are grouped according to transaction account identifiers, and each group is sorted by transaction time to identify the transaction sequence between each pair of accounts. Then, based on the alternating changes in transaction direction, adjacent outflows and inflows are paired to form round-trip events, and the number of round-trip events, the amount of round-trip events, and the time interval between round-trip events are counted.

[0039] S102, For any two trading accounts, determine the degree of correlation between the funds flow between the trading accounts based on the number of fund flows, the amount of fund flows, and the time interval between fund flows. Among them, the degree of correlation between fund transfers is a comprehensive indicator used to quantify the closeness and regularity of fund transfers between two accounts. The higher the degree of correlation, the more the two accounts tend to work together functionally.

[0040] One way to determine the degree of correlation between fund transfers between trading accounts is as follows: For any two trading accounts, count the number of fund transfers, the total amount of transfers, and the average time interval between transfers within the statistical period. Divide the number of transfers and the total amount of transfers by the maximum number of transfers and the maximum amount among all account pairs, respectively, to obtain normalized frequency and amount values. Take the reciprocal of the average time interval and divide it by the maximum reciprocal among all account pairs to obtain a normalized time-tightness value. Then, sum the above three normalized values ​​according to a set weight (where the weight ratio can be adaptively configured according to the fund flow characteristics of the enterprise's business type to reflect the emphasis on frequency, amount, or timeliness in different business scenarios). The resulting value is the degree of correlation between fund transfers between the account pairs. The larger the value, the more frequent the fund transfers between the two accounts, the larger the cumulative amount, and the shorter the interval.

[0041] S103, based on the transaction purpose information corresponding to each transaction account, classify the fund purpose types of the transaction accounts, and count the fund inflow and outflow of each transaction account under different fund purpose types; The types of fund usage are categorized into several major categories based on keywords, summary patterns, or business rules in the transaction usage information, such as: procurement payments, sales receipts, salaries and benefits, taxes and fees, internal transfers, and investment and financing.

[0042] S104. Based on the degree of correlation between fund transfers between transaction accounts and the type of fund usage of the corresponding transaction accounts, transaction accounts with the same type of fund usage and whose degree of correlation between fund transfers meets the preset conditions are classified and processed to form multiple fund function blocks. The preset conditions refer to the criteria used to determine whether a group of transaction accounts with the same type of fund usage should be classified into the same fund function block. The preset conditions include two sub-conditions: first, the degree of fund transfer correlation between any two accounts must be greater than or equal to a pre-set correlation threshold; second, the number of direct transactions between the two accounts within the statistical period must be at least one. The correlation threshold is set as follows: within the set of accounts of the same fund usage type, all account pairs within the set are sorted in descending order of their fund transfer correlation, and the lower limit of the interval in the top preset percentile (e.g., the top 60%) is selected as the correlation threshold. This percentile value can be dynamically adjusted according to the company size and the number of accounts to ensure that occasional transactions are filtered out and accounts with stable business relationships are retained. Furthermore, the number of direct transactions between two accounts within the statistical period must be at least one to exclude accounts that only have indirect connections but no actual direct fund transfers.

[0043] One method for classifying trading accounts with the same type of fund usage and a fund transfer correlation that meets preset conditions is as follows: Group all trading accounts according to fund usage type, and put accounts within the same usage type into the same candidate set; then, for each candidate set, take out all accounts in the set and the fund transfer correlation between each pair of these accounts. When the correlation between two accounts reaches or exceeds the correlation threshold in the preset conditions, mark the two accounts as a mergeable relationship, and construct an undirected connected graph with accounts as nodes and mergeable relationships as edges; in this connected graph, if there is a path that connects several accounts directly or indirectly, then these accounts belong to the same connected component, and all accounts in each connected component are assigned to the same fund function block; if the correlation between an account and any other account in the candidate set is lower than the threshold, then the account forms a separate fund function block.

[0044] S105: Read the corresponding fund transaction data between transaction accounts belonging to different fund function blocks, and determine the directional flow relationship between fund function blocks based on the outflow and inflow directions of funds in the fund transaction data. The outflow of funds refers to the transfer of funds from an account within a certain fund function block to an account within another fund function block; the inflow of funds refers to the transfer of funds from an account outside a certain fund function block to an account within the same fund function block.

[0045] Directed flow relationship: Identifies the flow channel of funds from the source fund functional area to the target fund functional area, reflecting the transfer path of enterprise funds between different functional groups.

[0046] Specifically, it iterates through all cross-block inter-account transaction records. For each transaction, it identifies the fund function block to which the fund transfer account belongs as the source block and the fund function block to which the fund transfer account belongs as the target block, thereby determining a directed flow relationship from the source block to the target block.

[0047] S106: Calculate the inflow and outflow of funds corresponding to each directional flow relationship, and generate a set of directional flow relationships between fund functional blocks.

[0048] The functional block partitioning module of this application categorizes all transaction accounts of the target enterprise into multiple fund function blocks according to business functions, and counts the fund inflows and outflows between blocks to construct a set of directional flow relationships between blocks. This reduces the number of nodes that need to be analyzed independently in the original fund transaction data from the actual number of accounts to the number of fund function blocks, thereby effectively compressing the graph structure scale faced by subsequent transmission path generation, closed-loop identification and other modules. Correspondingly, the computational complexity of traversing node pairs in the graph structure to find fund flow relationships is reduced from quadratic to quadratic of the number of accounts. Since the number of blocks is much smaller than the original number of accounts, the system can undertake a deeper and wider search for risk transmission paths under the same hardware resources without exhaustively scanning all account pairs. Furthermore, the functional block partitioning module retains the clustering characteristics of funds at the business function level, summarizing the originally scattered transactions among multiple accounts into targeted fund flows with business meaning between blocks. This allows subsequent modules to directly perform risk state evolution and transmission analysis based on the block level without repeatedly parsing the original account-level flow. Thus, the functional block partitioning module reduces computational and memory overhead while providing a feasible data foundation for mining deep multi-hop fund loops.

[0049] Example 2: This application provides a comprehensive assessment and check-up system for enterprise financial risk control. The system includes a risk sequence generation module, as detailed below: The risk sequence generation module is used to extract the financial indicator sequence of each funding functional block from the funding transaction data, and input the financial indicator sequence into the pre-trained risk state transition model to perform risk evolution and obtain the risk generation sequence.

[0050] In this embodiment, a brief introduction to the following related terms is given: Financial indicator sequence: refers to a set of multi-dimensional financial indicator values ​​arranged in chronological order for a specific funding functional area. Each statistical period corresponds to a multi-dimensional value group consisting of net cash flow, cash turnover rate, and short-term debt coverage ratio. The value groups of each statistical period are combined in chronological order to form a time series structure reflecting the dynamic changes in the financial status of the funding functional area within the historical statistical period.

[0051] The risk state transition model refers to a sequence modeling network that takes a time-series financial indicator sequence as input and a risk state level sequence as output. It employs a three-stage structure of feature extraction, hidden state recursion, and risk classification. This model captures the evolution of the financial state of financial functional blocks over continuous time steps and maps it to discrete risk state labels. Before deployment, the model undergoes supervised training using historical financial data. The training samples are historical financial indicator sequences of enterprises with labeled risk levels, and the supervision labels are manually labeled risk levels (such as normal, warning, high risk, etc.) at corresponding time steps. Through iterative optimization of the network parameters, the classification error between the model's output risk level and the labeled results is continuously reduced until the model's risk state classification accuracy on the validation set converges to above a preset threshold, completing the pre-training process.

[0052] Risk generation sequence: refers to the structured time series record output by the risk state transition model after performing inference on the financial indicator sequence of a certain funding functional block. Each record contains a time step number, the risk state label of the corresponding time step, and the state change value between adjacent time steps, reflecting the current status of the risk level of the funding functional block and its dynamic transition in each statistical period.

[0053] In one feasible implementation, the risk sequence generation module is specifically used for: S201, From the fund transaction data, according to the division results of each fund functional block, extract the net flow of funds, the turnover rate of funds and the short-term debt coverage ratio of each fund functional block in each statistical period as financial indicators, and arrange the financial indicators in the order of time steps to form a financial indicator sequence for the corresponding fund functional block. Among them, financial indicators refer to quantitative indicators reflecting the health of the capital operation of a capital function block within a certain statistical period. These include at least net cash flow, capital turnover rate, and short-term debt coverage ratio. Net cash flow refers to the difference between the total cash inflow and the total cash outflow of the capital function block within a certain statistical period. A positive value indicates net cash inflow, and a negative value indicates net cash outflow. It is used to indicate the profit or loss status of the capital function block in the current period. Capital turnover rate refers to the ratio of the total cash outflow of the capital function block to the average account balance of the block within a certain statistical period. It reflects the speed of capital circulation within the capital function block. The higher the value, the stronger the capital activity. Short-term debt coverage ratio refers to the ratio of the available cash balance of the account within the capital function block to the corresponding short-term debt amount at the end of a certain statistical period. It is used to indicate the ability of the capital function block to repay short-term debts with existing working capital. When the value is less than 1, it indicates that there is short-term capital flow pressure.

[0054] Financial indicator sequence: refers to a set of multi-dimensional financial indicator data organized continuously according to the statistical period for a single capital function block, used to characterize the changes in the capital operation status of the capital function block within the statistical period; each time step in the financial indicator sequence corresponds to a set of financial indicator values, which includes at least net cash flow, capital turnover rate, and short-term debt coverage ratio. The sets of financial indicator values ​​corresponding to each time step are arranged in chronological order to form a time-series structure with a progressive relationship, used to reflect the evolution of the capital liquidity, capital activity level, and short-term solvency of the capital function block within the continuous statistical period.

[0055] S202, the financial indicator sequence is input into a pre-trained risk state transition model, the risk state transition model including a time-series feature extraction layer, a hidden state transition layer and a risk state output layer; The time-series feature extraction layer performs a linear transformation on the financial indicator sequence at each time step to generate a local financial feature vector for the corresponding time step. The hidden state transition layer receives the local financial feature vector at the current time step and performs gated fusion calculation with the hidden state vector at the previous time step to generate the hidden state vector at the current time step. The risk state output layer performs mapping calculation on the hidden state vector at each time step, outputs the risk state probability distribution at the corresponding time step, and determines the risk state level corresponding to the maximum probability as the risk state label for the time step. It should be noted that the local financial feature vector refers to the intermediate representation vector obtained by the temporal feature extraction layer after performing a linear transformation on the input numerical group at a single time step. It is used to perform feature space mapping on the original financial indicators before entering the hidden state recursion, thereby improving the model's ability to express different combination patterns of indicators.

[0056] One approach to performing linear transformation is as follows: The temporal feature extraction layer maintains a set of trainable weight parameters and bias parameters. For the three-dimensional financial indicator numerical set based on financial indicator sequence analysis input at the current time step, the weight parameters and the input numerical set are weighted and summed, and then the bias parameters are added to obtain the local financial feature vector for the corresponding time step. These weight parameters and bias parameters are learned and updated through backpropagation during the pre-training phase, ensuring that the transformed feature vector effectively carries the combined information related to risk status judgment in the original financial indicators.

[0057] Hidden state vector: refers to the recursive state vector output by the hidden state transition layer at each time step. It has time memory and contains both the local financial feature information of the current time step and the historical state information accumulated in previous time steps. It is the core input for the risk state output layer to classify risk levels.

[0058] One approach to gated fusion computation involves a gated recursive unit structure in the hidden state transition layer. Two independent gating mechanisms adjust the fusion method between historical hidden state information and the current local financial feature vector. Specifically, the update gate determines a fixed proportion of information from the previous hidden state vector to be retained in the current time step output; the reset gate controls the proportion of the previous hidden state vector involved in the computation when generating candidate hidden states, enabling the model to selectively forget historical states from the current input. Both gating signals are obtained by concatenating the previous hidden state vector and the current local financial feature vector, followed by parameter transformation and activation function compression to the zero-to-one interval. Candidate hidden states are obtained by concatenating the weighted previous hidden state and the current local financial feature vector through the reset gate, followed by parameter transformation and nonlinear activation. The hidden state vector at the current time step is formed by weighting the previous hidden state vector and the candidate hidden state proportionally according to the update gate signal.

[0059] Risk state probability distribution: refers to the multi-dimensional probability vector obtained by the risk state output layer after performing mapping calculation on the hidden state vector of the current time step. Its dimension is consistent with the number of preset risk state levels (for example, if four levels are set: normal, mild warning, moderate warning, and high risk, then the probability vector is four-dimensional). The value of each dimension in the vector represents the probability that the time step belongs to the corresponding risk level, and the sum of the values ​​of each dimension is 1.

[0060] Risk status label: refers to the level number corresponding to the maximum probability value in the risk status probability distribution, represented in integer form. The larger the value, the higher the risk level of the corresponding time step.

[0061] The mapping calculation to output the risk state probability distribution for the corresponding time step is performed as follows: The risk state output layer maintains a set of trainable weight parameters and bias parameters. For the hidden state vector of the current time step, the weight parameters and the hidden state vector are weighted and summed, and then the bias parameters are added to obtain an original score vector with the same dimensions as the number of risk levels. Normalization is performed on this score vector, converting the original scores of each dimension into probability values ​​for the corresponding levels, where the sum of all level probability values ​​is 1. The result is the risk state probability distribution for the current time step. During the pre-training phase, manually labeled risk levels at each time step are used as supervision signals. The weight parameters and bias parameters are iteratively optimized by minimizing the classification error between the model's output probability distribution and the labeled levels.

[0062] S203, calculate the risk status label corresponding to each time step step by step based on the time step sequence, take the difference of risk status label between adjacent time steps as the status change value, and encapsulate the status change value and risk status label into status change information; Among them, the state change value refers to the difference between the risk state labels of two adjacent time steps. A positive value indicates that the risk level has increased, that is, the risk has intensified. A negative value indicates that the risk level has decreased, that is, the risk has eased. A zero value indicates that the risk state remains unchanged at that time step.

[0063] Status change information: refers to a structured record unit that combines the risk status label of a certain time step with the corresponding status change value. It uses fixed fields to represent the current status of the risk level at that time step and its direction and magnitude of change relative to the previous time step.

[0064] The encapsulation is as follows: For each time step, the risk status label and status change value are organized into a structure according to the fixed order of the risk level field and the status change field. The field content corresponds to the risk level integer number of the corresponding time step and the difference between it and the risk level number of the previous time step, thus forming a standardized record unit that can be directly indexed by field name.

[0065] S204, combine the state change information with the corresponding time step and record it to generate a risk generation sequence for the corresponding funding function block.

[0066] The method of combining records is as follows: For a certain funding functional block, all statistical periods are traversed in ascending order of time steps. For each time step, the time step number and the corresponding status change information are merged into a three-field record containing the time step number, risk level field and status change field. This record is then appended to the ordered list structure in ascending order of time steps to form an ordered record sequence with the same length as the total number of statistical periods. This is the risk generation sequence for the funding functional block.

[0067] The risk sequence generation module of this application compresses the financial indicator sequences of each funding functional block into a lightweight structure sequence containing only discrete risk state labels and state change values ​​through risk state transition model inference. The data scale of each time step is reduced from multi-dimensional continuous numerical values ​​to two integer fields, significantly reducing the total amount of data that other modules need to reside in memory when performing joint analysis of multiple blocks. The gated recursive structure ensures that the hidden state vector is updated in a fixed dimension. After inference, the system does not need to retain historical original indicator windows for transmission analysis, effectively controlling the memory usage that increases with the number of blocks and the time span. The state change values ​​are pre-calculated and fixed in the risk generation sequence during the encapsulation stage of the risk sequence generation module, providing a directly readable transition signal for the transmission path generation module when performing time sequence alignment. This avoids redundant I / O operations caused by repeated scanning of original data and repeated differential calculations across modules, thus effectively reducing the overall computational resource consumption and data read / write overhead in multi-block long-cycle batch processing scenarios.

[0068] Example 3, as Figure 1 As shown, this application provides a comprehensive assessment and check-up system for enterprise financial risk control. The system includes a transmission path generation module, as detailed below: The transmission path generation module is used to align the set of directional flow relationships with the risk generation sequence of the corresponding capital functional blocks in time, establish the transmission mapping relationship of the risk generation sequence between different capital functional blocks, and generate a set of risk transmission paths between capital functional blocks according to the transmission mapping relationship.

[0069] In this embodiment, a brief introduction to the following related terms is given: Transmission mapping relationship: This refers to the structured association record established after temporal alignment and transition directionality verification for a directed edge in the directed graph of the capital functional blocks. It includes the risk status labels of the upstream and downstream capital functional blocks connected by the directed edge, the corresponding risk transmission time lag values, and the criteria for marking the directed edge as a valid transmission edge. The core function of the transmission mapping relationship is to transform the directed flow relationship at the capital level into a description of an effective transmission channel at the risk level, providing time lag parameters and direction verification results for the closed-loop identification module to perform risk transmission calculations along the path.

[0070] The risk transmission path set refers to the aggregated set of all valid transmission paths selected from the directed graph of the capital function blocks based on the transmission mapping relationship. Each risk transmission path in the set consists of several sequentially connected valid transmission edges, reflecting the complete channel through which risk status signals propagate step by step along the direction of capital flow between multiple capital function blocks. The risk transmission path set provides the loop closure identification module with the traversal starting point and path range, ensuring that the loop closure search is performed within the verified valid transmission structure, rather than exhaustively searching through all directed edges.

[0071] In one feasible implementation, the transmission path generation module is specifically used for: S301, construct a directed graph of fund function blocks using each fund function block as a node and the directed flow relationship in the set of directed flow relationships as edges; The directed graph of fund function blocks refers to a graph structure with all fund function blocks as vertices and directed edges formed by statistically confirmed directed fund flow relationships between blocks. The direction of each directed edge indicates the direction of fund flow from the source block to the target block, and the edges are accompanied by corresponding statistical values ​​of fund inflow and outflow. The purpose of constructing this directed graph is to abstract the dispersed account-level fund transaction relationships into a block-level topology, providing a unified structural foundation for performing risk transmission time series analysis on the graph structure, while compressing the graph node size to the order of fund function blocks, reducing the computational complexity of subsequent path traversal.

[0072] S302, for each directed edge in the directed graph of the fund function block, determine the upstream fund function block and the downstream fund function block of the directed edge, and perform feature extraction based on the state change information of the risk generation sequence of the upstream fund function block and the downstream fund function block to obtain the risk state label and the first time step of the risk state label transition of the upstream fund function block, and the risk state label and the second time step of the risk state label transition of the downstream fund function block. In this context, a directed edge refers to a directed arc connecting two capital function blocks in a directed graph. The starting point of the arc is the capital function block in the direction of capital outflow, called the upstream capital function block, and the ending point is the capital function block in the direction of capital inflow, called the downstream capital function block. The direction of the directed edge is consistent with the direction of capital flow, used to characterize the potential directionality of risk signals propagating from upstream to downstream along the capital flow channel.

[0073] First time step: refers to the set of time step numbers in the risk generation sequence of the upstream capital function block where the risk status label changes relative to the previous time step (i.e., the status change value is not zero). Each time step number that meets the conditions constitutes an element in the first time step set, representing the moment when a risk status transition occurs in the upstream block.

[0074] The second time step refers to the set of time step numbers in the risk generation sequence of the downstream capital function block where the risk status label changes relative to the previous time step. Its meaning corresponds to the first time step, representing the moment when the risk status transition occurs in the downstream block.

[0075] One method for feature extraction is as follows: For the upstream funding functional block corresponding to the current directed edge, read the state change value field in its risk generation sequence step by step, and record the time step numbers where the state change value is not zero in sequence to form a list of transition time steps for the upstream block. At the same time, retain the risk state label and transition direction (i.e., the positive or negative sign of the state change value) corresponding to each transition time step. Perform the same operation for the downstream funding functional block to form a list of transition time steps for the downstream block and record the corresponding risk state label and transition direction. The above operation directly reads the pre-encapsulated state change value field in the risk generation sequence without rescanning the original financial indicator data.

[0076] S303, based on the first time step and the second time step, calculate the time step spacing value between each upstream capital function block and the downstream capital function block, and take the time step spacing value with the highest frequency as the risk transmission delay value of the directed edge. The time step spacing value refers to the difference obtained by subtracting the first time step number of the upstream block from each second time step number of the downstream block for a given pair of upstream and downstream funding functional blocks. A positive time step spacing value indicates that the downstream block's transition occurs after the upstream block's transition, conforming to the time logic of risk propagation from upstream to downstream. A time step spacing value of zero or negative indicates that the downstream transition does not lag behind the upstream transition, which does not conform to the expected propagation sequence. In statistical analysis, the difference is calculated for each pair of upstream and downstream transition time steps to form a frequency distribution of the time step spacing values. The highest frequency positive spacing value is taken as the risk transmission lag value for that directed edge.

[0077] Risk transmission lag: This refers to the typical number of time steps required for a risk state transition signal to propagate from the upstream block to the downstream block within a directed edge. It is represented by the time step interval with the highest frequency of occurrence. The risk transmission lag reflects the actual transmission cycle of capital flows along this channel, providing a quantitative basis for time series alignment and closed-loop time estimation.

[0078] S304, based on the risk transmission delay value, perform time shift processing on the transition time steps of each risk status label of the downstream capital function block, and subtract the risk transmission delay value from the second time step to obtain the corresponding alignment time step; Transition time step: refers to the time step in the risk generation sequence where the state change value is not zero, that is, the specific time step number where the risk state label changes relative to the previous time step. Transition time steps are a subset of time steps; not all time steps are transition time steps. Only when the state change value field of the corresponding time step is not zero is the time step marked as a transition time step and included in the time series alignment calculation.

[0079] The time-shifting process is as follows: For each second time step number of the downstream funding function block, the risk propagation delay value corresponding to the current directed edge is subtracted from the number to obtain a new time step number after time shifting. The meaning of this shifting operation is to trace back the time step corresponding to the risk propagation delay value from the actual time of the downstream block's transition, restoring the equivalent upstream trigger time after removing the propagation delay effect, so as to compare it with the first time step of the upstream block on the same time axis.

[0080] Alignment time step: This refers to the time step number obtained after performing time shift processing on the second time step of the downstream funding functional block. It represents the equivalent trigger time on the time axis corresponding to a certain jump event in the downstream block after eliminating the impact of risk transmission lag. The numerical difference between the alignment time step and the second time step is always equal to the risk transmission lag value, and it serves as the benchmark time coordinate for matching the upstream and downstream jump times.

[0081] S305, Match the alignment time step with the first time step of the upstream funding function block in chronological order to determine the alignment time step corresponding to the first time step and form a corresponding time step pair; A corresponding time step pair is a binary record formed by pairing a first time step of an upstream funding functional block with an aligned time step of a downstream funding functional block that is numerically closest and chronologically ordered after time shifting. The corresponding time step pair is formed by arranging all aligned time steps and all first time steps according to their numerical values ​​and matching them sequentially. When the value of an aligned time step is equal to or closest to the value of a first time step within the allowable error range, they are identified as a corresponding time step pair. Each first time step corresponds to at most one aligned time step. Corresponding time step pairs are the basis for comparing the directionality of transitions. The allowable error range reflects situations where upstream and downstream transition times are difficult to align precisely due to statistical granularity or occasional delays in actual business operations. It can be dynamically adjusted based on the confidence interval width of the risk transmission lag value (e.g., taking a multiple of the standard deviation of the frequency distribution corresponding to the lag value).

[0082] S306, For each corresponding time step pair, extract the transition direction of the upstream capital function block at the first time step corresponding to the risk status label, and the transition direction of the downstream capital function block at the alignment time step corresponding to the risk status label. The transition direction refers to the direction of the change in the risk status label at a given transition time step. It is determined by the sign of the corresponding state change value: a positive state change indicates an upward transition, meaning the risk level increases at that time step; a negative state change indicates a downward transition, meaning the risk level decreases at that time step. The transition direction only considers either upward or downward values, without considering the specific magnitude of the state change value, to ensure that the comparison of transition directions between upstream and downstream blocks is not affected by differences in the absolute values ​​of risk levels across blocks.

[0083] S307, count the number of time steps with the same transition direction, and calculate the proportion of the number to the total number of risk state label transition time steps of the upstream capital function block, and determine the proportion as the risk state transition same direction rate of the corresponding directed edge. Among them, the risk state transition direction consistency rate refers to the proportion of time step pairs corresponding to a certain directed edge in which the upstream and downstream blocks have the same transition direction, out of the total number of transition time steps of the upstream block. The higher the proportion, the stronger the temporal consistency between the risk state transition of the downstream block and that of the upstream block, and the higher the credibility of the directed edge as a risk transmission channel; the lower the proportion, the weaker the correlation between the transition directions of the two blocks, and the more likely the directed edge reflects a financial transaction relationship rather than a risk transmission relationship.

[0084] S308, mark the directed edges whose risk state transition homing rate exceeds the preset threshold as valid transmission edges, and encapsulate the valid transmission edges and the corresponding risk transmission delay values ​​into a transmission mapping relationship; The preset threshold refers to the lower limit of the risk state transition unidirectional rate used to determine whether a directed edge constitutes an effective risk transmission channel. The preset method is as follows: after calculating the risk state transition unidirectional rate for all directed edges, the unidirectional rate values ​​of all directed edges are sorted in ascending order, and the lower limit of the interval located in the preset percentile (e.g., the top 60%) is taken as the preset threshold. This percentile value can be adaptively adjusted based on the number of capital function blocks of the target enterprise and the historical risk transmission frequency.

[0085] Effective transmission edge: refers to a directed edge whose risk state transition homing rate exceeds a preset threshold. This indicates that there is a statistically credible pattern of synchronous risk state transition between the upstream and downstream capital functional blocks connected by the directed edge, and the directed edge is confirmed as an effective transmission channel for risk signals.

[0086] S309, Based on the transmission mapping relationship, identify effective transmission paths from the directed graph of the capital function block as risk transmission paths, and summarize the risk transmission paths to obtain a risk transmission path set.

[0087] The risk transmission path refers to a directed channel in the directed graph of the capital function block, which is formed by connecting at least one effective transmission edge in sequence. Each edge in the path has been verified by the risk state transition unidirectional rate and is accompanied by a corresponding risk transmission time delay value.

[0088] The risk transmission path is identified as follows: taking all valid transmission edges as passable edges, traversing each funding function block and taking it as the starting point of the path; starting from the starting point, recursively visiting downstream blocks along the direction of valid transmission edges, incorporating each edge into the current path, until the current block has no valid transmission edges pointing to other blocks; recording the edges traversed in this traversal in the order of access as a path sequence.

[0089] The transmission path generation module of this application aligns the set of directional flow relationships with the risk generation sequence in a temporal manner, and uses the risk state transition simultaneity rate as the screening criterion for effective transmission channels. This filters out directed edges in the original capital flow topology that only reflect capital flow relationships but do not have risk transmission patterns, narrowing the traversal range of the closed-loop identification module to the statistically verified effective transmission structure, eliminating the need for exhaustive search on all directed edges. The introduction of risk transmission delay values ​​quantifies and solidifies the transmission delay of each directed edge in the transmission mapping relationship. Subsequent modules can directly accumulate the delay values ​​of each edge when performing multi-hop transmission time estimation, without repeating cross-block temporal alignment calculations. This effectively reduces the path search calculation overhead of the system in complex multi-block topology scenarios while ensuring the accuracy of risk transmission path identification.

[0090] Example 4, as Figure 1 As shown, this application provides a comprehensive assessment and check-up system for enterprise financial risk control. The system includes a closed-loop identification module, as detailed below: The closed-loop identification module is used to perform risk transmission processing along the risk transmission paths in the risk transmission path set according to the state change information of the risk generation sequence, and to filter the risk closed-loop path based on the processing result.

[0091] In this embodiment, a brief introduction to the following related terms is given: Among them, the risk closed-loop path refers to a directed circular path in the set of risk transmission paths where the starting node and the ending node are the same funding functional block, and after sequential transmission along each directed edge of the path, the risk state transition direction at the ending node at the target time step is consistent with the risk state transition direction at the starting node at the activation timestamp. The existence of a risk closed-loop path indicates that a risk state change occurring in a certain funding functional block, after being transmitted step by step through several intermediate funding functional blocks, flows back to the block itself in the form of a unidirectional transition, forming a risk closed loop. It is an important indicator of the accumulation of potential risks in a company's capital structure.

[0092] The risk transmission process is as follows: Starting with the risk-activated block, the risk state transition direction at the activation timestamp is read, and the process proceeds downstream level by level along each risk transmission path starting from this block in the risk transmission path set. In each level of advancement, the target time step of the current downstream block is calculated by accumulating the risk transmission delay value corresponding to the current directed edge based on the activation timestamp. At the target time step, the risk state change value of the downstream block is read, and its transition direction is extracted and compared with the transition direction of the starting node. The above operations are performed sequentially on each level of downstream blocks in the path. If all downstream blocks undergo a transition in the same direction as the starting node at the corresponding target time step, the path is marked as a candidate risk closed-loop path. Then, it is determined whether the termination node of the candidate path is the same funding function block as the starting node. Those that meet the conditions are determined as risk closed-loop paths.

[0093] In one feasible approach, such as Figure 2 As shown, the closed-loop identification module is specifically used for: S401, read the risk state label step by step from the state change information of the risk generation sequence, and compare the changes of risk state label between adjacent time steps step by step to generate a risk state change sequence. Among them, the risk status change sequence refers to the set of status change value records arranged in time step order for a certain funding functional block. Each time step corresponds to a status change value, which is used to reflect the gradual change of the risk level of the block in the entire statistical period. It is a direct data source for identifying the transition time and transition direction.

[0094] The step-by-step comparison method is as follows: the risk status label field of each time step in the risk generation sequence of the target funding function block is read sequentially, and the difference between the risk status label of each time step and the risk status label of the immediately preceding time step is calculated. The difference is recorded as the status change value of that time step in the risk status change sequence.

[0095] It should be noted that the risk generation sequence has pre-encapsulated the state change value field in the risk sequence generation module. This step directly reads this field to complete the construction of the risk state change sequence, without having to repeatedly perform the difference calculation, thus avoiding redundant calculation overhead across modules.

[0096] S402, based on the risk state change sequence, identify the time step of the risk state transition, determine the corresponding capital function block that has transitioned as the risk activation block, and determine the time step of the first transition as the activation timestamp. Among them, the risk activation block refers to a funding function block in the risk state change sequence that has at least one time step with a non-zero state change value, i.e., a block that has experienced a change in risk level within the statistical period. The risk activation block is the starting search unit for closed-loop identification. It should be noted that only funding function blocks whose own risk state has actually transitioned are qualified as the starting node of the risk closed-loop path; if the risk state of a block has not changed throughout the entire statistical period, it does not have the factual basis to transmit risk transition signals downstream and is not included in the scope of the closed-loop search starting point.

[0097] Activation timestamp: This refers to the time step number when a risk-activated block first exhibits a non-zero state change value in the risk state change sequence, representing the moment when the block's risk state first transitions. The activation timestamp plays at least two roles in closed-loop identification: first, as the starting point for calculating the target time step of each downstream block along the path; and second, as a directional benchmark for comparing the consistency of transition directions, used to compare the transition directions of each downstream block with the transition direction at the activation timestamp.

[0098] S403, taking the risk activation block as the starting node, traversing each risk transmission path in the risk transmission path set, and determining the subsequent funding function blocks level by level according to the directed edge order in the risk transmission path. Subsequent funding functional blocks refer to the downstream funding functional blocks located after the currently traversed node and arranged sequentially according to the direction of the directed edges in a certain risk transmission path. During the path-by-path traversal, the risk activation block is taken as the first-level node. The second-level node is determined along the first directed edge of the path, the third-level node is determined along the second directed edge, and so on, until all nodes in the path are determined in sequence. The determination of subsequent funding functional blocks strictly follows the arrangement order of the directed edges in the path, without skipping levels, to ensure that the accumulation order of the time lag values ​​at each level in subsequent time calculations is consistent with the actual direction of funding transmission.

[0099] S404. Based on the risk transmission time delay value corresponding to each directed edge, the time step of the risk state transition of the subsequent capital function block is calculated to obtain the corresponding target time step. The target time step refers to the expected transition time calculated by progressively accumulating the risk transmission lag value from the activation timestamp. It represents the time step number at which the transition signal from the risk-activated block at the activation timestamp, after being transmitted through the path, is expected to reach the current downstream funding function block and trigger a unidirectional transition. The target time step is the time coordinate for reading and comparing the transition direction within the risk generation sequence.

[0100] The time calculation method is as follows: using the activation timestamp as the base time step, the process proceeds downstream from the starting node. For each directed edge in the path, the current accumulated time step is added to the risk transmission delay value corresponding to that directed edge to obtain the target time step for the next downstream funding function block. This accumulation operation is performed sequentially on all directed edges in the path until the target time step for each subsequent funding function block in the path is obtained. Specifically, the target time step of the downstream block of the first directed edge in the path is equal to the activation timestamp plus the risk transmission delay value of the first directed edge; the target time step of the downstream block of the second directed edge in the path is equal to the first-level target time step plus the risk transmission delay value of the second directed edge, and so on. The accumulation process proceeds unidirectionally along the directed edges of the path.

[0101] S405, at the target time step, read the risk status label of the corresponding fund function block and compare it with the risk status label at the activation timestamp to determine whether a risk status transition in the same direction has occurred, and mark the risk transmission path corresponding to the risk status transition in the same direction as a candidate risk closed loop path. The method for determining whether a consistent risk state transition has occurred is as follows: Read the state change value at the activation timestamp in the risk state change sequence of the risk activation block, and extract its sign as the baseline transition direction. Positive values ​​correspond to an upward risk direction, and negative values ​​correspond to a downward risk direction. For each subsequent funding function block in the path, locate the corresponding target time step in its risk generation sequence, read the state change value at that time step, and extract its sign. Compare this sign with the baseline transition direction. If the signs are the same, it is determined that the block at that level has experienced a risk state transition in the same direction as the starting node at the target time step. If the signs are different or the state change value is zero, it is determined that they are inconsistent. Perform the above comparison level by level for all subsequent funding function blocks in the path. Only when all subsequent funding function blocks in the path pass the consistency judgment is the risk transmission path marked as a candidate risk closed-loop path.

[0102] Candidate risk closed-loop path: This refers to a risk transmission path in the set of risk transmission paths where, after performing the target time step calculation and transition direction consistency judgment step by step along the path, all subsequent funding function blocks in the path undergo a risk state transition in the same direction as the activation timestamp of the risk activation block at the corresponding target time step. A candidate risk closed-loop path indicates that the path meets the statistical consistency conditions of risk transmission in both temporal and directional dimensions, but it has not yet been confirmed whether the path termination node flows back to the starting block.

[0103] S406, when the termination node of the candidate risk closed loop path is the same funding function block as the risk activation block, the candidate risk closed loop path is determined as the risk closed loop path.

[0104] Termination Node: This refers to the last node in the directed edge sequence of the risk transmission path, i.e., the funding function block pointed to by the last directed edge in the path. In the verification of termination nodes in candidate risk closed-loop paths, the funding function block identifier of the termination node is compared with the identifier of the risk activation block. If the two identifiers match, it is confirmed that the risk transmission signal of the candidate path, after being transmitted step-by-step through intermediate blocks, eventually flows back to the starting block, forming a complete risk closed loop. The candidate risk closed-loop path is then upgraded and confirmed as a risk closed-loop path and included in the subsequent contribution assessment module. If the termination node and risk activation block identifiers do not match, the candidate path does not constitute a closed loop and is removed from the candidate set.

[0105] The closed-loop identification module of this application combines the temporal calculation of risk state transitions with directional consistency verification. It performs closed-loop search within the effective transmission path range filtered by the transmission path generation module, without exhaustively verifying all directed edge combinations. The activated timestamp mechanism anchors the time base of the closed-loop search to the moment of the first transition, enabling the target time step of each downstream block to be calculated by accumulating the transmission delay value step by step, without brute-force matching of all time step combinations. The transition direction consistency judgment replaces the comparison of absolute level values ​​with symbol comparison, avoiding the problem of incomparable absolute values ​​of risk levels caused by differences in business scale among different funding function blocks. Thus, while ensuring the accuracy of risk closed-loop path identification, it effectively reduces the computational overhead of the closed-loop search in complex multi-block topology scenarios.

[0106] Example 5, as Figure 1 and Figure 3 As shown, this application provides a comprehensive assessment and check-up system for enterprise financial risk control. The system includes a contribution assessment module, as detailed below: The contribution assessment module is used to determine the path risk intensity identifier based on the path characteristics of the risk closed loop path, determine the risk increment propagation value of each funding functional block in the corresponding risk closed loop path based on the transmission mapping relationship, and generate the block risk contribution identifier of each funding functional block according to the correlation between the risk increment propagation value and the path risk intensity identifier of the corresponding risk closed loop path.

[0107] In this embodiment, a brief introduction to the following related terms is given: Path characteristics: refers to the quantifiable descriptive information of the risk closed-loop path in two dimensions: structure and behavior. This includes the number of funding functional blocks in the path, the time span of closed-loop transmission, the sequence of risk status changes between adjacent blocks in the path, and the proportional relationship between the amount of funds flowing back and the amount of funds flowing out initially.

[0108] Path risk intensity label: refers to the discrete risk intensity level label assigned to the path based on the path characteristics of the risk closed loop path and after mapping by the preset risk propagation association rules. It is used to characterize the overall risk propagation energy level of the closed loop path.

[0109] Risk increment propagation value: This refers to the incremental contribution of a certain funding functional block to the risk status within its corresponding risk closed-loop path, after time delay correction and funding flow offset correction. It reflects the actual degree to which the block amplifies or attenuates the upstream risk signal during its propagation downstream. The larger the risk increment propagation value, the stronger the amplification effect of the block on the propagation of risk signals within the closed-loop path.

[0110] Block Risk Contribution Identifier: This refers to a discrete level label generated based on the ranking of the risk contribution values ​​of each funding function block within the same risk closed-loop path. It is used to characterize the degree of risk contribution of the block relative to other blocks in the corresponding closed-loop path.

[0111] In one feasible approach, the contribution assessment module is specifically used for: S501: Number the positions of each capital function block in the risk closed-loop path according to the transmission order to obtain the path block set; The location number refers to an integer sequence number assigned sequentially to each funding functional block along the directed edges in the risk closed-loop path, starting from the initial node. The initial node is numbered first, and the number increments by one for each directed edge traversed along the path. The number of the path termination node (i.e., the node returning to the initial block) is the largest sequence number corresponding to the total number of blocks in the path. The purpose of the location number is to provide an ordered index for subsequently reading the state change values ​​of adjacent blocks in the transmission order, ensuring that the calculation of state change quantities strictly follows the directional order of risk transmission.

[0112] Path Block Set: An ordered set formed by arranging all funding functional blocks in the risk closed-loop path from smallest to largest according to their position numbers. Each element in the set contains the identifier of the funding functional block and its corresponding position number.

[0113] S502, based on the path block set, extract the state change values ​​corresponding to adjacent funding function blocks from the state change information of the risk generation sequence, and calculate the state change amount between adjacent funding function blocks to obtain the path state change sequence. Among them, the state change amount refers to the difference in state change values ​​between two adjacent funding function blocks in the path block set at their respective activation transition times. Specifically, it is the difference between the state change value corresponding to the later position number block and the state change value corresponding to the earlier position number block. It is used to measure the amplitude change of the risk signal when it propagates between two adjacent blocks. A positive value indicates that the risk signal is amplified during propagation, and a negative value indicates that it is attenuated.

[0114] The method for calculating the state change between adjacent funding functional blocks is as follows: according to the position number order in the path block set, the identifiers of two adjacent funding functional blocks are extracted in sequence, and the corresponding transition time step is located in their respective risk generation sequence. The state change value field of that time step is read. The state change value of the next block is subtracted from the state change value of the previous block to obtain the state change amount of the adjacent block pair. The above operation is performed sequentially on all adjacent block pairs in the path block set to form a state change amount sequence with a length of one less than the total number of path blocks.

[0115] Path state change sequence: refers to the ordered numerical sequence formed by arranging the state changes of all adjacent block pairs in the path block set according to their position numbers, recording the amplitude changes of risk signals between adjacent blocks as they propagate along the risk closed-loop path.

[0116] S503, based on the path block set, count the number of fund function blocks from the initial fund function block to the return to the initial fund function block, and combine the timestamp difference between the first and last fund function blocks in the path block set to determine the closed-loop transmission frequency value. The initial funding functional block refers to the funding functional block with the first position number in the path block set, which is the starting node of the risk closed-loop path and corresponds to the risk activation block determined in the closed-loop identification module. The initial funding functional block is also the target block for the return flow from the path termination node, and its risk state transition at the activation timestamp is the time reference starting point for the risk transmission analysis of the entire closed-loop path.

[0117] Timestamp difference: This refers to the difference between the target time step of the funding function block at the terminating node in the path block set and the activation timestamp of the initial funding function block. It reflects the total time step span from the starting block, through all intermediate blocks in the path, and back to the starting block. The smaller the timestamp difference, the faster the risk transmission speed of the closed-loop path and the shorter the response cycle of the risk loop.

[0118] Closed-loop transmission frequency value: refers to the estimated frequency of a risk signal completing a full closed-loop transmission within a unit time step. It is determined by the number of funding function blocks in the path and the timestamp difference, and is used to measure the transmission activity of the risk closed-loop path in the time dimension.

[0119] The closed-loop transmission frequency value is determined by dividing the total number of funding function blocks in the path block set by the difference between the first and last timestamps. The quotient is the closed-loop transmission frequency value of that risk closed-loop path. The more blocks there are and the smaller the timestamp difference, the more synchronous risk responses are involved in the closed-loop path within a short time span. The higher the closed-loop transmission frequency value, the stronger the corresponding risk propagation activity.

[0120] It should be noted that the introduction of closed-loop transmission frequency values ​​aims to address the technical shortcomings of assessing path risk solely based on the magnitude of path state changes. If risk intensity is measured only by the cumulative magnitude of the path state change sequence, two risk closed-loop paths with the same total magnitude will be assigned the same rating, failing to distinguish between two fundamentally different risk propagation topologies: first, a path involving fewer blocks where risk signals propagate slowly over a longer time span; second, a path involving more blocks where risk signals spread rapidly within a shorter time span. In actual corporate financial networks, the latter implies a wider risk propagation chain, faster response speed, and a greater potential threat of systemic contagion, corresponding to a higher systemic risk exposure. By combining the number of path blocks with the difference between the first and last timestamps to construct a closed-loop propagation frequency value, and multiplying it by the total amplitude of path state changes to calculate the basic path risk value, the quantitative result of path risk intensity can simultaneously reflect the amplitude characteristics and temporal activity of risk signal propagation. Thus, under the same amplitude conditions, risk closed-loop paths with faster propagation speed and wider impact range are assigned a higher risk intensity rating, significantly improving the ability of risk assessment results to distinguish the actual systemic risk level and the sensitivity of early warning.

[0121] S504, Calculate the basic path risk value of the corresponding risk closed-loop path based on the path state change sequence and the closed-loop transmission frequency value; The basic path risk value refers to the quantitative value of path risk calculated solely based on the magnitude of changes in the risk state within the path and the level of transmission activity. It does not yet include corrections for the characteristics of capital return behavior and reflects the inherent propagation strength of the risk closed-loop path at the level of risk state evolution.

[0122] The basic path risk value for a corresponding risk closed-loop path is calculated as follows: The absolute values ​​of all state changes in the path state change sequence are summed to obtain the total amplitude of risk state changes within the path. This total amplitude value is then multiplied by the closed-loop propagation frequency value; the resulting product is the basic path risk value for that risk closed-loop path. The total amplitude value reflects the cumulative intensity of the risk signal propagation within the path, while the closed-loop propagation frequency value reflects the temporal activity of the propagation. Multiplying the two allows the basic path risk value to simultaneously reflect both the amplitude and temporal characteristics of the risk signal.

[0123] It should be noted that the calculation of the basic path risk value uses the multiplication of the total amplitude value and the closed-loop propagation frequency value because the overall propagation strength of the risk signal in the closed-loop path depends on two independent dimensions: the cumulative amplitude of the risk state changing step by step along the path (total amplitude value), and the density of these changes over time (closed-loop propagation frequency value). If only the total amplitude value is used, a path that propagates slowly over a long time span will have the same basic risk value as a path that propagates rapidly over a short time span, which contradicts the common sense that "the shorter the time and the faster the propagation, the more urgent the risk accumulation" in actual risk propagation. If only the frequency value is used, the magnitude of the actual change between each hop is ignored. Multiplying the two values ​​ensures that the basic path risk value reflects both the cumulative energy and temporal activity of the risk signal propagation, thus more accurately quantifying the inherent risk propagation strength of the path.

[0124] S505, Calculate the proportional relationship between the amount of funds flowing back and the amount of funds flowing out in the risk closed loop path, and correct the basic path risk value to obtain the path behavior risk value; The amount of funds flowing back refers to the actual inflow of funds from the path's termination node back to the initial fund functional block within the risk closed-loop path. It originates from the statistical value of the fund inflow corresponding to the last directed edge of the path in the set of directed flow relationships. The ratio between the amount of funds flowing back and the amount of funds flowing out initially reflects the completeness of the fund loop in this closed-loop path: the closer the ratio is to or higher than 1, the higher the proportion of funds flowing back in the closed loop, and the more significant the corresponding closed-loop behavior characteristics; the lower the ratio, the greater the proportion of funds lost or diverted during transmission, and the weaker the closed-loop characteristics.

[0125] Path behavior risk value: This refers to the risk quantification value obtained by adjusting the ratio between the amount of funds flowing back and the amount of funds flowing out, based on the basic path risk value. It comprehensively reflects the joint risk level of the risk closed loop path in two dimensions: risk state evolution and fund loop behavior.

[0126] The method for correcting the basic path risk value is as follows: divide the amount of returned funds by the initial outflow amount to obtain the fund return ratio; multiply the basic path risk value by the fund return ratio, and the product is the path behavior risk value. When the fund return ratio is greater than 1, the path behavior risk value is higher than the basic path risk value, indicating that the fund return has excessively amplified the overall risk of the path; when the fund return ratio is less than 1, the path behavior risk value is lower than the basic path risk value, indicating that the diversion and attenuation of funds during the transmission process has a suppressive effect on the intensity of path risk.

[0127] S506, Based on the path behavior risk value and the preset risk propagation association rule, determine the path risk intensity identifier of the corresponding risk closed-loop path.

[0128] Among them, the preset risk propagation association rules refer to the segmented judgment rules that map path behavior risk values ​​to discrete path risk intensity levels. The rule content includes the numerical boundaries of several risk intensity level intervals and the corresponding intensity level identifiers (e.g., divided into four levels: low risk, medium risk, high risk, and extremely high risk). The preset method is as follows: before system deployment, based on the distribution characteristics of path behavior risk values ​​in historical risk closed-loop path samples, the sample values ​​are statistically analyzed using quantiles. The sample values ​​are then segmented according to quantiles, and the boundary values ​​of each segment interval are used as the numerical boundaries of the rules. The risk level corresponding to each segment interval is used as the output identifier of the rule. The number of segments and the quantile values ​​of the rule can be adaptively configured according to the risk control precision requirements of the target enterprise.

[0129] The method for determining the path risk intensity identifier of the corresponding risk closed loop path is as follows: compare the path behavior risk value corresponding to the current risk closed loop path with the numerical boundaries of each level interval in the preset risk propagation association rules in sequence, determine the interval in which the path behavior risk value falls, and assign the intensity level identifier of the corresponding interval to the risk closed loop path as its path risk intensity identifier.

[0130] In one feasible approach, the contribution assessment module is specifically used for: S511. Along the corresponding risk closed-loop path, according to the transmission order of the capital function blocks in the risk closed-loop path, read the state change value corresponding to the adjacent capital function block, and calculate the state offset of the state change value corresponding to the next capital function block relative to the state change value corresponding to the previous capital function block. Among them, the transmission order refers to the arrangement order of the capital function blocks in the risk closed loop path from the starting node to the ending node according to the direction of the directed edge. It is consistent with the position numbering order in the path block set and is the unified operation order benchmark for reading the state change value of adjacent blocks, calculating the state offset, and executing delay correction.

[0131] State offset: refers to the difference between the state change value of the latter block at its corresponding transition time step and the state change value of the former block at its corresponding transition time step in two adjacent capital function blocks along the transmission sequence. It reflects the original amplitude deviation that occurs when the risk signal propagates between the adjacent block pairs.

[0132] One method for calculating the state offset is as follows: take adjacent block pairs in sequence according to the transmission order, read the state change value at the corresponding transition time step in their respective risk generation sequences, subtract the state change value of the previous block from the state change value of the subsequent block, and the difference is the state offset of the adjacent block pair.

[0133] S512, Based on the transmission mapping relationship, determine the fund transmission time interval between adjacent fund functional blocks, and perform time delay correction processing on the state offset according to the fund transmission time interval to obtain the path state increment value; The fund transmission time interval refers to the risk transmission lag value between two adjacent fund function blocks recorded in the transmission mapping relationship. It is expressed as the number of time steps and reflects the typical delay period experienced by the risk signal from the upstream block to the downstream block. The fund transmission time interval is directly read from the lag field of the transmission mapping relationship, without the need to re-perform time alignment calculation.

[0134] The method for determining the time interval of fund transmission between adjacent fund functional blocks is as follows: For each pair of adjacent blocks in the path, the corresponding record is retrieved in the transmission mapping relationship record set using the upstream block identifier and the downstream block identifier as indexes, and its risk transmission time lag value field is read, which is the time interval of fund transmission for that pair of adjacent blocks.

[0135] Path state increment value: refers to the value obtained after performing delay correction processing on the state offset. It reflects the actual amplitude increment of risk signal propagation between adjacent block pairs after eliminating the influence of propagation delay. It is an intermediate result for subsequent propagation offset correction processing to obtain the risk increment propagation value.

[0136] The delay correction process is performed by dividing the state offset of the current adjacent block pair by the corresponding fund transmission time interval. The quotient is the path state increment value. This division by the time interval normalizes the state offset to a unit time step granularity, eliminating the incomparability of absolute values ​​of state offsets caused by different transmission delays between adjacent block pairs, thus making the path state increment value comparable among adjacent block pairs within the path.

[0137] It should be noted that the reason for implementing delay correction is as follows: Due to the objective differences in the time intervals for fund transmission between different adjacent block pairs, the absolute values ​​of the state offset accumulated by a block pair with a longer transmission lag over a longer time span are not directly comparable in magnitude to the absolute values ​​of the state offset formed by a block pair with a shorter transmission lag over a shorter time span. If the risk propagation contribution of each block pair is directly measured by the original state offset, the absolute value of the state offset of a block pair with a longer transmission lag will naturally be larger due to the longer accumulation time, and it is easy to be misjudged as a channel with stronger risk propagation capabilities; while even if the risk propagation intensity of a block pair with a shorter transmission lag is higher per unit time, it may be underestimated due to its smaller absolute value. By dividing the state offset by the time interval of fund transmission, the magnitude of risk state changes of each adjacent block pair is uniformly converted to the unit time step granularity. This enables block pairs under different transmission time delay conditions to make horizontal comparisons of risk propagation intensity on the same time base, ensuring that the calculation results of subsequent risk increment propagation values ​​truly reflect the actual propagation capacity of each block per unit time in the path, rather than being affected by the length of the transmission time delay.

[0138] S513: Extract the inflow and outflow of funds between adjacent fund function blocks in the corresponding risk closed-loop path, and perform propagation offset correction processing on the corresponding path state increment value according to the offset relationship between the inflow and outflow of funds to obtain the risk increment propagation value of the corresponding fund function block.

[0139] The offset relationship between capital inflows and outflows refers to the ratio between the actual capital inflow received by the downstream block and the actual capital outflow issued by the upstream block in a pair of adjacent blocks connected by a directed edge. This ratio reflects the actual transmission efficiency of funds in this transmission channel: a ratio of 1 indicates complete capital transmission, a ratio less than 1 indicates diversion or interception, and a ratio greater than 1 indicates an amplification effect caused by external capital inflows. The data for capital inflows and outflows are derived from the statistical fields of the corresponding directed edges in the set of directed flow relationships.

[0140] The method for performing propagation offset correction is as follows: divide the inflow of funds in the downstream block by the outflow of funds in the upstream block to obtain the fund transfer ratio of the adjacent block pair; multiply the path state increment value by this fund transfer ratio value, and the product is the risk increment propagation value of the corresponding downstream fund function block. The correction effect of the fund transfer ratio value on the path state increment value is that: when the fund transfer ratio is less than 1, the actual risk signal strength received by the downstream block is reduced accordingly; when the fund transfer ratio is greater than 1, the inflow of external funds leads to a corresponding amplification of the risk signal strength of the downstream block, so that the risk increment propagation value can simultaneously reflect the joint influence of risk state evolution and fund flow behavior.

[0141] It should be noted that the propagation offset correction is applied because the path state increment only reflects the time-normalized difference in the magnitude of risk state changes between adjacent blocks, but does not include the impact of the actual scale of fund transfers on the intensity of risk propagation. In actual corporate fund operations, the propagation intensity of risk signals depends not only on the magnitude of risk state changes but also on the actual flow of funds in the transmission channel: if only a small portion of the funds flowing out of upstream blocks actually enter downstream blocks, the risk exposure borne by downstream blocks will be correspondingly reduced, and the path state increment will overestimate the actual risk propagation intensity of the channel; if downstream blocks receive external funds in addition to upstream funds, their actual risk exposure will be higher than what can be reflected by the magnitude of state changes alone, and the path state increment will underestimate the actual risk propagation intensity of the channel. If the above-mentioned fund flow deviation is not corrected, the differences in propagation intensity caused by different levels of fund diversion or convergence between different block pairs will be ignored, and the risk increment propagation value cannot truly reflect the actual amount of risk propagation borne by each downstream block in the path. By introducing a proportion of fund transfer to correct the incremental value of the path state, the amplification or reduction effect of fund flow behavior on risk propagation is included in the calculation, so that the incremental risk propagation value can simultaneously reflect the combined effect of the magnitude of risk state evolution and the actual scale of fund circulation, thereby achieving effective integration of information from both the risk state and fund behavior dimensions in the risk contribution assessment of each block.

[0142] In one feasible approach, the contribution assessment module is specifically used for: S521, Obtain the set of path blocks corresponding to the risk closed-loop path, and read the position number of each fund function block in the risk closed-loop path from the set of path blocks; S522, Read the path risk intensity identifier corresponding to the risk closed-loop path, and associate the path risk intensity identifier as the path risk benchmark value with each capital function block in the capital function block sequence; The path risk benchmark value refers to the quantitative benchmark obtained by numerically representing the risk level corresponding to the path risk intensity identifier. It is used to provide a unified upper limit reference for the allocation of risk contributions of each funding functional block within the same risk closed-loop path. The role of the path risk benchmark value is to ensure that the risk contribution assessment of each funding functional block should not be conducted independently of the overall risk level of its closed-loop path. The contributions of each block in a high-intensity path should be allocated under a higher overall benchmark, while the contributions of each block in a low-intensity path should be allocated under a lower overall benchmark, thereby ensuring the cross-path comparability of block risk contribution values ​​between different risk closed-loop paths.

[0143] The path risk benchmark value is associated with each funding functional block in the funding functional block sequence as follows: For each funding functional block in the path block set, the path risk benchmark value corresponding to that path is written as an additional field into the processing record of that block under the current path. This allows the path risk benchmark value to be directly read from this field when calculating the risk contribution value for each block, without having to repeatedly query the path risk intensity identifier. This association operation is performed to push path-level information down to the block-level processing unit, avoiding frequent cross-level retrieval of path attributes during block-by-block calculations and reducing data access overhead in the calculation process.

[0144] S523, for each funding functional block in the path block set, extract the corresponding risk increment propagation value, and calculate the total propagation of the risk closed loop path based on the risk increment propagation value corresponding to each funding functional block; Among them, the total propagation amount refers to the sum of the risk increment propagation values ​​of all capital function blocks in the risk closed loop path, reflecting the cumulative intensity of risk signal propagation in the overall path, and is the denominator benchmark for calculating the propagation ratio of each block.

[0145] The method for calculating the total propagation of the risk closed-loop path is as follows: traverse all the capital function blocks in the path block set, read the risk increment propagation value corresponding to each block in turn, sum all the risk increment propagation values, and the sum is the total propagation of the risk closed-loop path.

[0146] S524, calculate the propagation ratio of each funding functional block based on the total propagation amount, and allocate the path risk benchmark value to the corresponding funding functional block according to the propagation ratio to obtain the path allocation risk value of each funding functional block; The propagation percentage refers to the proportion of the incremental risk propagation value of a specific funding functional block to the total propagation value of that risk closed-loop path, reflecting the relative contribution weight of that block in the overall risk propagation of the path. The propagation percentage is obtained by dividing the incremental risk propagation value of that block by the total propagation value, and the sum of the propagation percentages of all blocks is 1.

[0147] Path allocation risk value: This refers to the value obtained by allocating the path risk benchmark value to the corresponding funding functional blocks according to their propagation proportion. It reflects the amount of path-level risk that the block should bear based on its actual risk propagation contribution weight in the path. The path allocation risk value is obtained by multiplying the path risk benchmark value by the propagation proportion of the block.

[0148] S525, calculate the path allocation risk value of each funding functional block and the risk increment propagation value of the corresponding funding functional block to obtain the risk contribution value of each funding functional block; Among them, the risk contribution value is a joint quantitative indicator that comprehensively reflects the absolute propagation intensity and relative allocation weight of a certain funding function block in the corresponding risk closed-loop path. It reflects both the actual amplitude of the risk signal propagation of the block itself and its contribution share under the overall risk intensity framework of the path.

[0149] The method for calculating the risk contribution value of a funding function block is as follows: multiply the path allocation risk value of the block by the corresponding risk increment propagation value. The product is the risk contribution value of the block. The path allocation risk value reflects the block's allocation weight under the total path benchmark, while the risk increment propagation value reflects the block's own propagation strength. Multiplying the two ensures that the risk contribution value is simultaneously constrained by both the overall risk level of the path and the individual propagation capacity of the block, avoiding a one-sided assessment of the block's contribution based solely on the proportion of propagation or the absolute size of the propagation value.

[0150] S526. Based on the ranking relationship of the risk contribution values ​​of each funding functional block in the same risk closed-loop path, generate the block risk contribution identifier of each funding functional block.

[0151] The method for generating the block risk contribution identifier for each funding functional block is as follows: All funding functional blocks in the same risk closed-loop path are arranged in descending order of risk contribution value. Based on the arrangement, each block is divided into a preset contribution level range and assigned a corresponding discrete level identifier (e.g., divided into three levels: primary contribution block, secondary contribution block, and general contribution block). Blocks with higher risk contribution values ​​receive a higher-level block risk contribution identifier, while blocks with lower values ​​receive a lower-level identifier. When a block in the same risk closed-loop path participates in the contribution calculation of multiple paths simultaneously, the maximum risk contribution value of that block in each path is taken as the final risk contribution value of that block in the current closed-loop path, and then the sorting and identifier assignment operations are performed.

[0152] Traditional financial risk contribution assessment systems, when handling multiple risk closed-loop paths, require repeated reading and calculation of the original transaction flow data of each block within the path. As the number and length of paths increase, this leads to a surge in disk I / O requests caused by concurrent access to the original data of the same block across multiple paths, and a linear increase in memory usage due to maintaining cross-block time-aligned intermediate states. In large-scale multi-path concurrency scenarios, this creates a significant data processing bottleneck. The contribution assessment module in this application compresses the original multi-dimensional financial indicator data along the path into a one-dimensional path state change sequence, eliminating the need for repeated reading of original transaction data during the path assessment stage and fundamentally eliminating redundant I / O access during multi-path concurrency. The latency correction process converts the state offset to a unit time step granularity, making the risk propagation intensity of each block directly comparable, and eliminating the need for the system to maintain a separate time-aligned intermediate state cache for each pair of blocks. The propagation offset correction embeds the impact of cash flow into the risk increment propagation value in the form of a constant factor, avoiding secondary retrieval of cash flow data during the contribution allocation stage. This application compresses the data scale required for block-level evaluation of each path to a single risk increment propagation value through a contribution evaluation module. This allows contribution allocation calculation to only require performing arithmetic operations on this single value. In the scenario of concurrent evaluation of a large-scale risk closed-loop path set, this effectively reduces the computational resource consumption and data read / write latency of a single path evaluation, and improves the overall data processing throughput of the system.

[0153] Example 6, as Figure 1 As shown, this application provides a comprehensive assessment and check-up system for enterprise financial risk control. The system includes a result generation module, as detailed below: The results generation module is used to construct a risk heatmap based on the block risk contribution identifier and path risk intensity identifier of the fund transaction data, and generate enterprise financial risk assessment results.

[0154] In this embodiment, a brief introduction to the following related terms is given: Risk Heatmap: A visual image generated by mapping the risk contribution indicators of each block and the risk intensity indicators of each path onto a two-dimensional graphic structure using visual encoding methods such as color saturation, brightness, and color overlay, with the capital function blocks as nodes and risk closed-loop paths as connecting channels. The risk heatmap intuitively presents the risk contribution level of each block in the enterprise's capital network and the risk propagation intensity of each closed-loop path through changes in color depth and brightness. This allows risk distribution characteristics and transmission structure to be directly perceived within the same graphical interface, eliminating the need to read numerical reports one by one.

[0155] Corporate financial risk assessment results: refers to the final visualized risk analysis report output by the results generation module. With a risk heat map as the core carrier, it comprehensively displays the risk contribution level of each capital function block of the target enterprise within the statistical period, the risk propagation intensity of each risk closed loop path, and the spatial distribution of risk cluster areas. This allows risk management personnel to directly read the report and formulate targeted risk intervention measures accordingly.

[0156] In one feasible approach, the result generation module is specifically used for: S601, read the block risk contribution identifier corresponding to each fund function block and the path risk intensity identifier corresponding to each risk closed loop path; S602, associate the block risk contribution identifier corresponding to each fund function block with the fund inflow and fund outflow in the corresponding fund transaction data to determine the block display area corresponding to each fund function block; The block display area refers to the area within the risk heat map. Figure 2 A graphical placeholder area is allocated on the 3D plane for a specific funding functional block. Its shape uses rectangular or circular nodes, and the area is determined by the sum of the funding inflow and outflow within that block during the statistical period. Blocks with a larger sum of inflow and outflow have a larger area, ensuring that the visual weight of the block on the graph matches its actual funding volume. The planar position of the block display area is arranged according to the topological relationships of the funding functional blocks in the directed graph. Blocks with direct funding flow relationships are placed closer together on the plane to preserve the topological structure information of the funding network.

[0157] S603, associate the path risk intensity identifier corresponding to each risk closed loop path with the transmission order of the capital function block in the risk closed loop path to determine the path connection area corresponding to each risk closed loop path. The path connection region refers to the area within the risk thermal zone. Figure 2 On the 3D plane, a set of directed line segments or curves are formed by sequentially connecting the corresponding block display areas along the propagation direction of each directed edge in the risk closed-loop path. The starting point and ending point of each line segment correspond to the boundary of the upstream and downstream block display areas of the corresponding directed edge, respectively. The width of the line segments connecting the paths is determined based on the path risk intensity indicator of the corresponding risk closed-loop path. The higher the level of the path risk intensity indicator, the wider the line segment, so as to visually reflect the difference in risk propagation energy of different paths.

[0158] S604. Based on the block risk contribution identifier corresponding to each capital function block and the path risk intensity identifier corresponding to each risk closed loop path, a risk heat map is drawn for the block display area and the path connection area to obtain a visualized corporate financial risk assessment result. The corporate financial risk assessment result is used to display the risk assessment status of corporate financial risk control.

[0159] The method for drawing the risk heatmap is as follows: a graphics rendering tool (such as an SVG vector graphics rendering engine or a Canvas 2D drawing interface) is used to render the block display area and the path connection area layer by layer. The node graphic layer is drawn according to the area and position parameters of each block display area, and the path graphic layer is drawn according to the line segment parameters of each path connection area. The graphic elements are color-coded according to the block risk contribution indicator and the path risk intensity indicator. All layers are merged and output into a single risk heatmap image, which serves as a visualization carrier of the enterprise's financial risk assessment results.

[0160] In one feasible approach, the result generation module is also used for: S611, based on the block risk contribution identifier corresponding to each funding function block, perform color layering processing on the corresponding block display area, wherein the higher the risk contribution value corresponding to the block risk contribution identifier, the higher the color saturation of the corresponding block display area. Color layering processing refers to assigning the saturation values ​​of each level in the preset color scheme to the fill color of the corresponding block display area based on the discrete level corresponding to the risk contribution identifier of the block. This makes the blocks with different risk contribution levels present visually distinguishable color depths in the heat map. Blocks with high contribution levels have dark and saturated colors, while blocks with low contribution levels have light colors. This allows risk control personnel to directly identify the spatial location of high-risk contribution blocks by color depth without reading numerical values.

[0161] The method for performing color layering is as follows: pre-configure corresponding color saturation parameter values ​​for each risk contribution level (for example, assigning saturation values ​​of 20%, 50%, 80%, and 100% respectively from low to high). When rendering the block display area, read the corresponding saturation parameter value according to the block risk contribution identifier, assign the saturation parameter value to the saturation channel of the fill color of the block display area, keep the hue and brightness channels unchanged, and complete the color layering rendering.

[0162] S612, based on the path risk intensity identifier corresponding to each risk closed loop path, perform color gradient processing on the corresponding path connection area, wherein the color brightness of the subsequent path connection area along the transmission direction of the risk closed loop path is gradually increased. The color gradient processing method is as follows: For a specific risk closed-loop path, the base hue of the path's color gradient is determined based on its risk intensity indicator, and all directed edges in the path are numbered according to the transmission sequence. The starting and ending values ​​of the color brightness are evenly divided according to the total number of directed edges, and brightness parameter values ​​that increase progressively from the starting brightness to the ending brightness are sequentially assigned to the path connection areas corresponding to each directed edge. This makes the color of the path connection areas gradually brighten from dark to light along the transmission direction, visually presenting the directionality and hierarchy of the risk signal transmission along the path. The higher the risk intensity indicator level of the path, the more vivid the color scheme corresponding to its base hue, so as to clearly distinguish it from low-intensity paths.

[0163] S613, when the same block display area corresponds to multiple path connection areas, perform color overlay processing on the color gradient areas corresponding to the multiple path connection areas, and generate risk clustering identifiers for the corresponding fund function blocks based on the coverage of the color overlay areas.

[0164] The color overlay processing method is as follows: When multiple paths intersect at the boundary of the same block display area, the color gradient termination segment of each path connection area near the boundary of the block display area is extracted as an independent color layer. The multiple color layers are then overlaid and rendered sequentially to the edge area of ​​the block display area using a transparency blending overlay method. The overlaid mixed color carries the hue information of multiple paths at the same time. The more layers there are and the higher the color saturation of each layer, the more striking the visual effect after blending. The actual coverage area of ​​the color overlay area expands as the number of paths involved in the overlay increases.

[0165] Risk clustering indicator: This refers to a discrete level marker generated based on the proportion of the area covered by the color-overlay region on the boundary of the corresponding block display area to the total boundary length of that block display area. A higher coverage ratio indicates a greater number of risk closed-loop paths converging into that block, a stronger degree of risk aggregation at that block, and a higher risk clustering indicator level. Risk clustering indicators are displayed on the heatmap as a ring-shaped highlighted mark at the edge of the block display area, allowing risk management personnel to quickly locate core node blocks in the enterprise's financial network where risk paths are highly concentrated.

[0166] The result generation module of this application maps block-level risk contribution identifiers to the color saturation of the block display area, path-level risk intensity identifiers to the color brightness gradient of the path connection area, and multi-path convergence features to the color overlay coverage and risk cluster identifiers. This simultaneously encodes information from three dimensions—block-level risk contribution, path-level risk intensity, and network-level risk convergence—in a single thermal image. This allows risk management personnel to directly perceive the overall distribution characteristics of the enterprise's financial risk network within the same graphical interface without performing cross-view data retrieval. The differentiated encoding design of the three visual channels—color saturation, brightness, and overlay area—makes the three types of information visually independent and non-interfering, avoiding the visual aliasing problem commonly encountered when multi-dimensional information is overlaid in a single view. This reduces the cognitive burden of information reading while ensuring the readability and spatial positioning accuracy of the risk assessment results.

[0167] The foregoing has shown and described the basic principles, main features, and advantages of this application. Those skilled in the art should understand that this application is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of this application. Various changes and modifications can be made to this application without departing from the spirit and scope thereof, and all such changes and modifications fall within the scope of this application as claimed. The scope of protection of this application is defined by the appended claims and their equivalents.

Claims

1. A comprehensive assessment and diagnostic system for enterprise financial risk control, characterized in that, The system includes: The functional block division module is used to perform functional division processing on the target company's fund transaction data based on the fund transfer relationship between accounts and the purpose of funds, forming multiple fund functional blocks, and statistically analyzing the fund inflow and outflow between any two fund functional blocks to construct a set of directional flow relationships between fund functional blocks; The risk sequence generation module is used to extract the financial indicator sequence of each funding functional block from the funding transaction data, and input the financial indicator sequence into a pre-trained risk state transition model to perform risk evolution and obtain a risk generation sequence. The transmission path generation module is used to time-align the set of directional flow relationships with the risk generation sequence of the corresponding capital functional blocks, establish the transmission mapping relationship of the risk generation sequence between different capital functional blocks, and generate a set of risk transmission paths between capital functional blocks according to the transmission mapping relationship. The closed-loop identification module is used to perform risk transmission processing along the risk transmission paths in the risk transmission path set according to the state change information of the risk generation sequence, and to filter the risk closed-loop path based on the processing result; The contribution assessment module is used to determine the path risk intensity identifier based on the path characteristics of the risk closed loop path, determine the risk increment propagation value of each fund functional block in the corresponding risk closed loop path based on the transmission mapping relationship, and generate the block risk contribution identifier of each fund functional block according to the correlation between the risk increment propagation value and the path risk intensity identifier of the corresponding risk closed loop path. The results generation module is used to construct a risk heatmap based on the block risk contribution identifier and path risk intensity identifier of the fund transaction data, and generate enterprise financial risk assessment results.

2. The system according to claim 1, characterized in that, The conduction path generation module is specifically used for: A directed graph of fund functional blocks is constructed, with each fund functional block as a node and the directed flow relationships in the set of directed flow relationships as edges. For each directed edge in the directed graph of the fund function blocks, determine the upstream and downstream fund function blocks of the directed edge, and perform feature extraction based on the state change information of the risk generation sequence of the upstream and downstream fund function blocks to obtain the risk state label of the upstream fund function block and the first time step of the risk state label transition, and the risk state label of the downstream fund function block and the second time step of the risk state label transition. The time step interval between each upstream funding functional block and the downstream funding functional block is calculated based on the first time step and the second time step, and the time step interval with the highest frequency is taken as the risk transmission delay value of the directed edge. Based on the risk transmission lag value, time shift processing is performed on the transition time steps of each risk status label of the downstream capital function block, and the risk transmission lag value is subtracted from the second time step to obtain the corresponding alignment time step; The alignment time step is matched with the first time step of the upstream funding function block in chronological order to determine the alignment time step corresponding to the first time step, thus forming a corresponding time step pair. For each corresponding time step pair, extract the transition direction of the upstream funding function block at the first time step corresponding to the risk status label, and the transition direction of the downstream funding function block at the aligned time step corresponding to the risk status label. The number of time steps with the same transition direction is counted, and the proportion of the number to the total number of risk state label transition time steps in the upstream capital function block is calculated. The proportion is determined as the risk state transition same direction rate of the corresponding directed edge. Directed edges whose risk state transition homing rate exceeds a preset threshold are marked as valid transmission edges, and the valid transmission edges and their corresponding risk transmission delay values ​​are encapsulated into a transmission mapping relationship. Based on the aforementioned transmission mapping relationship, valid transmission paths are identified from the directed graph of the capital function block as risk transmission paths, and the risk transmission paths are summarized to obtain a risk transmission path set.

3. The system according to claim 1, characterized in that, The closed-loop identification module is specifically used for: The risk state label is read step by step from the state change information of the risk generation sequence, and the changes in risk state label between adjacent time steps are compared step by step to generate a risk state change sequence. Based on the risk state change sequence, the time step of the risk state transition is identified, the corresponding capital function block that has the transition is determined as the risk activation block, and the time step of the first transition is determined as the activation timestamp. Starting with the risk activation block as the starting node, each risk transmission path in the risk transmission path set is traversed path by path, and the subsequent funding function blocks are determined level by level according to the directed edge order in the risk transmission path. Based on the risk transmission time lag value corresponding to each directed edge, the time step of the risk state transition of the subsequent capital function block is calculated to obtain the corresponding target time step. At the target time step, the risk status label of the corresponding fund function block is read and compared with the risk status label at the activation timestamp to determine whether a risk status transition in the same direction has occurred, and the risk transmission path corresponding to the risk status transition in the same direction is marked as a candidate risk closed loop path. When the termination node of the candidate risk closed loop path is the same funding function block as the risk activation block, the candidate risk closed loop path is determined as the risk closed loop path.

4. The system according to claim 1, characterized in that, The contribution evaluation module is specifically used for: The location of each funding functional block in the risk closed-loop path is numbered according to the transmission order to obtain the path block set; Based on the path block set, extract the state change values ​​corresponding to adjacent funding function blocks from the state change information of the risk generation sequence, and calculate the state change amount between adjacent funding function blocks to obtain the path state change sequence. Based on the set of path blocks, the number of fund function blocks between the initial fund function block and the return to the initial fund function block is counted, and the closed-loop transmission frequency value is determined by combining the timestamp difference between the first and last fund function blocks in the set of path blocks. Based on the path state change sequence and the closed-loop transmission frequency value, calculate the basic path risk value of the corresponding risk closed-loop path; The proportional relationship between the amount of funds flowing back and the amount of funds flowing out in the risk closed loop path is statistically analyzed, and the basic path risk value is corrected to obtain the path behavior risk value. Based on the path behavior risk value and the preset risk propagation association rules, the path risk intensity identifier of the corresponding risk closed-loop path is determined.

5. The system according to claim 1, characterized in that, The contribution evaluation module is specifically used for: Along the corresponding risk closed loop path, according to the transmission order of the capital function blocks in the risk closed loop path, read the state change values ​​corresponding to the adjacent capital function blocks, and calculate the state offset of the state change value corresponding to the next capital function block relative to the state change value corresponding to the previous capital function block. Based on the transmission mapping relationship, the fund transmission time interval between adjacent fund functional blocks is determined, and the state offset is subjected to time delay correction processing according to the fund transmission time interval to obtain the path state increment value. Extract the inflow and outflow of funds between adjacent fund function blocks in the corresponding risk closed-loop path, and perform propagation offset correction processing on the corresponding path state increment value according to the offset relationship between the inflow and outflow of funds to obtain the risk increment propagation value of the corresponding fund function block.

6. The system according to claim 1, characterized in that, The contribution evaluation module is specifically used for: Obtain the set of path blocks corresponding to the risk closed-loop path, and read the position number of each funding function block in the risk closed-loop path from the set of path blocks; Read the path risk intensity identifier corresponding to the risk closed-loop path, and associate the path risk intensity identifier as the path risk benchmark value with each capital function block in the capital function block sequence; For each funding functional block in the path block set, the corresponding risk increment propagation value is extracted, and the total propagation of the risk closed loop path is calculated based on the risk increment propagation value corresponding to each funding functional block. The propagation ratio of each funding functional block is calculated based on the total propagation volume, and the path risk benchmark value is allocated to the corresponding funding functional block according to the propagation ratio to obtain the path allocation risk value of each funding functional block. The path allocation risk value of each funding functional block and the risk increment propagation value of the corresponding funding functional block are calculated to obtain the risk contribution value of each funding functional block. Based on the ranking of the risk contribution values ​​of each funding functional block within the same risk closed-loop path, a block risk contribution identifier is generated for each funding functional block.

7. The system according to claim 1, characterized in that, The risk sequence generation module is specifically used for: From the aforementioned fund transaction data, according to the division results of each fund functional block, the net cash flow, cash turnover rate, and short-term debt coverage ratio of each fund functional block in each statistical period are extracted as financial indicators, and the financial indicators are arranged in the order of time steps to form a sequence of financial indicators for the corresponding fund functional block. The financial indicator sequence is input into a pre-trained risk state transition model, which includes a temporal feature extraction layer, a hidden state transition layer, and a risk state output layer. The temporal feature extraction layer performs a linear transformation on the financial indicator sequence at each time step to generate a local financial feature vector for that time step. The hidden state transition layer receives the local financial feature vector at the current time step and performs gated fusion calculations with the hidden state vector from the previous time step to generate the hidden state vector for the current time step. The risk state output layer performs mapping calculations on the hidden state vectors at each time step, outputs the risk state probability distribution for that time step, and determines the risk state level corresponding to the maximum probability as the risk state label for that time step. Based on the time step sequence, the risk status label corresponding to each time step is calculated step by step. The difference between the risk status labels between adjacent time steps is used as the status change value, and the status change value and the risk status label are encapsulated into status change information. The state change information is combined with the corresponding time step and recorded to generate a risk generation sequence for the corresponding funding function block.

8. The system according to claim 1, characterized in that, The functional block division module is specifically used for: Extract transaction account identifiers, transaction times, transaction amounts, transaction directions, and transaction purposes from the target company's fund transaction data. Based on the transaction account identifiers, conduct fund flow analysis on the transaction times, transaction amounts, and transaction directions to obtain fund transfer relationships. The fund transfer relationships include at least the number of fund transfers, the amount of fund transfers, and the time interval between fund transfers. For any two trading accounts, the degree of correlation between the funds flow between the trading accounts is determined based on the number of fund flows, the amount of fund flows, and the time interval between fund flows. Based on the transaction purpose information corresponding to each transaction account, the types of fund usage of the transaction accounts are classified, and the fund inflow and outflow of each transaction account under different fund usage types are statistically analyzed. Based on the degree of correlation between fund transfers between trading accounts and the type of fund usage of the corresponding trading accounts, trading accounts with the same type of fund usage and whose degree of correlation between fund transfers meets the preset conditions are classified and processed to form multiple fund function blocks; Read the corresponding fund transaction data between transaction accounts belonging to different fund function blocks, and determine the directed flow relationship between fund function blocks based on the outflow and inflow directions of funds in the fund transaction data; The inflow and outflow of funds corresponding to each directional flow relationship are statistically analyzed to generate a set of directional flow relationships between fund functional blocks.

9. The system according to claim 1, characterized in that, The result generation module is specifically used for: Read the block risk contribution identifier corresponding to each funding function block and the path risk intensity identifier corresponding to each risk closed loop path; The block risk contribution identifier corresponding to each fund function block is associated with the fund inflow and outflow in the corresponding fund transaction data to determine the block display area corresponding to each fund function block; Associate the path risk intensity identifier corresponding to each risk closed loop path with the transmission order of the capital function block in the risk closed loop path to determine the path connection area corresponding to each risk closed loop path. Based on the block risk contribution identifier corresponding to each capital function block and the path risk intensity identifier corresponding to each risk closed loop path, a risk heat map is drawn for the block display area and path connection area to obtain a visualized corporate financial risk assessment result. The corporate financial risk assessment result is used to display the risk assessment status of corporate financial risk control.

10. The system according to claim 1, characterized in that, The result generation module is also used for: Based on the block risk contribution identifier corresponding to each funding function block, the corresponding block display area is processed with color layering. The higher the risk contribution value corresponding to the block risk contribution identifier, the higher the color saturation of the corresponding block display area. Based on the path risk intensity indicator corresponding to each risk closed loop path, the corresponding path connection area is subjected to color gradient processing, wherein the color brightness of the subsequent path connection area along the risk closed loop path transmission direction is gradually increased. When the same block display area corresponds to multiple path connection areas, color overlay processing is performed on the color gradient areas corresponding to the multiple path connection areas, and risk aggregation identifiers of the corresponding fund function blocks are generated based on the coverage of the color overlay areas.

Images