Contract settlement and automatic reconciliation system and method based on deep integration of industry and finance data

Abstract

This invention relates to the field of contract reconciliation management technology, specifically disclosing a contract settlement and automatic reconciliation system and method that deeply integrates business and financial data. After acquiring business performance data sequences and financial settlement data sequences, the system maps them into a two-dimensional time-series directed graph. Upon receiving an abnormal settlement trigger instruction, it traces back along the flow of funds and penetrates forward along the business-finance mapping to capture affected node clusters. Then, based on the comparison results between the accounting period identifier and the current closing time, it generates inter-period reversal action instructions and original order reduction action instructions, constructs alternative reversal path chains, and imports them into a virtual sandbox memory for verification. When the updated amount value is lower than a preset lower threshold, it triggers backtracking and reorganization until a legal reversal path scheme is selected. Finally, it calculates a physical disturbance score by combining the total number of nodes and the total number of inter-period actions, selects the scheme with the lowest score, and outputs it to the accounting platform to complete the synchronous update of business-finance reconciliation.

Description

Technical Field

[0001] This invention belongs to the field of contract reconciliation management technology, and involves a contract settlement and automatic reconciliation system and method that deeply integrates business and financial data. Background Technology

[0002] In corporate contract performance and financial settlement activities, the settlement results directly affect the accuracy of fund payments, revenue recognition, and subsequent reconciliation, serving as a crucial foundation for ensuring a closed-loop contract performance and the authenticity and consistency of financial data. As contract management models become increasingly complex, many business scenarios have evolved from single-contract settlements to multi-level nested settlement networks. For example, a general contracting contract often involves multiple subcontracting contracts, each of which is further linked to multiple orders and payment orders, forming a settlement chain topology with clear hierarchical relationships and monetary transfer relationships. In this structure, if a completed settlement is identified as overpayment, incorrect payment, or duplicate payment, simply rolling back the erroneous settlement node can easily disrupt the original accounting balance between its upstream and downstream nodes, leading to problems such as abnormal settlement amounts for subsequent contracts, distorted subcontract balances, and cascading deviations in the payment chain.

[0003] Current settlement reversal technology primarily relies on unidirectional reversals for single settlements, typically only statically canceling erroneous settlements themselves. It lacks the ability to automatically identify the impact paths of related nodes in complex contract topologies and also lacks a verification mechanism to determine whether reversal actions will cause accounting imbalances upstream or downstream. For settlement networks with intertwined multi-level contracts and orders, current technology struggles to automatically plan processing paths that balance business relationships and accounting constraints, and it cannot dynamically adjust the amount allocation range during the reversal process. This can easily lead to problems such as negative amounts for some nodes after reversal, broken mapping relationships, and inability to close historical vouchers. Summary of the Invention

[0004] In view of the problems existing in the prior art, the present invention provides a contract settlement and automatic reconciliation system and method with deep integration of business and financial data to solve the above-mentioned technical problems.

[0005] To achieve the above and other objectives, the technical solution adopted by the present invention is as follows:

[0006] The first aspect of this invention provides a method for contract settlement and automatic reconciliation based on deep integration of business and financial data, the method comprising:

[0007] Obtain business performance data sequences and financial settlement data sequences;

[0008] The business performance data sequence and the financial settlement data sequence are mapped into a business-finance two-dimensional time-series directed graph. The nodes of the business-finance two-dimensional time-series directed graph contain state vectors with monetary values, business volume values ​​and accounting period identifiers. The edges of the business-finance two-dimensional time-series directed graph contain fund flow constraint edges and business-finance mapping constraint edges.

[0009] Receive abnormal settlement trigger instructions and extract the abnormal target nodes corresponding to the abnormal settlement trigger instructions;

[0010] Starting from the abnormal target node, perform a source traversal along the opposite direction of the capital flow constraint edge in the business and finance two-dimensional time-series directed graph, and perform a penetration traversal along the positive direction of the business and finance mapping constraint edge to extract the connected affected node clusters.

[0011] Enter the offsetting strategy generation loop, which includes:

[0012] Extract the accounting period identifier for each node in the affected node cluster;

[0013] Compare the accounting period identifier with the current system closing time;

[0014] If the accounting period identifier is earlier than the current system closing time, an inter-period reversal instruction with a negative amount will be generated.

[0015] If the accounting period identifier is not earlier than the current system closing time, then an original order deduction action instruction will be generated;

[0016] Based on the inter-period reversal action instructions and the original order reduction action instructions, and combined with the reverse connection relationship of the fund flow constraint edge, an alternative reversal path chain is generated.

[0017] The alternative reversal path chain is imported into the virtual sandbox memory and put into a validity check loop. The validity check loop includes:

[0018] According to the alternative offsetting path chain, the amount and business volume values ​​in the state vectors of each node in the affected node cluster are deducted from the virtual sandbox memory to obtain the updated state vectors.

[0019] Determine whether the monetary value in the updated state vector is lower than a preset lower threshold.

[0020] If the amount in the updated state vector is lower than the preset lower threshold, the backtracking mechanism is triggered, and the amount corresponding to the abnormal target node is split according to the preset splitting ratio and the reversal strategy generation loop is re-entered.

[0021] If the amount values ​​in the updated state vector are not lower than the preset lower threshold, then the corresponding alternative reversal path chain is extracted as a legal reversal path scheme.

[0022] The physical disturbance score is calculated based on the total number of nodes in the legal offsetting path scheme and the total number of cross-period offsetting action instructions.

[0023] The legal reversal path scheme with the lowest physical disturbance score is selected and output to the accounting platform for synchronous update of business and financial reconciliation data.

[0024] The second aspect of this invention provides a contract settlement and automatic reconciliation system that deeply integrates business and financial data, the system comprising:

[0025] The graph construction module is used to generate a business-finance two-dimensional time-series directed graph based on the business performance data sequence and the financial settlement data sequence. The nodes of the business-finance two-dimensional time-series directed graph carry a state vector consisting of a monetary value, a business volume value, and an accounting period identifier. The edges of the business-finance two-dimensional time-series directed graph include cash flow constraint edges and business-finance mapping constraint edges.

[0026] An anomaly receiving module is used to receive an anomaly settlement trigger instruction and extract the anomaly target node corresponding to the anomaly settlement trigger instruction;

[0027] The affected node extraction module is used to perform source traversal in the business and finance two-dimensional time-series directed graph along the opposite direction of the capital flow constraint edge, starting from the abnormal target node, and to perform penetration traversal along the positive direction of the business and finance mapping constraint edge, and to extract the connected clusters of affected nodes.

[0028] The reversal strategy generation module is used to enter the reversal strategy generation loop, extract the accounting period identifier of each node in the affected node cluster, and compare the accounting period identifier with the current system closing time. When the accounting period identifier is earlier than the current system closing time, a cross-period reversal action instruction with a negative amount is generated. When the accounting period identifier is not earlier than the current system closing time, an original order reduction action instruction is generated. Based on the cross-period reversal action instruction and the original order reduction action instruction, and combined with the reverse connection relationship of the fund flow constraint edge, an alternative reversal path chain is generated.

[0029] The sandbox verification module is used to import the alternative offsetting path chain into the virtual sandbox memory and enter a legality verification loop. According to the alternative offsetting path chain, it deducts the amount value and business volume value of each node in the affected node cluster from the state vector in the virtual sandbox memory to obtain the updated state vector. It determines whether the amount value in the updated state vector is lower than the preset lower threshold. When the amount value in the updated state vector is lower than the preset lower threshold, a backtracking mechanism is triggered. When the amount value in the updated state vector is not lower than the preset lower threshold, the corresponding alternative offsetting path chain is extracted as the legal offsetting path scheme.

[0030] The scoring output module is used to calculate the physical disturbance score based on the total number of nodes in the legal reversal path scheme and the total number of inter-period reversal action instructions; select the legal reversal path scheme with the lowest physical disturbance score and output it to the accounting platform.

[0031] As described above, the contract settlement and automatic reconciliation system and method with deep integration of business and financial data provided by the present invention has at least the following beneficial effects:

[0032] This invention introduces a mechanism to compare accounting period identifiers with the current system closing time within the affected node cluster. Based on this, it automatically distinguishes between inter-period reversal instructions and original order reduction instructions. This allows the reversal strategy to no longer use static, fixed rules, but rather to adaptively select a processing path that better suits actual financial constraints based on the accounting period status. By generating alternative reversal path chains by combining the reverse connection relationship of the fund flow constraint edges, the system can simulate the deduction of amount and business volume values ​​in the virtual sandbox memory in advance, and perform legality verification on the updated state vector. This allows it to identify risky paths that may lead to amount penetration, business volume imbalance, and broken accounting relationships before formal execution. This simulation-then-confirmation processing method not only reduces the pressure of manual review but also significantly improves the security and controllability of abnormal settlement reversal processes, enabling the system to automatically correct errors in complex contract topologies.

[0033] This invention calculates physical disturbance scores for legitimate reversal path schemes, comprehensively evaluating the impact of the total number of nodes and the total number of inter-period reversal action instructions on the business network and financial compliance. It then selects the optimal path with the least disturbance from multiple feasible schemes and outputs it to the accounting platform to synchronize and update business and financial reconciliation data. Unlike existing technologies that only pursue successful amount rollback, this invention focuses more on the impact of reversal actions on the stability of the entire settlement network structure. It can minimize the impact on related nodes and the amount of modification to historical data links while ensuring accounting balance, thereby reducing system recalculation costs and subsequent audit risks. Through this mechanism, this invention achieves a closed-loop process of automatic location, planning, verification, and execution of abnormal settlements, significantly improving the accuracy, efficiency, and traceability of contract settlement error correction in complex business and financial scenarios. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a schematic diagram showing the connections between the steps of the method of the present invention.

[0036] Figure 2 This is a schematic diagram showing the connections of the various modules in the system of the present invention. Detailed Implementation

[0037] The following description, in conjunction with the implementation of this invention, is merely an example and illustration of the concept of this invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the inventive concept or exceed the scope defined in these claims, all of which should fall within the protection scope of this invention.

[0038] Example 1:

[0039] Please see Figure 1 As shown, the contract settlement and automatic reconciliation method that deeply integrates business and financial data includes:

[0040] Obtain business performance data sequences and financial settlement data sequences; map the business performance data sequences and financial settlement data sequences into a business-finance two-dimensional time-series directed graph; the nodes of the business-finance two-dimensional time-series directed graph contain state vectors with monetary values, business volume values ​​and accounting period identifiers, and the edges of the business-finance two-dimensional time-series directed graph contain fund flow constraint edges and business-finance mapping constraint edges.

[0041] Preferably, the business performance data sequence and the financial settlement data sequence are mapped into a two-dimensional time-series directed graph, including:

[0042] Based on the data generation time, the business performance data sequence and the financial settlement data sequence are time-series aligned to construct a time-series aligned data set;

[0043] For time-series aligned datasets, business volume values ​​are parsed from business performance data sequences, and monetary values ​​and accounting period identifiers are parsed from financial settlement data sequences.

[0044] The parsed business volume, amount, and accounting period identifiers are vectorized and aggregated to generate a state vector, and basic nodes are created by combining the state vector.

[0045] Extract the settlement document flow details from the financial settlement data sequence, and establish fund flow constraint edges between basic nodes based on the upstream and downstream fund allocation relationships in the settlement document flow details;

[0046] Extract business event transformation instructions from the business performance data sequence, and establish a business-finance mapping constraint edge by connecting the basic node that records business execution actions with the basic node that records settlement vouchers based on the business event transformation instructions.

[0047] By using the constraints of fund flow and business-finance mapping, the basic nodes carrying state vectors are connected in a global topology network to generate a two-dimensional time-series directed graph of business and finance.

[0048] In this embodiment of the invention, the system first acquires the business performance data sequence from the underlying business system and the financial settlement data sequence from the ERP system. Since actual business operations often involve time lags such as "delivery before invoicing" or "construction before pricing," the system performs time-series alignment processing on the two sets of sequences based on the data generation time, constructing a time-series aligned data set. This alignment process employs a dynamic time window matching mechanism, using the business occurrence time as a benchmark, and searching for corresponding financial settlement records within a preset valid reconciliation time window. Next, for the time-series aligned data set, the system parses the specific business volume values ​​(e.g., the volume of poured concrete in an engineering project or the tonnage shipped in the supply chain, with the units of measurement denoted as business entities, such as...) from the form fields of the business performance data sequence. or The system then parses the corresponding monetary values ​​and accounting period identifiers containing year and month information from the financial settlement data sequence. Subsequently, the parsed business volume values, monetary values, and accounting period identifiers are vectorized and aggregated to generate state vectors for each data entity. These state vectors are then used as node attribute parameters in memory to instantiate and create the basic nodes of the network topology.

[0049] Next, the system further extracts settlement document transaction details from the financial settlement data sequence. Based on the payment relationship between the payer and payee accounts recorded in the transaction details, it establishes fund flow constraint edges pointing to the fund recipient between different basic nodes. Following this, to address the long-standing issue of accounting traceability gaps caused by the separation of business and finance, the system extracts business event transformation instructions (such as the operation log of converting an acceptance form into a pricing form) from the business performance data sequence. Based on these transformation instructions, the system connects the basic nodes recording business execution actions with the basic nodes recording settlement vouchers across dimensions, thereby establishing business-finance mapping constraint edges. To ensure that the business-finance mapping constraint edges have accurate mathematical evaluation metrics to support subsequent sandbox accounting, the system simultaneously calculates the mapping transformation accounting amount when establishing the mapping edges. The calculation formula is set as follows:

[0050]

[0051] in, The accounting amount generated for the mapping; The benchmark unit price is as stipulated in the contract, and the unit of measurement is yuan per business entity unit. This represents the actual volume of fulfillment transactions as determined by the analysis. The actual time lag between business execution and financial accounting is obtained by subtracting the business execution time from the settlement time. The above variation introduces two correction factors: one is the time lag depreciation coefficient. Its dimensions are This is used to characterize the implicit cost of capital occupation or the inter-period discount rate caused by delays in fund settlement. The first parameter is the preset value, which is usually obtained by calculating the daily interest rate based on the company's weighted average cost of capital for the current period; the second is the quality depreciation ratio. , which is a dimensionless coefficient, is automatically calculated by extracting deduction items from the business acceptance form, and its value range is defined in [0,1]. Finally, the system uses fund flow constraint edges with definite fund transfer amounts and business-finance mapping constraint edges with accounting mapping characteristics to connect all basic nodes carrying state vectors in a global topology network, forming a business-finance two-dimensional time-series directed graph that includes both fund transfer paths and business performance fulfillment paths.

[0052] Receive abnormal settlement trigger instructions and extract the abnormal target nodes corresponding to the abnormal settlement trigger instructions;

[0053] Starting from the abnormal target node, a source traversal is performed in the opposite direction of the capital flow constraint edge in the business and finance two-dimensional time-series directed graph, and a penetration traversal is performed in the positive direction of the business and finance mapping constraint edge to extract the connected affected node clusters.

[0054] Preferably, the affected clusters of connected nodes include:

[0055] Record the physical memory address of the abnormal target node in the business and finance two-dimensional time-series directed graph;

[0056] Starting from the physical memory address, read the starting address of the constraint edge of the fund flow, and push the upstream node corresponding to the starting address into the fund traceability stack;

[0057] Starting from the physical memory address, read the endpoint address of the business mapping constraint edge, and push the business node corresponding to the endpoint address into the business penetration stack;

[0058] Compare the monetary values ​​in the state vectors of each node in the fund traceability stack and the business finance penetration stack with the monetary values ​​of the abnormal target node;

[0059] Nodes whose monetary value in the state vector is greater than a preset zero threshold are retained, while nodes whose monetary value in the state vector is equal to the preset zero threshold are removed. The retained nodes are then aggregated to form a cluster of affected nodes.

[0060] In this embodiment of the invention, when the system's reconciliation risk control engine or front-end financial personnel initiates a request to reverse an incorrect account, the system first receives an abnormal settlement trigger instruction through an interface and parses the corresponding settlement order number or business voucher number from the load data of the instruction. Based on this, it extracts the corresponding abnormal target node in the graph database. To ensure the efficiency and accuracy of the underlying graph traversal, the system records the physical memory address of the abnormal target node in the business-finance dual-dimensional time-series directed graph. Subsequently, the system starts a bidirectional breadth-first search with this physical memory address as the origin. In the direction of fund tracing, the system starts from the physical memory address, reads the starting address pointer of the fund flow constraint edge in reverse, and pushes the upstream node pointing to the starting address (i.e., the fund provider, such as the general contract or upstream invoice node) into the "fund tracing stack" pre-opened by the system. At the same time, in the direction of business penetration, the system reads the ending address in forward along the pointer of the business-finance mapping constraint edge and pushes the corresponding underlying business execution node (such as the associated warehouse entry order, acceptance bill of quantities node) into the "business-finance penetration stack".

[0061] After loading upstream and downstream related nodes into the memory stack, to prevent a single point of failure from indiscriminately locking the entire massive historical contract ledger, the system needs to compare the amount values ​​in the state vectors of each node in the stack with the original abnormal amount of the target node. The calculation formula is set as follows:

[0062]

[0063] This formula is a variation of the traditional full proportional allocation model. Traditional linear allocation often assumes that funds are transferred losslessly and without time delay within the contract network, which is highly unreasonable in the complex fund flows of actual general contracting and subcontracting. Therefore, this variant model introduces a two-dimensional attenuation mechanism involving both topological space and time. In the formula, This represents the actual amount of money affected by the abnormal target node affecting the node within the stack. The absolute value of the initial erroneous amount recorded in the state vector of the abnormal target node; The topology traversal depth of the current node from the abnormal target node (i.e., the number of constraint edge jumps between them, which is directly provided by the depth counter of the breadth-first search). This is a preset topology isolation attenuation coefficient, a dimensionless pure numerical value that represents the degree of business isolation generated each time funds pass through a layer of contract or order settlement. This is the time difference between the generation time of the current node in the stack and the generation time of the abnormal target node. Let be the time-sedimentation decay constant, with dimensions . It is used to simulate the financial inertia that, as time goes by, early accounts have been solidified (such as settled accounts) and are difficult to be shaken by new errors. It is usually taken as the reciprocal of the financial month-end cycle.

[0064] To determine the actual amount of impact for each pushed node The system then compares it with a preset zero threshold. Considering the precision overflow issue in computer floating-point arithmetic and the minimum accounting precision of the financial ledger, the preset zero threshold is not directly set to absolute 0, but rather recommended to be set to the smallest accounting unit of legal tender. The system determines the calculated value... Nodes exceeding a preset zero threshold are considered "substantially impacted financially" and are retained in memory; conversely, if the impact is less than a certain threshold due to network attenuation, resulting in... If the value is less than or equal to a preset zero threshold, the system determines that the node has escaped the affected radius and removes it from the current processing stack. Finally, the system uses all the high-value related nodes that have been filtered and retained to re-aggregate them in memory into a directed connected subgraph, which is the precisely generated cluster of affected nodes.

[0065] Enter the offsetting strategy generation loop, which includes:

[0066] Extract the accounting period identifier for each node in the affected node cluster; compare the accounting period identifier with the current system closing time.

[0067] If the accounting period identifier is earlier than the current system closing time, a cross-period reversal instruction with a negative amount will be generated; if the accounting period identifier is not earlier than the current system closing time, a deduction instruction for the original document will be generated.

[0068] Based on the inter-period reversal action instructions and the original order reduction action instructions, and combined with the reverse connection relationship of the fund flow constraint edge, an alternative reversal path chain is generated.

[0069] In this embodiment of the invention, the accounting period identifier encapsulated in the state vector of each node in the affected node cluster is extracted one by one from memory, and the real-time closing calendar of the ERP financial platform is retrieved to compare the accounting period identifier with the current system closing time. If the accounting period identifier is earlier than the current system closing time, it indicates that the corresponding accounting period has been physically fixed and cannot be directly overwritten. The system then generates an inter-period reversal action instruction carrying a negative amount value, which triggers the red-ink voucher offset logic.

[0070] If the accounting period identifier is not earlier than the current system closing time, it indicates that the accounting system is still open, and the system will directly generate the original document reduction action instruction for the current period's blue-ink deduction. When generating the inter-period reversal action instruction, in order to accurately quantify the time-occupancy loss of funds and implicit tax adjustment costs caused by crossing financial closing periods, the system has a built-in and runs a dynamic inter-period reversal amount calculation model; the calculation formula of this model is derived from a variation of the classic formula for calculating the future value of principal and interest, and the specific calculation formula is set as follows:

[0071]

[0072] In the formula, This represents the actual negative amount carried in the inter-period offsetting action instruction; Represents the absolute value of the base amount of pending exceptions actually allocated to the current node; This represents the absolute point in time when the system performs the reversal operation. This represents the closing time point mapped by the accounting period identifier of the current node; This represents the preset monthly financial benchmark cycle constant, whose physical value is fixed at 30 and its dimension is days. The pre-defined inter-period penalty adjustment factor for funds is a dimensionless pure coefficient. Its pre-defined logic is based on the square amplification effect of the monthly weighted average cost of capital of enterprises. It aims to suppress long-term accounting delays and compensate sunk costs from the algorithm level. Representing the tax damping coefficient, it is a dimensionless pure numerical value. Its value is dynamically assigned based on the current node's invoice authentication status, which is obtained through polling via the tax interface. The value is assigned when an invoice is verified by the tax authorities as having been authenticated and deducted. 1.06 is used to cover the implicit frictional resistance of input tax credit reversal when the invoice verification is in an unverified state. Take 1.00.

[0073] It should be added that, ( The time difference is obtained by subtracting the two, with the dimension of time difference being days, and the value of time difference is obtained by converting the current time point and the node time point.

[0074] It should be added that, The value is based on the company's annual weighted average cost of capital. Calculate the square: λ is the risk preference coefficient, with a default value of 1.0. If the company cannot provide WACC, λ can be directly configured in the range [0.0001, 0.001].

[0075] Finally, based on the calculated inter-period offsetting action instructions and the original single-entry reduction action instructions, and combined with the reverse direction pointer connection relationship of the fund flow constraint edge in the underlying memory stack, the system sequentially mounts and connects the action instructions with compensation and correction attributes to each associated node, thereby automatically assembling and generating a multi-level continuous alternative offsetting path chain.

[0076] The alternative reversal path chain is imported into the virtual sandbox memory and put into a validity check loop. The validity check loop includes:

[0077] According to the alternative offsetting path chain, the amount and business volume values ​​in the state vectors of each node in the affected node cluster are deducted from the virtual sandbox memory to obtain the updated state vectors.

[0078] Determine whether the monetary value in the updated state vector is lower than a preset lower threshold.

[0079] If the amount in the updated state vector is lower than the preset lower threshold, the backtracking mechanism is triggered, and the amount corresponding to the abnormal target node is split according to the preset splitting ratio and the reversal strategy generation loop is re-entered.

[0080] If the amount values ​​in the updated state vector are all not lower than the preset lower threshold, then the corresponding alternative offsetting path chain is extracted as a valid offsetting path scheme.

[0081] Preferably, the state vector also includes a tax lock identifier;

[0082] After extracting the accounting period identifier for each node in the affected node cluster, the reversal strategy generation loop also includes:

[0083] Extract the tax lock identifier for each node in the affected node cluster;

[0084] Determine if the tax lock icon indicates a tax filing status;

[0085] If the tax lock mark is in the "tax filed" status, the generation process of the original single reduction action instruction will be intercepted, and a cross-period reversal action instruction with a tax reversal mark will be forcibly generated.

[0086] If the tax lockout status is "not filed", the judgment logic of generating inter-period reversal instructions and original document reduction instructions based on the accounting period status will continue.

[0087] Preferably, the updated state vector is obtained by deducting the amount and transaction volume values ​​from the state vectors of each node in the affected node cluster according to the alternative offsetting path chain in the virtual sandbox memory, including:

[0088] Analyze the split amount of the amount to be offset carried in the alternative offsetting path chain;

[0089] Lock the current processing node within the affected node cluster corresponding to the split amount of the amount to be offset in the virtual sandbox memory;

[0090] Read the original amount value from the state vector of the current processing node;

[0091] Subtract the amount to be offset from the original amount to obtain the recalculated amount.

[0092] Based on the unit price conversion ratio carried in the business and finance mapping constraint edge, calculate the amount of business to be deducted that is equivalent to the amount to be offset.

[0093] Read the original business volume value from the state vector of the current processing node, and subtract the business volume to be reduced from the original business volume value to obtain the recalculated business volume value.

[0094] The updated state vector is generated by overwriting the original amount and the original business volume with the recalculated amount and the recalculated business volume.

[0095] Preferably, if the amount in the updated state vector is lower than a preset lower threshold, a backtracking mechanism is triggered, including:

[0096] Record the memory coordinates of the breakpoint node that caused the amount value to fall below the preset lower threshold.

[0097] Extract all upstream parallel nodes associated with the memory coordinates of the breakpoint node;

[0098] Read the total remaining available amount of all upstream parallel nodes;

[0099] The preset splitting ratio is set based on the percentage of the remaining available amount of each upstream parallel node relative to the total remaining available amount.

[0100] According to the preset splitting ratio, the monetary value corresponding to the abnormal target node is mathematically divided into multiple split monetary values.

[0101] The multiple split amount values ​​are allocated to the corresponding upstream parallel nodes and the program pointer is triggered to jump to the starting point of the offsetting strategy generation loop.

[0102] Preferably, after determining whether the amount value in the updated state vector is lower than a preset lower threshold, the validity check loop further includes:

[0103] If the amount values ​​in the updated state vector are not lower than the preset lower threshold, then extract the amount values ​​and business volume values ​​in the updated state vector.

[0104] Multiply the business volume value by the pre-determined contract unit price corresponding to the current processing node to obtain the theoretical recovery amount;

[0105] Compare the absolute value of the deviation between the theoretically restored amount and the actual amount;

[0106] If the absolute value of the deviation is higher than the preset tolerance threshold, it is determined that the consistency of business and financial mapping is broken, the current alternative reversal path chain is abandoned and the backtracking mechanism is triggered.

[0107] If the absolute value of the deviation is not higher than the preset tolerance threshold, then the step of extracting the corresponding alternative reversal path chain as a legal reversal path scheme is executed.

[0108] Preferably, after triggering the backtracking mechanism, the following also applies:

[0109] Count the cumulative number of loops that trigger the backtracking mechanism;

[0110] Determine if the cumulative number of loop iterations has reached the preset dead loop interception limit;

[0111] If the cumulative number of loops reaches the preset dead loop interception limit, the validity check loop will be terminated.

[0112] Extract the memory coordinates of the breakpoint node at the latest trigger of the backtracking mechanism and the corresponding alternative reversal path chain;

[0113] Mark the memory coordinates of the breakpoint node as a historical debt deadlock node, generate a business intervention instruction chain containing the historical debt deadlock node and the corresponding alternative write-off path chain, and output it to the terminal interface.

[0114] In this embodiment of the invention, before entering the core sandbox simulation, the system first performs a tax-based pre-interception to ensure the compliance of the reversal strategy. The system extracts the tax lock identifier of each node in the affected node cluster and determines whether the tax lock identifier is in a tax-reported state. If it is determined to be in a tax-reported state, it means that the underlying accounts have been solidified, and the system then intercepts the generation process of the original document reduction action instruction, forcibly generating an inter-period reversal action instruction with a tax reversal red mark; if it is determined to be in a tax-unreported state, the normal logic of generating the corresponding instruction based on the accounting period identifier is maintained.

[0115] After completing compliance interception, the system imports the alternative offsetting path chain into a separately allocated virtual sandbox memory for physical-level deduction simulation. The system parses the amount to be offset carried in the alternative offsetting path chain and uses pointer addressing in the virtual sandbox memory to lock the current processing node within the affected node cluster corresponding to that amount. Next, the system reads the original amount value from the state vector of the current processing node, subtracts the amount to be offset from the original amount value, and calculates the recalculated amount value. In order to ensure that the underlying business execution volume is synchronized and matched while deducting funds, the system needs to calculate the amount of business to be deducted that is equivalent to the amount to be offset based on the unit price conversion ratio carried in the business-finance mapping constraint edge. The specific calculation formula is set as follows:

[0116]

[0117] In the formula, This represents the calculated amount of business to be reduced, measured in business entity units. This represents the amount to be split into separate portions for the amount to be offset. The unit price of the pre-determined contract recorded in the business-finance mapping constraint edge is expressed in yuan per business entity unit. Represents the original monetary value of the current processing node; This represents the preset penalty coefficient for returns, a dimensionless pure numerical value. Its preset logic is based on industry-standard return penalty ratios or tiered fee rate volatility, with a recommended value range of 0.02 to 0.05. The system reads the original business volume value from the current processing node's state vector, subtracts the calculated business volume to be reduced to obtain the recalculated business volume value, and uses the recalculated amount value and the recalculated business volume value to overwrite the original data, generating an updated state vector.

[0118] After completing the single-point simulation, the system immediately determines whether the amount in the updated state vector is lower than a preset lower threshold. This threshold is typically set to 0 yuan to absolutely prevent negative asset penetration in the financial accounts. If the amount is lower than the preset lower threshold, it indicates that the available balance of the current node is insufficient to withstand the write-off, and the system immediately triggers a backtracking mechanism. The system records the memory coordinates of the breakpoint node that caused the write-off, traces all upstream parallel nodes associated with those coordinates, and reads the total remaining available amount of each upstream parallel node. Based on the percentage of remaining available amount of each upstream parallel node relative to the total, the system sets a preset splitting ratio. According to this ratio, the initial amount of the abnormal target node is split into multiple parts, which are then allocated to the corresponding upstream parallel nodes. Simultaneously, the program pointer jumps to the starting point of the write-off strategy generation loop to replan the path. If the updated amount is not lower than the preset lower threshold, the system further extracts the updated amount and the business volume, multiplies the business volume by the predetermined contract unit price to obtain the theoretical restored amount, and compares the absolute value of the deviation between the theoretical restored amount and the recalculated amount. If the absolute value of the deviation is higher than the preset tolerance threshold (to be compatible with computer floating-point precision and currency rounding, the preset tolerance threshold is recommended to be set to 0.05 yuan), the system determines that the consistency of business and financial mapping has been broken, and then abandons the current alternative reversal path chain and triggers the backtracking mechanism; if it is not higher than the preset tolerance threshold, the path is extracted as a valid reversal path scheme.

[0119] To prevent system computing power exhaustion due to insufficient balances across all nodes in an extreme historical bad debt network, the system automatically counts the cumulative loop count after each trigger of the backtracking mechanism and determines whether the cumulative loop count has reached the preset deadlock interception limit. Once the preset deadlock interception limit is reached, the system immediately terminates the legality verification loop, extracts the memory coordinates of the latest breakpoint node, marks it as a historical debt deadlock node, and finally generates a business intervention instruction chain containing the deadlock breakpoint and unfinished paths, outputting it to the terminal interface to prompt human financial experts to intervene and handle the situation accordingly.

[0120] The physical disturbance score is calculated based on the total number of nodes in the legal reversal path scheme and the total number of inter-period reversal action instructions; the legal reversal path scheme with the lowest physical disturbance score is selected and output to the accounting platform for synchronous update of business and financial reconciliation data.

[0121] Preferably, the physical disturbance score is calculated based on the total number of nodes included in the legal offsetting path scheme and the total number of inter-period offsetting action instructions, including:

[0122] Analyze the valid reversal path scheme to obtain the total number of nodes involved in the modification;

[0123] Extract the total number of inter-period reversal action instructions included in the legal reversal path scheme;

[0124] Configure a first preset penalty weight for the total number of nodes, and configure a second preset penalty weight for the total number of inter-period reversal action instructions;

[0125] The topological turbulence index is obtained by multiplying the total number of nodes by the first preset penalty weight;

[0126] The financial compliance cost index is obtained by multiplying the total number of inter-period write-off instructions by the second preset penalty weight;

[0127] The physical disturbance score is obtained by summing the topological turbulence index and the financial compliance cost index.

[0128] In this embodiment of the invention, after the virtual sandbox parallel simulation ends, multiple legal reversal path schemes are retained due to different amount splitting ratios and backtracking path branches. To minimize the impact on the enterprise's real business network and underlying database while mitigating accounting gaps, the system needs to calculate a physical disturbance score based on the total number of nodes and the total number of inter-period reversal action instructions in each legal reversal path scheme. First, the system analyzes the memory topology of each legal reversal path scheme, obtaining the total number of nodes involved in state updates and modifications by traversing a counter. This value reflects the breadth of the reversal action's impact across the entire business chain. Simultaneously, the system extracts the total number of inter-period reversal action instructions included in the legal reversal path scheme, representing how many red-ink reversal vouchers requiring tax regulatory intervention will be generated by the scheme. Next, the system configures a first preset penalty weight for the total number of nodes and a second preset penalty weight for the total number of inter-period reversal action instructions. In traditional linear evaluation models, weights are often fixed constants, which cannot reflect the exponential deadlock risk to the database caused by large-scale concurrent modifications. Therefore, this embodiment constructs a nonlinear adaptive weight allocation algorithm based on a variation of the system steady-state entropy model in information theory. Specifically, the first preset penalty weight calculation formula is as follows:

[0129]

[0130] The formula for calculating the second preset penalty weight is:

[0131]

[0132] In the formula, This represents the total number of nodes involved in the modification. Represents the fundamental topological penalty constant, with dimensions of Its default logic is to characterize the average overhead of a single accounting modification on the concurrent performance of the underlying data, with a recommended value range of 0.05 to 0.10; This represents the expected maximum node capacity constant in the graph network, which is usually extracted directly from the maximum number of database connections configured in an enterprise ERP system. Represents the fundamental intertemporal penalty constant, with dimensions of ; The regional tax audit risk elasticity coefficient is a dimensionless pure numerical value, which is dynamically input based on the frequency of audits.

[0133] The system will count the total number of nodes With dynamically generated first preset penalty weight Multiplying them yields the topological turbulence index; similarly, the total number of inter-period offsetting orders is calculated. With the second preset penalty weight Multiplying them yields the financial compliance cost index; finally, the topological turbulence index and the financial compliance cost index are summed to obtain the physical disturbance score.

[0134] After calculating the scores of all candidate schemes, the system selects the legal cancellation path scheme with the lowest physical disturbance score through global sorting and comparison.

[0135] Example 2:

[0136] like Figure 2 As shown, the contract settlement and automatic reconciliation system deeply integrates business and financial data. The system includes a graph construction module, an anomaly receiving module, an affected node extraction module, an offsetting strategy generation module, a sand table verification module, and a scoring output module.

[0137] The various modules are connected via wired and / or wireless connections to enable data transmission between them;

[0138] The graph construction module is used to generate a business-finance two-dimensional time-series directed graph based on the business performance data sequence and the financial settlement data sequence. The nodes of the business-finance two-dimensional time-series directed graph carry a state vector consisting of a monetary value, a business volume value, and an accounting period identifier. The edges of the business-finance two-dimensional time-series directed graph include cash flow constraint edges and business-finance mapping constraint edges.

[0139] An anomaly receiving module is used to receive an anomaly settlement trigger instruction and extract the anomaly target node corresponding to the anomaly settlement trigger instruction;

[0140] The affected node extraction module is used to perform source traversal in the business and finance two-dimensional time-series directed graph along the opposite direction of the capital flow constraint edge, starting from the abnormal target node, and to perform penetration traversal along the positive direction of the business and finance mapping constraint edge, and to extract the connected clusters of affected nodes.

[0141] The reversal strategy generation module is used to enter the reversal strategy generation loop, extract the accounting period identifier of each node in the affected node cluster, and compare the accounting period identifier with the current system closing time. When the accounting period identifier is earlier than the current system closing time, a cross-period reversal action instruction with a negative amount is generated. When the accounting period identifier is not earlier than the current system closing time, an original order reduction action instruction is generated. Based on the cross-period reversal action instruction and the original order reduction action instruction, and combined with the reverse connection relationship of the fund flow constraint edge, an alternative reversal path chain is generated.

[0142] The sandbox verification module is used to import the alternative offsetting path chain into the virtual sandbox memory and enter a legality verification loop. According to the alternative offsetting path chain, it deducts the amount value and business volume value of each node in the affected node cluster from the state vector in the virtual sandbox memory to obtain the updated state vector. It determines whether the amount value in the updated state vector is lower than the preset lower threshold. When the amount value in the updated state vector is lower than the preset lower threshold, a backtracking mechanism is triggered. When the amount value in the updated state vector is not lower than the preset lower threshold, the corresponding alternative offsetting path chain is extracted as the legal offsetting path scheme.

[0143] The scoring output module is used to calculate the physical disturbance score based on the total number of nodes in the legal reversal path scheme and the total number of inter-period reversal action instructions; select the legal reversal path scheme with the lowest physical disturbance score and output it to the accounting platform.

[0144] It should be noted that the interval and threshold sizes are set for ease of comparison. The size of the threshold depends on the amount of sample data and the base number set by those skilled in the art for each set of sample data, as long as it does not affect the proportional relationship between the parameter and the quantized value. Furthermore, the above formulas are all dimensionless calculations, and the formulas are derived from software simulations using a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0145] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0146] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0147] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for contract settlement and automatic reconciliation that deeply integrates business and financial data, characterized in that: include: Based on the business performance data sequence and the financial settlement data sequence, a two-dimensional time-series directed graph of business and finance is generated. Receive abnormal settlement trigger instructions and extract the abnormal target nodes corresponding to the abnormal settlement trigger instructions; Starting from the abnormal target node, perform a source traversal along the opposite direction of the capital flow constraint edge in the business and finance two-dimensional time-series directed graph, and perform a penetration traversal along the positive direction of the business and finance mapping constraint edge to extract the connected affected node clusters. Enter the offsetting strategy generation loop, including: Extract the accounting period identifier for each node in the affected node cluster; compare the accounting period identifier with the current system closing time. If the accounting period identifier is earlier than the current system closing time, a cross-period reversal action instruction with a negative amount value is generated; otherwise, an original order deduction action instruction is generated. Based on the inter-period reversal action instructions and the original order reduction action instructions, and combined with the reverse connection relationship of the fund flow constraint edge, an alternative reversal path chain is generated. Import the alternative cancellation path chain into the virtual sandbox memory and enter a validity verification loop, including: According to the alternative offsetting path chain, the amount and business volume values ​​in the state vectors of each node in the affected node cluster are deducted from the virtual sandbox memory to obtain the updated state vectors. Determine whether the amount in the updated state vector is lower than the preset lower threshold; if yes, trigger the backtracking mechanism; if no, extract the corresponding alternative reversal path chain as the legal reversal path scheme. The physical disturbance score is calculated based on the total number of nodes in the legal offsetting path scheme and the total number of cross-period offsetting action instructions. The legal reversal path scheme with the lowest physical disturbance score is selected and output to the accounting platform for synchronous update of business and financial reconciliation data.

2. The contract settlement and automatic reconciliation method based on deep integration of business and financial data according to claim 1, characterized in that, Mapping the business performance data sequence and the financial settlement data sequence into a two-dimensional time-series directed graph, including: Based on the data generation time, the business performance data sequence and the financial settlement data sequence are time-series aligned to construct a time-series aligned data set; For time-series aligned datasets, business volume values ​​are parsed from business performance data sequences, and monetary values ​​and accounting period identifiers are parsed from financial settlement data sequences. The parsed business volume, amount, and accounting period identifiers are vectorized and aggregated to generate a state vector, and basic nodes are created by combining the state vector. Extract the settlement document flow details from the financial settlement data sequence, and establish fund flow constraint edges between basic nodes based on the upstream and downstream fund allocation relationships in the settlement document flow details; Extract business event transformation instructions from the business performance data sequence, and establish a business-finance mapping constraint edge by connecting the basic node that records business execution actions with the basic node that records settlement vouchers based on the business event transformation instructions. By using the constraints of fund flow and business-finance mapping, the basic nodes carrying state vectors are connected in a global topology network to generate a two-dimensional time-series directed graph of business and finance.

3. The contract settlement and automatic reconciliation method based on deep integration of business and financial data according to claim 1, characterized in that, The affected cluster of connected nodes is extracted, including: Record the physical memory address of the abnormal target node in the business and finance two-dimensional time-series directed graph; Starting from the physical memory address, read the starting address of the constraint edge of the fund flow, and push the upstream node corresponding to the starting address into the fund traceability stack; Starting from the physical memory address, read the endpoint address of the business mapping constraint edge, and push the business node corresponding to the endpoint address into the business penetration stack; Compare the monetary values ​​in the state vectors of each node in the fund traceability stack and the business finance penetration stack with the monetary values ​​of the abnormal target node; Nodes whose monetary value in the state vector is greater than a preset zero threshold are retained, while nodes whose monetary value in the state vector is equal to the preset zero threshold are removed. The retained nodes are then aggregated to form a cluster of affected nodes.

4. The contract settlement and automatic reconciliation method based on deep integration of business and financial data according to claim 1, characterized in that, The state vector also contains a tax lock identifier; After extracting the accounting period identifier for each node in the affected node cluster, the reversal strategy generation loop also includes: Extract the tax lock identifier for each node in the affected node cluster; Determine if the tax lock icon indicates that taxes have been filed; If the tax lock mark is in the "tax filed" status, the generation process of the original single reduction action instruction will be intercepted, and a cross-period reversal action instruction with a tax reversal mark will be forcibly generated. If the tax lockout status is "not filed", the judgment logic of generating inter-period reversal instructions and original document reduction instructions based on the accounting period status will continue.

5. The contract settlement and automatic reconciliation method based on deep integration of business and financial data according to claim 1, characterized in that, According to the alternative offsetting path chain, the amount and business volume values ​​in the state vectors of each node in the affected node cluster are deducted from the virtual sandbox memory to obtain the updated state vectors, including: Analyze the split amount of the amount to be offset carried in the alternative offsetting path chain; Lock the current processing node within the affected node cluster corresponding to the split amount of the amount to be offset in the virtual sandbox memory; Read the original amount value from the state vector of the current processing node; Subtract the amount to be offset from the original amount to obtain the recalculated amount. Based on the unit price conversion ratio carried in the business and finance mapping constraint edge, calculate the amount of business to be deducted that is equivalent to the amount to be offset. Read the original business volume value from the state vector of the current processing node, and subtract the business volume to be reduced from the original business volume value to obtain the recalculated business volume value. The updated state vector is generated by overwriting the original amount and the original business volume with the recalculated amount and the recalculated business volume.

6. The contract settlement and automatic reconciliation method based on deep integration of business and financial data according to claim 1, characterized in that, If the amount in the updated state vector is lower than a preset lower threshold, a backtracking mechanism is triggered, including: Record the memory coordinates of the breakpoint node that caused the amount value to fall below the preset lower threshold. Extract all upstream parallel nodes associated with the memory coordinates of the breakpoint node; Read the total remaining available amount of all upstream parallel nodes; The preset splitting ratio is set based on the percentage of the remaining available amount of each upstream parallel node relative to the total remaining available amount. According to the preset splitting ratio, the monetary value corresponding to the abnormal target node is mathematically divided into multiple split monetary values. The multiple split amount values ​​are allocated to the corresponding upstream parallel nodes and the program pointer is triggered to jump to the starting point of the offsetting strategy generation loop.

7. The contract settlement and automatic reconciliation method based on deep integration of business and financial data according to claim 1, characterized in that, After determining whether the amount in the updated state vector is lower than the preset lower threshold, the validity check loop also includes: If the amount values ​​in the updated state vector are not lower than the preset lower threshold, then extract the amount values ​​and business volume values ​​in the updated state vector. Multiply the business volume value by the pre-determined contract unit price corresponding to the current processing node to obtain the theoretical recovery amount; Compare the absolute value of the deviation between the theoretically restored amount and the actual amount; If the absolute value of the deviation is higher than the preset tolerance threshold, it is determined that the consistency of business and financial mapping is broken, the current alternative reversal path chain is abandoned and the backtracking mechanism is triggered. If the absolute value of the deviation is not higher than the preset tolerance threshold, then the step of extracting the corresponding alternative reversal path chain as a legal reversal path scheme is executed.

8. The contract settlement and automatic reconciliation method based on deep integration of business and financial data according to claim 1, characterized in that, After the backtracking mechanism is triggered, it also includes: Count the cumulative number of loops that trigger the backtracking mechanism; Determine if the cumulative number of loop iterations has reached the preset dead loop interception limit; If the cumulative number of loops reaches the preset dead loop interception limit, the validity check loop will be terminated. Extract the memory coordinates of the breakpoint node at the latest trigger of the backtracking mechanism and the corresponding alternative reversal path chain; Mark the memory coordinates of the breakpoint node as a historical debt deadlock node, generate a business intervention instruction chain containing the historical debt deadlock node and the corresponding alternative write-off path chain, and output it to the terminal interface.

9. The contract settlement and automatic reconciliation method based on deep integration of business and financial data according to claim 1, characterized in that, The physical disturbance score is calculated based on the total number of nodes in the legitimate offsetting path scheme and the total number of inter-period offsetting action instructions, including: Analyze the valid reversal path scheme to obtain the total number of nodes involved in the modification; Extract the total number of inter-period reversal action instructions included in the legal reversal path scheme; Configure a first preset penalty weight for the total number of nodes, and configure a second preset penalty weight for the total number of inter-period reversal action instructions; The topological turbulence index is obtained by multiplying the total number of nodes by the first preset penalty weight; The financial compliance cost index is obtained by multiplying the total number of inter-period write-off instructions by the second preset penalty weight; The physical disturbance score is obtained by summing the topological turbulence index and the financial compliance cost index.

10. A contract settlement and automatic reconciliation system that deeply integrates business and financial data, characterized in that: It is based on the contract settlement and automatic reconciliation method of deep integration of business and financial data as described in any one of claims 1-9, and includes a graph construction module, an anomaly receiving module, an affected node extraction module, a reversal strategy generation module, a sand table verification module, and a scoring output module.

Images