Visual display method for multi-dimensional incidence relation of financial statements

Abstract

The invention relates to the technical field of financial data analysis, in particular to a financial statement multi-dimensional incidence relation visual display method which comprises the following steps: extracting subject codes, amount values and business dimension labels, and generating a standardized financial data set with timestamps; generating an initial edge set based on a preset subject association rule base, endowing each edge with a weight value reflecting the sensitivity of a service scene through a dynamic weight calculation module, and forming a multi-dimensional association network with a weight label; identifying a cross-level association path through a path traversal algorithm, and generating a recessive relationship set including a path intensity index; and dynamically and visually displaying. According to the method, the fusion expression capability of the complex financial flow is enhanced, and a clear data basic support is provided for subsequent multi-dimensional modeling and analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of financial data analysis, and in particular to a method for visually displaying multi-dimensional association relationships in financial statements. Background Art

[0002] With the continuous improvement of the enterprise informatization level, financial data is gradually showing the characteristics of high frequency, multi-source, and heterogeneity. Data from multiple business systems such as enterprise resource planning systems (ERP), supply chain management platforms, and tax declaration systems often have significant differences in structure, granularity, and semantic standards, making it difficult for traditional financial analysis methods to achieve unified processing and association modeling. In addition, the relationships between various items in financial statements not only include explicit checking relationships based on accounting standards but also implicitly contain a large number of potential logical connections based on business dimension labels (such as supply chain levels, fund uses, transaction risk levels, etc.).

[0003] In the prior art, the display of financial relationships mostly presents in the form of static charts, preset templates, or two-dimensional matrices, which are difficult to reveal the high-dimensional path-based association structures hidden behind multi-layer business structures and cross-cycle fund flows. At the same time, the lack of a comprehensive quantification mechanism for path strength, scenario sensitivity, and propagation characteristics also results in insufficient ability to identify potential risk chains and abnormal flows. Summary of the Invention

[0004] The present invention provides a method for visually displaying multi-dimensional association relationships in financial statements, which integrates multi-source financial data, mines explicit and implicit item relationships, and supports an interactive and highly interpretable dynamic visualization display method to assist financial auditing, internal control management, and strategic decision-making.

[0005] A method for visually displaying multi-dimensional association relationships in financial statements includes the following steps: S1. Standardization processing of financial data: Connect to multi-source financial systems, extract item codes, amount values, and business dimension labels, and generate a set of standardized financial data with timestamps; S2. Construction of a multi-dimensional association network: Using the set of standardized financial data as input, generate an initial edge set based on a preset item association rule library, and assign a weight value reflecting business scenario sensitivity to each edge through a dynamic weight calculation module to form a multi-dimensional association network with weight labels; S3. Mining of implicit association relationships: Input the multi-dimensional association network into an implicit association analysis model based on a graph attention mechanism, identify cross-level association paths through a path traversal algorithm, and generate a set of implicit relationships including path strength indicators; S4. Dynamic visualization display: According to the path strength indicators in the set of implicit relationships, call a parameterized layout engine to generate a dynamic visualization interface, and support triggering the focus diffusion and dimension folding of the association network through interactive operations.

[0006] Optionally, the S1 specifically includes: S11, connecting ERP, supply chain management system and tax declaration platform through heterogeneous data interface adapter, extracting original financial data stream, wherein the original financial data stream includes account code, amount value, business occurrence time and source system identifier; S12, cross-system consistency processing of subject codes: detect the differences in code naming in different source systems, perform code correction through the preset industry standard subject tree, and generate global subject codes in a unified format; S13, amount attribute standardization: convert the multi-currency amount into the base currency according to the official exchange rate at the time of the transaction, and attach the currency label and exchange rate snapshot information; S14, business dimension tag binding: according to a preset dimension mapping rule library, the unstructured business description field in the source system is converted into a standardized business dimension tag set, wherein the business dimension tag set includes transaction type, related party level, and fund use classification; S15, timestamp generation and data integration: add data collection timestamp and business timestamp to each data record, and finally output a standardized financial data set with timestamp.

[0007] Optionally, the S2 specifically includes: S21, initial edge set generation: based on the preset account association rule library, traverse the account codes and business dimension labels in the standardized financial data set, perform logical association, and generate explicit and implicit associated edges; S22, dynamic weight calculation: calculate the composite weight value for each associated edge through the dynamic weight calculation module. The composite weight value specifically includes basic weight calculation, scene sensitivity correction and time attenuation compensation; S23, network structure optimization: prune the associated edges whose composite weight values are lower than the weight threshold, merge repeated associated paths, and generate a multi-dimensional associated network. The multi-dimensional associated network includes features such as node attributes, associated edge attributes and network metadata.

[0008] Optionally, in S21: Explicit association edges are generated according to accounting standards: automatic connections are established for pairs of accounts that have direct debit / credit relationships or accounting cross-checking relationships; Implicit association edges are generated based on business dimension labels: cross-dimensional connections are generated for account pairs that share the same supply chain level label, risk level label, or fund usage classification label.

[0009] Optionally, the basic weight calculation includes assigning an initial weight based on the association type. The basic weight of the explicit association edge is the preset accounting standard weight value, and the basic weight of the implicit association edge increases exponentially according to the number of shared tags. The scenario sensitivity correction includes invoking the scenario parameters in the industry feature template to strengthen the weights of the association edges involved in sensitive services. Sensitive services include the capital flow path of the subjects related to supply chain finance; the association relationship of the cross-year deferred subjects, and the fund transfer between the subjects with a high-risk level label of high risk. The time decay compensation is to reduce the weight value of the association edge whose business timestamp exceeds the preset period according to the logarithmic function.

[0010] Optionally, the node attributes include inheriting the subject code, amount attribute, and business dimension label in the standardized financial data; The association edge attributes include storing the dynamic weight value, the association type identifier, and the original parameter snapshot of the weight calculation basis; The network metadata includes recording the version number of the industry feature template and the weight calculation timestamp.

[0011] Optionally, the S3 specifically includes: S31, Graph Attention Feature Fusion: Input the node attributes and edge attributes of the multi-dimensional association network into the multi-head graph attention network to generate the node feature vector that fuses the business semantics; construct an attention mask matrix based on the edge dynamic weights to suppress the influence of the association edges with weight values lower than the preset threshold on the feature propagation; S32, Cross-level Path Detection: Execute a hybrid search strategy of breadth-first traversal and depth-first traversal to detect the paths that meet the double constraint conditions, and perform dynamic pruning on the detected paths: remove the duplicate sub-paths and the redundant paths with a node coincidence degree higher than 80%. The double constraint conditions include: The path span ≥ 3 business levels; Include at least one implicit association edge triggered by the business dimension label; S33, Path Strength Quantification: Calculate the path basic strength value: , where represents an association path, which consists of several nodes and edges, represents an edge in the path , represents the dynamic weight value of the edge , represents the attention coefficient of the edge , is the length of the path , is the path length decay factor, represents the basic strength value of the path ; Apply business scenario correction factor: If the path contains supply chain finance label nodes, multiply the strength value by 1.5; if the path spans accounting years, multiply the strength value by 1 / (1 + log(annual span)); S34. Perform Monte Carlo robustness tests on the paths with the top 15% of strength values, including randomly perturbing node features 1000 times and retaining stable paths with strength volatility < 5%; Generate a set of implicit relationships: Aggregate the stable paths to generate a set of implicit association relationships. Each association record in the set of implicit association relationships includes the path node sequence, its business dimension label chain, the final strength index, and the critical impact factor marker.

[0012] Optionally, the S4 specifically includes: S41. Dynamic weight mapping: Normalize the path strength indicators in the set of implicit relationships to interval values in [0,1], establish a strength-visual parameter mapping rule. The strength-visual parameter mapping rule includes that the node size is positively correlated with the strength of the path start point, the edge transparency increases linearly with the path strength value, and the node color saturation reflects the heat value of the business dimension label to which it belongs; Generate an initial topological layout based on the strength-visual parameter mapping rule, where the core nodes are automatically positioned in the focus area of the canvas; S42. Focus diffusion trigger: In response to the interaction event of the user selecting a target node, perform real-time layout reconstruction, including taking the user-selected target node as the center, expanding three layers of associated nodes in descending order of path strength, with the layer spacing inversely proportional to the path strength value, applying a magnetic attraction effect to the nodes within the diffusion range to make strongly associated nodes automatically gather towards the center; Suspending and displaying the summaries of the top 5 cross-level paths in terms of strength at the edge of the diffusion area; S43. Dimension folding control: Build an interactive dimension filtering panel and respond to the following operations: When checking / unchecking business dimension labels, hide or display the nodes and associated edges of the corresponding labels in real time; Slide the strength threshold bar to dynamically fold the path edges with strength values lower than the current threshold; Apply a Gaussian blur effect to the folded elements and retain a 0.5-second fade-out trajectory; S44. Multi-view collaborative update: Includes a global view, a local view, and a linkage mechanism.

[0013] Optionally, the global view is to retain the core skeleton network composed of high-strength paths for users to grasp the overall topological structure; The local view is to synchronously update all visible nodes and paths of the current focus node and its diffusion area; The linkage mechanism is defined as: When a certain path in the local view is marked as abnormal, automatically highlight its start and end nodes in the global view and thicken the edge along the path .

[0014] Advantages of the present invention: In the present invention, by constructing a heterogeneous data interface adapter and a standardized processing flow, semantic unification and format specification of financial data in multi-source systems are realized, operations such as cross-system mapping of account codes, currency amount conversion, and dimension label extraction are supported, the fusion expression ability of complex financial flows is enhanced, and a clear data basis support is provided for subsequent multi-dimensional modeling and analysis.

[0015] In the present invention, by introducing a graph attention mechanism to construct a multi-dimensional financial association network, and integrating dynamic weights, semantic labels, and time features, it supports path-level identification and intensity evaluation of implicit financial relationships hidden in cross-level business structures. Especially, a business scenario correction factor and an annual span adjustment mechanism are incorporated into the path intensity quantification, which improves the expression accuracy and analysis interpretability of financial risk propagation paths.

[0016] In the present invention, by dynamically mapping path intensity to node size, edge transparency, and label saturation, and combining mechanisms such as focus diffusion, dimension folding, and multi-view linkage, an interactive display environment supporting real-time exploration and hierarchical deduction is constructed. This mechanism can automatically focus on key nodes, significant paths, and abnormal relationship chains, enhancing data interpretability and operation flexibility in the processes of audit analysis, risk tracing, and strategic planning. Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only those of the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative efforts.

[0018] Figure 1 It is a schematic flowchart of the method according to an embodiment of the present invention. Detailed Embodiments

[0019] The present invention will be described in detail below in conjunction with the drawings and specific embodiments. At the same time, it should be noted here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments. For some well-known technologies, those skilled in the art can also adopt other alternative methods for implementation; and the drawing part is only for more specifically describing the embodiments, and is not intended to specifically limit the present invention.

[0020] It should be noted that in the specification, the mention of "an embodiment", "embodiment", "exemplary embodiment", "some embodiments", etc. indicates that the described embodiments may include specific features, structures or characteristics, but not necessarily every embodiment includes such specific features, structures or characteristics. Additionally, when combining an embodiment to describe a specific feature, structure or characteristic, implementing such feature, structure or characteristic in combination with other embodiments (whether explicitly described or not) should be within the knowledge of those skilled in the relevant art.

[0021] Generally, terms can be understood at least in part from their use in context. For example, at least in part depending on the context, the term "one or more" as used herein can be used to describe any feature, structure or characteristic in a singular sense, or can be used to describe a combination of features, structures or characteristics in a plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey a set of exclusive factors, but rather can alternatively, at least in part depending on the context, allow for the existence of other factors that are not necessarily explicitly described.

[0022] As Figure 1 shown, a method for visualizing multi-dimensional association relationships in financial statements includes the following steps: S1. Standardization processing of financial data: Connect to multi-source financial systems, extract account codes, amount values, and business dimension labels, and generate a set of standardized financial data with timestamps. S2. Construction of a multi-dimensional association network: Using the set of standardized financial data as input, generate an initial edge set based on a preset account association rule library, and assign a weight value reflecting the business scenario sensitivity to each edge through a dynamic weight calculation module to form a multi-dimensional association network with weight labels. S3. Mining of implicit association relationships: Input the multi-dimensional association network into an implicit association analysis model based on a graph attention mechanism, identify cross-level association paths through a path traversal algorithm, and generate a set of implicit relationships including path strength indicators. S4. Dynamic visualization display: According to the path strength indicators in the set of implicit relationships, call a parameterized layout engine to generate a dynamic visualization interface, supporting the triggering of focus diffusion and dimension folding of the association network through interactive operations.

[0023] S1 specifically includes: S11. Connect to the ERP, supply chain management system, and tax declaration platform through a heterogeneous data interface adapter, and extract the original financial data stream, where the original financial data stream includes fields , where represents the set of original financial data streams, represents the account code of the th record, represents the amount value of the th record, Indicates the business occurrence time, Indicates the source system identifier, Indicates the total number of original data records; S12. Perform cross - system consistency processing on the subject code. The steps include constructing an industry - standard subject tree ; Define the coding mapping function , and implement the conversion from the original code to the standard code: , where, Indicates the globally standardized subject code, Indicates the mapping function based on the system identifier and the standard subject tree, Indicates the preset industry - standard subject tree; S13. Standardize the amount attribute: For amount data involving multiple currencies, convert it to the home - currency amount according to the exchange rate at the time of business occurrence and attach the currency label and the exchange - rate snapshot , where, Indicates the home - currency amount after exchange - rate conversion, Indicates the exchange rate at the business - time point of the th record, Indicates the original currency, Indicates the exchange - rate snapshot, Indicates the standardized amount record unit; S14. Bind business - dimension labels: Through the business - dimension mapping function , map the unstructured description field to the standardized business - dimension label set: , where, Indicates the unstructured description text in the th record, Indicates the set of business - dimension labels, and the set form includes transaction type, related - party level, and fund - use classification. Its expression is = , Indicates the th dimension label in the th record, Indicates the number of dimension labels in a single record, Indicates the semantic mapping function based on the preset rule library. S15. Generate time stamps and integrate data: Define the data - collection time stamp

[0024] , Indicates the data - collection time stamp of the th record; Retain the business occurrence time ​As the business timestamp , the final standardized data record format is , indicating the business timestamp of the th record (i.e., the original business occurrence time ), indicating the th standardized financial data record.

[0025] S2 specifically includes: S21, Initial edge set generation: Traverse the standardized financial data set , indicating the standardized financial data set, extract the account codes and the business dimension label set , and perform the following association logic for any two records : Explicit association edge generation condition: If , then establish an explicit association edge ; Implicit association edge generation condition: If , then establish an implicit association edge ; Among them, indicates the data record index, indicates the explicit association edge (based on debit / credit or reconciliation relationship), indicates the implicit association edge (based on shared labels), indicates the initial edge set (unweighted or unfiltered), indicates the set of explicit accounting association rules, indicates the set of predefined business label types available for implicit association; S22, Dynamic weight calculation: For each edge , calculate the composite weight , indicates the edge connecting nodes and , indicates the final weighted value (composite weight value) of the edge , indicates the base weight (initial weight value) of the edge, indicates the scenario sensitivity correction coefficient, indicates the time decay compensation coefficient; (1) Base weight : The explicit edge is ; The implicit edge (the number of shared labels is ) is ; indicates the weight lookup table function for the explicit edge, Represents the weight base value of the implicit edge, Represents the number of business tags shared among the th records, Represents the exponential coefficient that controls the growth of the implicit edge tags; (2)Scenario sensitivity correction coefficient : If any of the following scenario conditions are met, including And Belongs to the supply chain finance path; there is cross-year deferral; there are high-risk tags; then set , otherwise, ; (3)Time decay compensation coefficient : Define the associated time span , if , then , otherwise, ; Represents the business time difference of the th record, Represents the preset business time difference threshold (time decay trigger condition); S23, Network structure optimization: For the constructed weighted edge set , if , then prune (delete); if there are duplicate edges And , merge and retain the higher weight; finally generate the association network: ; Represents the weight threshold (used to filter low-weight edges), Represents the finally generated multi-dimensional association network, Represents the node set in the multi-dimensional association network, Represents the finally retained weighted edge set.

[0026] S3 specifically includes: S31, Graph attention feature fusion: Input the node attributes (such as subject codes, business dimension tags) and edge attributes (such as dynamic weight values, association types) of the multi-dimensional association network into the multi-head graph attention network to generate node feature vectors that fuse business semantics.

[0027] Based on the dynamic weights of the edges, construct an attention mask matrix to suppress the influence of associated edges with weight values lower than the preset threshold on feature propagation.

[0028] S32, Cross-level path detection: Execute a hybrid search strategy of breadth-first search (BFS) and depth-first search (DFS) to detect paths that meet the following conditions: the path span ≥ 3 business levels (e.g., subject → supplier → regional market → tax entity); the path contains at least one implicit association edge triggered by business dimension labels; Perform dynamic pruning on the detected paths: remove duplicate sub-paths and redundant paths with a node overlap degree higher than 80%.

[0029] S33, Path strength quantization: Calculation of path basic strength value: Define the basic strength of path P as the product of the dynamic weight values of all edges on this path and the corresponding attention coefficients, considering the exponential decay of the path length: , where, represents an association path, consisting of several nodes and edges, represents the path in one edge, represents the edge the dynamic weight value of, represents the edge the attention coefficient of, is the path length of, is the path length decay factor, represents the path basic strength value; Application of business scenario correction coefficient: Introduce a business scenario-related correction coefficient to adjust the strength on the basis of the path basic strength value: If the path contains nodes with the "supply chain finance" label, multiply the strength value by ; If the path spans multiple fiscal years, multiply the strength value by 1 / (1 + log(annual span)) as the annual span correction coefficient.

[0030] S34, Significance verification and set generation: Robustness significance verification: Perform a Monte Carlo robustness test on the paths with the top 15% strength values. Specifically, perform 1000 random perturbations on the node features of each candidate path, recalculate the path strength each time, and statistically analyze the fluctuations of the strength values. If the strength volatility of a certain path is lower than 5%, it is considered that the path performs stably under random perturbations (the strength change is within the allowable range), that is, it passes the significance verification and is determined to be a robust implicit association path; Generation of implicit association relationship set: Summarize the paths that pass the significance verification to generate an implicit association relationship set. Each association record in the implicit association relationship set includes: The node sequence of the path and its corresponding business dimension label chain; The final strength index of the path (i.e., the value obtained by adjusting the path base strength by the service correction factor and then multiplying it by the stability probability of the path); Critical impact factor marker (identifying the 3 key nodes or edges that contribute the most to the path strength).

[0031] S4 specifically includes: S41, dynamic weight mapping: S411, for each path in the implicit relationship set , normalize its path strength index : , where represents the normalized strength value, represents the maximum and minimum strengths in the path set, represents the rd path's final strength value (from the output of S3); S412, establish the strength-visual parameter mapping rule: Node size , where represents the maximum normalized strength of the paths associated with the path starting point, represents the node radius; represents the minimum / maximum size of the node; Edge transparency , where represents the edge transparency, represents the normalized strength value of the path associated with the edge transparency, represents the minimum / maximum transparency of the edge; Node color saturation represents the color saturation, represents the heat value of the business dimension label is the global maximum heat; S413, the initial topological layout uses the force-directed algorithm, and the core node with the highest path strength is preferentially located in the central area of the canvas to construct the initial graph structure coordinate set: , where represents the initial graph topological layout coordinate set, is the layout algorithm, with the normalized strength as the weight input; S42, focus diffusion trigger: S421, when the user clicks on the target node , trigger the layout reconstruction process: Expand all paths related to , and among the top three-layer node subsets ranked by strength , according to the hierarchical propagation distance Set the node position: , where represents the target node currently selected by the user, represents the set of three-layer expanded nodes related to the target node, represents the layout propagation distance from the nodes of the layer to the center, represents the maximum diffusion radius, represents the maximum normalized intensity of the paths associated with the nodes of the layer; For strongly associated nodes within the diffusion range , apply a magnetic attraction effect to make them gather towards the center: ‖, where represents the coordinate vector of the node , represents the adsorption intensity coefficient, represents the magnetic attraction force received by the node , represents the coordinate vector of the focus node, ‖ represents the coordinate gradient (i.e., the modulus gradient of the position vector difference); S422, display the top 5 path summaries with the highest intensity in a floating window at the edge of the visible area ; S43, Dimension folding control: S431, The user performs operations through the dimension filtering panel: Label control: Check / uncheck business labels When , the corresponding node set and edge set are displayed / hidden; Intensity threshold slider control: where represents the business dimension label currently selected in the user interface, is the intensity threshold currently set by the user, represents the edge set of the entire graph; S432, Apply a dynamic blur special effect animation to the hidden or removed graph elements: Use a Gaussian blur kernel and perform the transition in the fade-out trajectory manner of represents the duration of the fade-out animation; S44, Multi-view collaborative update: S441, Global view: Retain the core skeleton network composed of high-intensity paths for the user to grasp the overall topological structure; S442, Local view: Synchronously update all visible nodes and paths of the current focus node and its diffusion area; S443, Linkage mechanism: When a certain path in the partial view is marked as abnormal (such as extremely high strength or label conflict), automatically highlight its start and end nodes in the global view, and thicken the edges along the path , where represents any edge in the path and is used to highlight in the global view.

[0032] This invention covers any alternatives, modifications, equivalent methods and solutions made within the essence and scope of this invention. For the public to have a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments of this invention. However, those skilled in the art can also fully understand this invention without the description of these details. Additionally, well-known methods, processes, procedures, components and circuits, etc. are not described in detail to avoid unnecessary confusion to the essence of this invention.

[0033] The above are only the preferred embodiments of this invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of this invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of this invention.

Claims

1. A method for visual display of multi-dimensional correlation relationships in financial statements, characterized in that, It includes the following steps: S1. Standardized processing of financial data: Connect to multiple-source financial systems, extract account codes, amount values, and business dimension tags, and generate a set of standardized financial data with timestamps; S2. Construction of a multi-dimensional association network: Using the set of standardized financial data as input, generate an initial edge set based on a preset account association rule library, and assign a weight value reflecting the business scenario sensitivity to each edge through a dynamic weight calculation module to form a multi-dimensional association network with weighted tags; S3. Mining of implicit association relationships: Input the multi-dimensional association network into an implicit association analysis model based on the graph attention mechanism, identify cross-level association paths through a path traversal algorithm, and generate a set of implicit relationships including path strength indicators; S4. Dynamic visualization display: According to the path strength indicators in the set of implicit relationships, call a parameterized layout engine to generate a dynamic visualization interface, supporting the triggering of focus diffusion and dimension folding of the association network through interactive operations.

2. The visualization method for multi-dimensional association relationships of financial statements according to claim 1, characterized in that The specific content of S1 includes: S11. Connect to the ERP, supply chain management system, and tax declaration platform through a heterogeneous data interface adapter, and extract the original financial data stream, where the original financial data stream includes account codes, amount values, business occurrence times, and source system identifiers; S12. Perform cross-system consistency processing on account codes: Detect the coding naming differences in different source systems, and perform coding correction through a preset industry standard account tree to generate a globally unified format account code; S13. Standardize the amount attribute: Convert multi-currency amounts into the local currency according to the official exchange rate at the time of business occurrence, and attach currency tags and exchange rate snapshot information; S14. Bind business dimension tags: According to a preset dimension mapping rule library, convert unstructured business description fields in the source system into a set of standardized business dimension tags, where the set of business dimension tags includes transaction types, related party levels, and fund usage classifications; S15. Generate timestamps and integrate data: Attach a data collection timestamp and a business timestamp to each data record, and finally output a set of standardized financial data with timestamps.

3. A method for visual display of multi-dimensional association relationships in financial statements according to claim 1, characterized in that, The specific content of S2 includes: S21. Generate an initial edge set: Based on a preset account association rule library, traverse the account codes and business dimension tags in the set of standardized financial data, perform logical associations, and generate explicit association edges and implicit association edges; S22. Calculate dynamic weights: Calculate a composite weight value for each association edge through a dynamic weight calculation module, and the composite weight value specifically includes basic weight calculation, scenario sensitivity correction, and time decay compensation; S23. Optimize the network structure: Prune the association edges with a composite weight value lower than the weight threshold, merge duplicate association paths, and generate a multi-dimensional association network, where the multi-dimensional association network includes node attributes, association edge attributes, and network metadata.

4. A method for visual display of multi-dimensional association relationships in financial statements according to claim 3, characterized in that, In S21: The explicit association edges are generated according to accounting standards: Automatically establish connections for account pairs with direct debit-credit relationships or accounting reconciliation relationships; The implicit association edges are generated based on business dimension tags: Generate cross-dimensional connections for account pairs sharing the same supply chain level tag, risk level tag, or fund usage classification tag.

5. A method for visual display of multi-dimensional association relationships in financial statements according to claim 3, characterized in that, The basic weight calculation includes assigning initial weights based on the association type. The basic weight of the explicit association edge is the preset accounting standard weight value, and the basic weight of the implicit association edge increases exponentially according to the number of shared tags. The scenario sensitivity correction includes calling the scenario parameters in the industry feature template to strengthen the weights of the association edges involved in sensitive services. Sensitive services include the capital flow path of the subjects related to supply chain finance; the association relationship of the cross-year deferred subjects, and the capital transfer between the subjects with a high-risk level label. The time decay compensation is to reduce the weight value of the association edge whose business timestamp exceeds the preset period according to the logarithmic function.

6. A method for visual display of multi-dimensional association relationships in financial statements according to claim 3, characterized in that The node attributes include inheriting the subject code, amount attribute, and business dimension label in the standardized financial data. The association edge attributes include storing the dynamic weight value, association type identifier, and the original parameter snapshot of the weight calculation basis. The network metadata includes recording the version number of the industry feature template and the weight calculation timestamp.

7. A method for visual display of multi-dimensional association relationships in financial statements according to claim 1, characterized in that, The specific steps of S3 are as follows: S31, Graph attention feature fusion: Input the node attributes and edge attributes of the multi-dimensional association network into the multi-head graph attention network to generate node feature vectors that fuse business semantics; construct an attention mask matrix based on the edge dynamic weights to suppress the influence of the association edges with weight values lower than the preset threshold on feature propagation. S32, Cross-level path detection: Execute a mixed search strategy of breadth-first traversal and depth-first traversal to detect paths that meet the double constraint conditions, and perform dynamic pruning on the detected paths: Remove duplicate sub-paths and redundant paths with a node overlap degree higher than 80%. The double constraint conditions include: The path span ≥ 3 business levels; Include at least one implicit association edge triggered by the business dimension label; S33, Path strength quantization: Calculate the basic path strength value: S base (P) = Π e∈P (w e ·α e ) × exp(-λL), where P represents an associated path composed of several nodes and edges, e ∈ P represents an edge in path P, w e represents the dynamic weight value of edge e, α e represents the attention coefficient of edge e, L is the length of path P, λ is the path length decay factor, and S base (P) represents the basic strength value of path P; Apply the business scenario correction coefficient: If the path contains a supply chain finance label node, the strength value is multiplied by 1.5; if the path spans accounting years, the strength value is multiplied by 1 / (1 + log(annual span)). S34, Perform a Monte Carlo robustness test on the paths with the top 15% strength values, including randomly perturbing the node features 1000 times and retaining the stable paths with a strength volatility < 5%. Generate the implicit relationship set: Aggregate the stable paths to generate the implicit association relationship set. Each association record in the implicit association relationship set includes the path node sequence, its business dimension label chain, the final strength index, and the key influence factor mark.

8. A method for visual display of multi-dimensional association relationships of financial statements according to claim 1, characterized in that, The specific steps of S4 are as follows: S41, Dynamic weight mapping: Normalize the path strength indicators in the implicit relationship set to the interval value of [0,1], and establish the strength-visual parameter mapping rule. The strength-visual parameter mapping rule includes that the node size is positively correlated with the strength of the path starting point, the edge transparency increases linearly with the path strength value, and the node color saturation reflects the heat value of the business dimension label to which it belongs; Generate the initial topological layout based on the strength-visual parameter mapping rule, where the core nodes are automatically positioned to the focus area of the canvas. S42, Focus Diffusion Trigger: In response to the user's interaction event of selecting a target node, perform real-time layout reconstruction, including expanding three layers of associated nodes centered on the user-selected target node in descending order of path strength. The layer spacing is inversely proportional to the path strength value, and a magnetic attraction effect is applied to the nodes within the diffusion range, causing strongly associated nodes to automatically gather towards the center. Display the top 5 cross-level path summaries with the highest strength floating at the edge of the diffusion area. S43, Dimension Folding Control: Build an interactive dimension filtering panel and respond to the following operations: When checking / unchecking business dimension labels, hide or display the nodes and associated edges of the corresponding labels in real time. Slide the strength threshold bar to dynamically fold the path edges with strength values lower than the current threshold. Apply a Gaussian blur effect to the folded elements and retain a 0.5-second fade-out trajectory. S44, Multi-View Collaborative Update: Includes a global view, a local view, and a linkage mechanism.

9. A method for visual display of multi-dimensional association relationships in financial statements according to claim 8, characterized in that The global view is to retain the core skeleton network composed of high-strength paths for the user to grasp the overall topological structure. The local view is to synchronously update all visible nodes and paths of the current focus node and its diffusion area. The linkage mechanism is defined as follows: When a certain path P in the local view * is marked as abnormal, the start and end nodes of the path are automatically highlighted in the global view, and the edges e ∈ P are thickened along the path * .

Images