A multi-source business data correlation analysis method based on digital twinning

Abstract

The application discloses a kind of multi-source business data association analysis methods based on digital twinning, comprising the following steps: obtaining multi-source business data, constructs digital twinning coupling body;Entity mapping processing is executed, and candidate association edge set is generated, multidimensional gate control constraint processing is executed, and digital twinning association graph is generated;Extract initial search state set, and construct Bron-Kerbosch maximum group recursive search process;Candidate expansion priority control is executed, and candidate group set is formed;Group formation semantic locking processing is carried out to candidate group set, and maximum group recursive search result is formed;Digital twinning entity state is monitored, and determines inversion source and reconstructs association influence domain, and forms local maximum group identification result;Maximum group recursive search result, local maximum group identification result is executed group verification convergence processing, and association analysis result is output.The application improves the accuracy and dynamic analysis capability of multi-source business data association identification.

Description

Technical Field

[0001] This invention relates to the field of business correlation analysis technology, and in particular to a method for multi-source business data correlation analysis based on digital twins. Background Technology

[0002] With the continuous integration of digital business systems, sensing devices, process platforms, and status monitoring methods, a large amount of business data with different sources, structures, and update rhythms has accumulated in the target business scenarios. In order to improve the ability to identify the relationship between business objects, existing technologies usually adopt methods such as multi-source data fusion, relationship modeling, graph structure analysis, rule matching, and event tracking to mine the relationship between business objects and use the analysis results for business monitoring, risk identification, process tracking, status analysis, and decision support.

[0003] While existing technologies can achieve the organization and correlation analysis of multi-source business data to a certain extent, they still have significant limitations in practical applications. Existing solutions typically handle multi-source business data at the level of data aggregation, entity mapping, relationship edge building, or rule matching, lacking a unified coupling organization mechanism for business objects, object relationships, business events, and business states. This makes it difficult to express the structural, semantic, and state correlations between multi-source business data within the same framework, thus affecting the completeness of subsequent correlation analysis. Most existing graph analysis solutions rely on general relationship retrieval, path analysis, neighborhood expansion, or ordinary clustering methods, which are insufficient in identifying highly coupled correlation groups. Especially in complex business scenarios, it is difficult to identify entity combinations with tightly connected features and business constraint features from a large number of candidate relationships, resulting in insufficient accuracy and weak structural stability of the correlation results.

[0004] Therefore, how to provide a method for multi-source business data correlation analysis based on digital twins is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a method for multi-source business data association analysis based on digital twins. This invention improves the accuracy of multi-source business data association structure identification, semantic constraint capability, and dynamic analysis capability under state change scenarios by constructing a digital twin couple and a digital twin association graph, and introducing Bron-Kerbosch maximal clique recursive search, candidate expansion priority control, clique formation semantic locking, local maximal clique identification, and clique verification convergence processing.

[0006] A method for multi-source business data correlation analysis based on digital twins according to an embodiment of the present invention includes the following steps:

[0007] Acquire multi-source business data from the target business scenario and construct a digital twin coupling based on the multi-source business data;

[0008] Entity mapping is performed based on the digital twin coupler to generate a candidate set of associated edges, and multi-dimensional gating constraint processing is performed on the candidate set of associated edges to generate a digital twin association graph.

[0009] An initial search state set is extracted from the digital twin association graph, and a Bron-Kerbosch maximal clique recursive search process is constructed based on the initial search state set;

[0010] In the recursive search process of the Bron-Kerbosch maximal clique, candidate expansion priority control is performed on the expansion node set, and recursive expansion is performed according to the expansion order of candidate nodes to form a candidate clique set;

[0011] Perform semantic locking processing on the candidate cluster set. While maintaining the pairwise connectivity of nodes within the candidate cluster set, filter target candidate clusters that meet the preset business semantic closure conditions, update the maximal cluster recursive search process, and form the maximal cluster recursive search results.

[0012] The state of digital twin entities in the digital twin entity set is monitored to determine the inversion source of the state transition. The associated influence domain is reconstructed according to the inversion source. Local maxima clique identification is performed within the associated influence domain to form the local maxima clique identification result.

[0013] Perform clique verification convergence processing on the recursive search results of maximal cliques and the identification results of local maximal cliques, and output the correlation analysis results.

[0014] Optionally, the construction of the digital twin coupler specifically includes:

[0015] The system performs access processing on multi-source business data in the target business scenario, converts each data record entering the processing flow into a unified record structure, and performs normalization processing on the record element set in the unified record structure to form the original dataset.

[0016] The original dataset is parsed and processed. Each data record is split, classified, and associated according to the record element set in the unified record structure to obtain a set of coupled components. Coupled organizational relationships are established according to the correspondence between each data record before and after splitting, forming coupled organizational data.

[0017] Based on the coupled organization data, a coupled body construction process is performed. A digital twin coupled association is established according to the coupled component set. A combination mapping is performed on each coupled component according to the coupled organization relationship to form a digital twin coupled body.

[0018] Optionally, obtaining the digital twin association graph specifically includes:

[0019] Entity mapping is performed based on the digital twin coupling. Corresponding retrieval, cross-comparison, and merging and positioning are performed on each coupling component in the digital twin coupling along the coupling organizational relationship. Coupling components representing the same business object are mapped to the same digital twin entity, and coupling components with organizational relationships are mapped to the corresponding digital twin entities to form a set of digital twin entities.

[0020] The digital twin entities in the digital twin entity set are paired up in pairs. For each entity pair, the organizational association results, co-occurrence results, action transmission results, and state linkage results between the corresponding coupling components are read. The association strength is calculated according to the preset association judgment rules. Entity pairs that meet the candidate edge construction conditions are determined as candidate association edges, forming a candidate association edge set.

[0021] Multidimensional gating constraints are applied to the candidate association edge set. For each candidate association edge, gating judgments are performed from the perspectives of organizational association consistency, relationship transmission continuity, event effect closure, and state change coordination. The results of each gating judgment are then jointly screened. Candidate association edges that meet the retention conditions in the joint screening are determined as valid association edges, forming a set of valid association edges.

[0022] The digital twin entities in the digital twin entity set are used as graph nodes, and the set of effective associated edges are used as connecting edges between nodes. The node connection relationship is established according to the entity pairing results corresponding to the effective associated edges, thus forming a digital twin association graph.

[0023] Optionally, the construction of the Bron-Kerbosch maximal clique recursive search process specifically includes:

[0024] Based on the digital twin association graph, the graph structure is read, and node adjacency relationships are established between each digital twin entity node and the digital twin entity nodes that maintain effective association edges, thus obtaining node adjacency relationship data;

[0025] Node order is constructed based on node adjacency relationship data. The node index order is formed according to the registration order of digital twin entity nodes in the node adjacency relationship data. For each digital twin entity node in the node index order, the corresponding search primitive is extracted to form node search primitive data.

[0026] Based on the node search primitive data, the initial search state set is constructed. Each digital twin entity node in the node index order is written into the corresponding search state unit to form an initial search state that corresponds one-to-one with each digital twin entity node. All initial search states are then organized into an initial search state set.

[0027] Based on the initial search state set, a compatibility propagation recursive expansion is performed. From the range of nodes to be expanded corresponding to any initial search state, a digital twin entity node to be expanded is selected. The selected digital twin entity node to be expanded is incorporated into the current clique primitive to form an expanded clique node combination. The selected digital twin entity node to be expanded is used to propagate and update the edge compatibility records in the current search state to form an expanded search state. The expanded search state includes the expanded clique node combination, the continued expansion range, the continued exclusion range, and the updated edge compatibility records.

[0028] The expanded search state is recursively continued. The digital twin entity nodes in the continued expansion range are repeatedly treated as digital twin entity nodes to be expanded and compatibility propagation recursive expansion is performed. When the continued expansion range is empty, there are no nodes in the continued exclusion range that can be incorporated into the current expansion clique combination, and the nodes maintain common edge compatibility with all digital twin entity nodes, the current expansion clique node combination is determined as a maximal clique. The recursive search states that have completed expansion and are formed corresponding to the initial search state set are organized into a Bron-Kerbosch maximal clique recursive search process.

[0029] Optionally, the formation of the candidate group set specifically includes:

[0030] For each recursive search state in the Bron-Kerbosch maximal clique recursive search process, the extended range is read, and the continued extended range in each recursive search state is read as the corresponding extended node set to form the candidate extended control input;

[0031] Based on the candidate expansion control input, the candidate priority value is calculated. After the digital twin entity node to be expanded is merged into the expansion clique node combination, the number of nodes that still maintain common edge compatibility with the continued expansion range, the continued exclusion range, and the updated edge compatibility record are counted. The expansion retention value, exclusion constraint value, and compatibility propagation value are formed. The candidate priority value corresponding to each digital twin entity node to be expanded in the expansion node set is obtained according to the preset calculation method.

[0032] The candidate node expansion order is constructed based on the candidate priority value. The candidate nodes to be expanded are arranged in order of priority from high to low to form the candidate node expansion order. The first candidate node to be expanded in the candidate node expansion order is determined as the priority expansion node in this round. The candidate node expansion order is written into the order control record of the corresponding recursive search state to form the priority control search state.

[0033] Based on the priority control search state, the recursive expansion is performed in sequence. Based on the digital twin entity node to be expanded, the scope of continued expansion, the scope of continued exclusion, and the updated edge compatibility record are synchronously shrunk and updated to obtain the next layer recursive search state corresponding to the currently read digital twin entity node to be expanded. The remaining digital twin entity nodes to be expanded in the candidate node expansion order are retained as nodes to be processed in the current layer.

[0034] Recursive continuation control is executed on each of the next-level recursive search states formed by the sequential recursive expansion process. The candidate priority value calculation process, the candidate node expansion order construction process, and the sequential recursive expansion process are repeatedly executed until the preset termination condition is reached. The current expansion clique node combination is recorded as a candidate clique, and all candidate cliques formed by the Bron-Kerbosch maximal clique recursive search process are organized into a candidate clique set.

[0035] Optionally, the formation of the maximal clique recursive search result specifically includes:

[0036] Perform cluster screening on the candidate cluster set, perform pairwise connectivity verification according to the connection status between the digital twin entity nodes in each candidate cluster, and retain the candidate clusters that meet the pairwise connectivity conditions as candidate clusters to be judged, forming a set of candidate clusters to be judged;

[0037] Based on the set of candidate clusters to be determined, a business semantic closure determination is performed. For each candidate cluster in the set of candidate clusters to be determined, the associated content of the digital twin coupler corresponding to the candidate cluster is read to form semantic closure determination information. The semantic closure determination information is then matched with the preset business semantic closure conditions. Candidate clusters whose matching results meet the preset business semantic closure conditions are determined as target candidate clusters, thus forming a target candidate cluster set.

[0038] The maximal clique recursive search process is updated based on the target candidate clique set. Each target candidate clique in the target candidate clique set is written back to the search state unit of the corresponding recursive search state. The state locking update is performed according to the search state corresponding to the target candidate clique. The extension branches that meet the preset business semantic closure conditions are retained in the maximal clique recursive search process, and the extension branches that do not meet the preset business semantic closure conditions are removed from the maximal clique recursive search process, forming the maximal clique recursive search result.

[0039] Optionally, the formation of the local maxima identification result specifically includes:

[0040] The state of digital twin entities in the digital twin entity set is monitored, the current state is compared with the historical state, digital twin entities whose state representation has changed are selected to form a state transition entity set, and the inversion source is determined based on the triggering effect of the state transition entities on the associated structure to form an inversion source set.

[0041] Based on the set of inversion sources, the associated influence domain is reconstructed. Starting from each inversion source, the influence transmission is tracked along the node connection relationship in the digital twin association graph. The associated range affected by state transition is extracted and organized into associated influence domains corresponding to each inversion source, forming a set of associated influence domains.

[0042] Local maximal clique identification is performed based on the set of associated influence domains. Local recursive expansion is performed in each associated influence domain according to the node connection relationship. The current expanded node combination is determined as a local maximal clique, forming the local maximal clique identification result.

[0043] Optionally, the output of the association analysis results specifically includes:

[0044] The clique results in the recursive search results of maximal cliques and the clique results in the local maximal clique identification results are written into the clique result set to form the clique set to be verified.

[0045] Clique verification is performed on the clique set to be verified. Correspondence comparison, consistency screening, and conflict identification are performed on the clique results in the clique set to be verified. Clique results that meet the retention conditions are determined as valid clique results, and clique results with overlapping, cross or conflicting relationships are determined as clique results to be converged, thus forming a clique verification result set.

[0046] Based on the clique verification result set, clique convergence processing is performed. The clique results to be converged are merged, overlaps are resolved, and conflicts are resolved. The converged clique results and the valid clique results are organized together into a set of associated core cliques, and the association analysis results are output based on the set of associated core cliques.

[0047] The beneficial effects of this invention are:

[0048] This invention acquires multi-source business data and constructs a digital twin coupling body, incorporating business object data, business relationship data, business event data, and business status data into a unified structured coupling carrier. Then, based on the digital twin coupling body, entity mapping processing and candidate association edge generation processing are performed. This enables the unified organization, association representation, and coupling mapping of multi-source business data in the same processing link, overcoming the shortcomings of existing technologies that only focus on the aggregation, splicing, or loose edge construction of multi-source business data. This improves the integrity of the associated information and the consistency of the structural expression in complex business scenarios.

[0049] This invention forms a digital twin association graph by performing multi-dimensional gating constraint processing on the candidate association edge set, extracts an initial search state set from the digital twin association graph, constructs a Bron-Kerbosch maximal clique recursive search process, further performs candidate expansion priority control on the expanded node set, and performs clique semantic locking processing on the candidate clique set. It can filter target candidate cliques that meet the preset business semantic closure conditions while maintaining the pairwise connectivity of nodes within the candidate clique set. This overcomes the problem in existing technologies that rely solely on general connectivity, ordinary path analysis, or conventional clustering methods, making it difficult to identify highly coupled entity groups. As a result, it improves the accuracy of multi-source business data association identification, the business interpretability of clique results, and the ability to identify complex association structures.

[0050] This invention monitors the state of digital twin entities to identify the inversion source of state transitions and reconstructs the associated influence domain according to the inversion source. It then performs local maximal clique identification within the associated influence domain. This enables targeted reanalysis of the affected association range in scenarios of state change, overcoming the problems of insufficient dynamic response capability and high processing cost caused by the use of global recalculation or static update methods in the prior art. This enhances the adaptability of the association analysis process to state changes and improves the timeliness and pertinence of local association identification in dynamic scenarios.

[0051] This invention performs clique verification and convergence processing on the recursive search results of maximal cliques and the identification results of local maximal cliques, and performs unified verification, convergence and output of clique results from different sources. This overcomes the problems of overlapping, conflicting, discrete output and insufficient consistency of clique results in the prior art, thereby improving the stability, uniformity and usability of the final association analysis results, and enabling the output results to more accurately represent the association relationship, event transmission relationship and state linkage relationship between business objects in the target business scenario. Attached Figure Description

[0052] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0053] Figure 1 This is a flowchart of a multi-source business data correlation analysis method based on digital twins proposed in this invention;

[0054] Figure 2 This diagram illustrates the construction of the Bron-Kerbosch maximal clique recursive search process for a multi-source business data correlation analysis method based on digital twins proposed in this invention. Detailed Implementation

[0055] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0056] refer to Figures 1-2 A method for multi-source business data correlation analysis based on digital twins includes the following steps:

[0057] Acquire multi-source business data from the target business scenario and construct a digital twin coupling based on the multi-source business data;

[0058] Entity mapping is performed based on the digital twin coupler to generate a candidate set of associated edges, and multi-dimensional gating constraint processing is performed on the candidate set of associated edges to generate a digital twin association graph.

[0059] An initial search state set is extracted from the digital twin association graph, and a Bron-Kerbosch maximal clique recursive search process is constructed based on the initial search state set;

[0060] In the recursive search process of the Bron-Kerbosch maximal clique, candidate expansion priority control is performed on the expansion node set, and recursive expansion is performed according to the expansion order of candidate nodes to form a candidate clique set;

[0061] Perform semantic locking processing on the candidate cluster set. While maintaining the pairwise connectivity of nodes within the candidate cluster set, filter target candidate clusters that meet the preset business semantic closure conditions, update the maximal cluster recursive search process, and form the maximal cluster recursive search results.

[0062] The state of digital twin entities in the digital twin entity set is monitored to determine the inversion source of the state transition. The associated influence domain is reconstructed according to the inversion source. Local maxima clique identification is performed within the associated influence domain to form the local maxima clique identification result.

[0063] Perform clique verification convergence processing on the recursive search results of maximal cliques and the identification results of local maximal cliques, and output the correlation analysis results.

[0064] In this embodiment, the construction of the digital twin coupler specifically includes:

[0065] The system performs access processing on multi-source business data in the target business scenario, converts each data record entering the processing flow into a unified record structure, and performs normalization processing on the record element set in the unified record structure to form the original dataset. The multi-source business data includes business object data, business relationship data, business event data, and business status data.

[0066] The original dataset is parsed and processed. Each data record is split, classified, and associated according to the record element set in the unified record structure to obtain a set of coupled components. Coupled organizational relationships are established according to the correspondence between each data record before and after splitting, forming coupled organizational data.

[0067] Based on the data of the coupled organization, the process of constructing the coupled body is performed. Digital twin coupling associations are established according to the coupling component sets. Combination mapping is performed on each coupled component according to the coupling organization relationship to form a digital twin coupled body. The digital twin coupled body is a structured coupling carrier built based on multi-source business data. It includes the coupling component sets and the coupling organization relationship. The coupling component sets include business object components, business relationship components, business event components, and business state components. The coupling organization relationship is used to represent the correspondence, association, and organization relationship between each coupled component. The digital twin coupled body is used to perform unified organization, association representation, and coupling mapping on business objects, object relationships, business events, and business states in the target business scenario.

[0068] In this embodiment, obtaining the digital twin association graph specifically includes:

[0069] Entity mapping is performed based on the digital twin coupling. Corresponding retrieval, cross-comparison, and merging and positioning are performed on each coupling component in the digital twin coupling along the coupling organizational relationship. Coupling components representing the same business object are mapped to the same digital twin entity, and coupling components with organizational relationships are mapped to the corresponding digital twin entities to form a set of digital twin entities.

[0070] The system pairs digital twin entities in the digital twin entity set. For each entity pair, it reads the organizational association results, co-occurrence results, action transmission results, and state linkage results between the corresponding coupling components. According to the preset association judgment rules, it performs association strength calculation and determines the entity pairs that meet the candidate edge establishment conditions as candidate association edges, forming a candidate association edge set. The organizational association result is the processing result that characterizes whether there is a coupling organizational relationship between two digital twin entities in the digital twin coupling body and the strength of the coupling organizational relationship. The co-occurrence result is the processing result that characterizes the synchronous occurrence of two digital twin entities in the same business process, the same event record, and the same state change process. The action transmission result is the processing result that characterizes whether the business event, business relationship, and business state corresponding to one digital twin entity has a transmission effect on another digital twin entity and whether the transmission path is continuous. The state linkage result is the processing result that characterizes whether there is a linkage relationship between two digital twin entities in terms of state change direction, state change timing, and state change amplitude.

[0071] Multidimensional gating constraints are applied to the candidate association edge set. For each candidate association edge, gating judgments are performed from the perspectives of organizational association consistency, relationship transmission continuity, event action closure, and state change coordination. The results of each gating judgment are then jointly screened. Candidate association edges that meet the retention conditions in the joint screening are determined as valid association edges, forming a set of valid association edges. Organizational association consistency is used to characterize whether the coupling organizational relationship between the entities corresponding to the candidate association edge is unified. Relationship transmission continuity is used to characterize whether the relationship transmission link between the entities corresponding to the candidate association edge is continuous. Event action closure is used to characterize whether the event action process between the entities corresponding to the candidate association edge is closed. State change coordination is used to characterize whether the state changes between the entities corresponding to the candidate association edge are coordinated.

[0072] The digital twin entities in the digital twin entity set are used as graph nodes, and the set of effective associated edges are used as connecting edges between nodes. The node connection relationship is established according to the entity pairing results corresponding to the effective associated edges, thus forming a digital twin association graph.

[0073] In this embodiment, the construction of the Bron-Kerbosch maximal clique recursive search process specifically includes:

[0074] Based on the digital twin association graph, the graph structure is read, and node adjacency relationships are established between each digital twin entity node and the digital twin entity nodes that maintain effective association edges. The node adjacency relationship data is used to characterize the node connection status in the digital twin association graph.

[0075] Node order construction is performed based on node adjacency relationship data. The node index order is formed according to the registration order of digital twin entity nodes in the node adjacency relationship data. For each digital twin entity node in the node index order, the corresponding search primitive is extracted. The search primitive includes the current clique primitive composed of the current digital twin entity node, the range of nodes to be expanded that are connected to the current digital twin entity node and are located after the node index order, the range of excluded nodes that are connected to the current digital twin entity node and are located before the node index order, and the connection compatibility record of the current digital twin entity node with other digital twin entity nodes in the digital twin association graph, forming the node search primitive data.

[0076] Based on the node search primitive data, the initial search state set is constructed. Each digital twin entity node in the node index order is written into the corresponding search state unit to form an initial search state that corresponds one-to-one with each digital twin entity node. All initial search states are organized into an initial search state set. The search state unit is a state record structure used to write, save and call the initial search state, including the current clique primitive, the range of nodes to be expanded, the range of excluded nodes, and the edge compatibility record.

[0077] Based on the initial search state set, a compatibility propagation recursive expansion is performed. From the range of nodes to be expanded corresponding to any initial search state, a digital twin entity node to be expanded is selected. The selected digital twin entity node to be expanded is incorporated into the current clique primitive to form an expanded clique node combination. The selected digital twin entity node to be expanded is used to perform propagation update on the edge compatibility records in the current search state. The propagation update includes keeping digital twin entity nodes that maintain common edge compatibility with the expanded clique node combination in the continued expansion range, removing digital twin entity nodes that do not maintain common edge compatibility with the expanded clique node combination from the continued expansion range, and keeping digital twin entity nodes that maintain common edge compatibility with the expanded clique node combination in the excluded node range in the continued exclusion range, forming the expanded search state. The expanded search state includes the expanded clique node combination, the continued expansion range, the continued exclusion range, and the updated edge compatibility records.

[0078] The expanded search state is recursively continued by repeatedly treating the digital twin entity nodes in the continued expansion range as digital twin entity nodes to be expanded, and performing compatibility propagation recursive expansion. Compatibility propagation recursive expansion includes continuously updating the expanded clique node combination, the continued expansion range, the continued exclusion range, and the updated edge compatibility record in each recursive continuation process. When the continued expansion range is empty, the current expanded search state is maintained as the recursive search state with completed expansion. When the continued expansion range is empty, there are no nodes in the continued exclusion range that can be incorporated into the current expanded clique node combination, and all digital twin entity nodes maintain common edge compatibility, the current expanded clique node combination is determined as a maximal clique, and all the recursive search states with completed expansion corresponding to the initial search state set are organized into a Bron-Kerbosch maximal clique recursive search process.

[0079] This invention constructs an initial search state set on a digital twin association graph, including the current clique primitive, the range of nodes to be expanded, the range of excluded nodes, and the edge compatibility record. It introduces a compatibility propagation recursive expansion and recursive continuation mechanism to realize the state organization, localized contraction, and compatibility constraint update of the maximal clique search process. This improves the accuracy of complex association structure identification, the stability of the search process, and the ability to extract highly coupled entity groups.

[0080] In this embodiment, the formation of the candidate cluster set specifically includes:

[0081] For each recursive search state in the Bron-Kerbosch maximal clique recursive search process, the extended range is read, and the continued extended range in each recursive search state is read as the corresponding extended node set to form the candidate extended control input;

[0082] Based on the candidate expansion control input, candidate priority values ​​are calculated. After the digital twin entity node to be expanded is incorporated into the expansion clique node combination, the number of nodes that still maintain common edge compatibility with the continued expansion range, the continued exclusion range, and the updated edge compatibility record are counted. This forms the expansion retention value, exclusion constraint value, and compatibility propagation value. The candidate priority value corresponding to each digital twin entity node to be expanded in the expansion node set is obtained according to the preset calculation method. The preset calculation method is that the candidate priority value is equal to the first weight multiplied by the expansion retention value, plus the second weight multiplied by the compatibility propagation value, and minus the third weight multiplied by the exclusion constraint value. The first weight, the second weight, and the third weight satisfy the sum of one. The candidate priority value is used to characterize the priority expansion degree of the corresponding digital twin entity node to be expanded.

[0083] The candidate node expansion order is constructed based on the candidate priority value. The candidate nodes to be expanded are arranged in order of priority from high to low to form the candidate node expansion order. The first candidate node to be expanded in the candidate node expansion order is determined as the priority expansion node in this round. The candidate node expansion order is written into the order control record of the corresponding recursive search state to form the priority control search state.

[0084] Based on the priority control search state, the recursive expansion is performed in sequence. Based on the digital twin entity node to be expanded, the scope of continued expansion, the scope of continued exclusion, and the updated edge compatibility record are synchronously shrunk and updated to obtain the next layer recursive search state corresponding to the currently read digital twin entity node to be expanded. The remaining digital twin entity nodes to be expanded in the candidate node expansion order are retained as nodes to be processed in the current layer.

[0085] Recursive continuation control is executed on each of the next-level recursive search states formed by the sequential recursive expansion process. The candidate priority value calculation process, the candidate node expansion order construction process, and the sequential recursive expansion process are repeatedly executed until the preset termination condition is reached. The preset termination condition is that the continued expansion range in the corresponding recursive search state is empty, and there are no digital twin entity nodes in the continued exclusion range that can be merged into the current expansion clique node combination and maintain common edge compatibility with all digital twin entity nodes in the current expansion clique node combination. The current expansion clique node combination is recorded as a candidate clique, and all candidate cliques formed by the Bron-Kerbosch maximal clique recursive search process are organized into a candidate clique set.

[0086] This invention enhances the targeting of expansion node selection and the ability to control recursive paths by introducing candidate priority value calculation, candidate node expansion order control, and recursive state synchronous contraction and update mechanisms into the Bron-Kerbosch maximal clique recursive search process. This reduces invalid expansion and redundant search, improves the efficiency of candidate clique formation, the stability of maximal clique search, and the accuracy of identifying highly coupled entity groups in complex business association scenarios.

[0087] In this embodiment, the formation of the recursive search result for the maximal clique specifically includes:

[0088] Perform cluster screening on the candidate cluster set, perform pairwise connectivity verification according to the connection status between the digital twin entity nodes in each candidate cluster, and retain the candidate clusters that meet the pairwise connectivity conditions as candidate clusters to be judged, forming a set of candidate clusters to be judged;

[0089] Based on the set of candidate clusters to be determined, a business semantic closure determination is performed. For each candidate cluster in the set of candidate clusters to be determined, the associated content of the digital twin coupler corresponding to the candidate cluster is read to form semantic closure determination information. The semantic closure determination information is then matched with the preset business semantic closure conditions. Candidate clusters whose matching results meet the preset business semantic closure conditions are determined as target candidate clusters, thus forming a target candidate cluster set.

[0090] The maximal clique recursive search process is updated based on the target candidate clique set. Each target candidate clique in the target candidate clique set is written back to the search state unit of the corresponding recursive search state. The state locking update is performed according to the search state corresponding to the target candidate clique. The extension branches that meet the preset business semantic closure conditions are retained in the maximal clique recursive search process, and the extension branches that do not meet the preset business semantic closure conditions are removed from the maximal clique recursive search process, forming the maximal clique recursive search result.

[0091] This invention introduces a business semantic closure judgment and recursive search process update mechanism on the basis of candidate cluster screening, and integrates structural connectivity constraints and business semantic constraints into the maximal cluster identification process. This avoids the result bias caused by retaining candidate clusters only based on node connectivity, and improves the business interpretability, screening accuracy and result reliability of the recursive search results of maximal clusters.

[0092] In this embodiment, the formation of the local maxima identification result specifically includes:

[0093] The state of digital twin entities in the digital twin entity set is monitored, the current state is compared with the historical state, digital twin entities whose state representation has changed are screened to form a state transition entity set, and the inversion source is determined based on the triggering effect of the state transition entities on the associated structure. The inversion source is determined by performing association transmission tracing on each digital twin entity in the state transition entity set, identifying digital twin entities that have a triggering effect on node connection relationship and local maxima clique identification range, and the identification results are determined as inversion sources to form an inversion source set. The state of digital twin entities is used to represent the current state representation information of digital twin entities in the target business scenario.

[0094] Based on the set of inversion sources, the associated influence domain is reconstructed. Starting from each inversion source, the influence transmission is tracked along the node connection relationship in the digital twin association graph. The associated range affected by state transition is extracted and organized into associated influence domains corresponding to each inversion source, forming a set of associated influence domains.

[0095] Local maximal clique identification is performed based on the set of associated influence domains. Within each associated influence domain, local recursive expansion is performed according to the node connection relationship. Local recursive expansion includes retaining digital twin entities that are compatible with the current expansion node combination in the continued expansion range, removing digital twin entities that are not compatible with the current expansion node combination from the continued expansion range, and determining the current expansion node combination as a local maximal clique when the continued expansion range is empty and there are no digital twin entities that can be incorporated into the current expansion node combination and are compatible with all digital twin entities in the current expansion node combination. This forms the local maximal clique identification result.

[0096] This invention introduces mechanisms for state monitoring, inversion source determination, correlation influence domain reconstruction, and local maximal clique identification to achieve targeted tracking and local response to correlation changes caused by state transitions. This avoids processing redundancy caused by global recalculation and improves the timeliness of correlation identification, the accuracy of range positioning, and the ability to analyze locally highly coupled structures in dynamic scenarios.

[0097] In this embodiment, the output of the correlation analysis results specifically includes:

[0098] The clique results in the recursive search results of maximal cliques and the clique results in the local maximal clique identification results are written into the clique result set to form the clique set to be verified.

[0099] Clique verification is performed on the clique set to be verified. Correspondence comparison, consistency screening, and conflict identification are performed on the clique results in the clique set to be verified. Clique results that meet the retention conditions are determined as valid clique results, and clique results with overlapping, cross or conflicting relationships are determined as clique results to be converged, thus forming a clique verification result set.

[0100] Based on the clique verification result set, clique convergence processing is performed. The clique results to be converged are merged, overlaps are resolved, and conflicts are resolved. The converged clique results and the valid clique results are organized together into a set of associated core cliques. Based on the set of associated core cliques, the association analysis results are output. The association analysis results are business association identification results formed based on multi-source business data, which are used to characterize the association relationship, event transmission relationship, and state linkage relationship between business objects in the target business scenario.

[0101] Example 1: To verify the feasibility of this invention in practice, it was applied to a complex business scenario involving warehouse fulfillment and equipment collaboration. This scenario continuously generates order flow records, task allocation records, warehouse location change records, equipment operation records, alarm event records, and status feedback records. Data from different sources differs in structure, field naming, update frequency, and granularity. In actual operation, a delay in a task often manifests not only as a single node status change but also further impacts warehouse location occupancy, equipment scheduling, transportation routes, and upstream / downstream collaboration. Traditional methods typically rely on direct field association, rule matching, or partial log tracing for investigation. These methods can only identify explicit direct relationships and struggle to identify truly highly coupled groups of relationships. They also fail to quickly pinpoint the affected area after a status change, resulting in fragmented association results, slow dynamic response, and high manual investigation workload.

[0102] In this embodiment, multi-source business data is first accessed, and the data from different sources is uniformly converted into a consistent record structure, allowing object information, relationship information, event information, and state information to enter the same processing link. Then, the original data is parsed, and related content within the same record is split and a corresponding relationship is established, forming coupled organizational data. A digital twin coupled entity is then constructed based on this coupled organizational data. This unifies the organization of objects, relationships, events, and states scattered across different sources, providing a consistent foundation for subsequent entity mapping and graph construction. Next, entity mapping is performed along the coupled organizational relationships, merging content representing the same business object into the same digital twin entity, and mapping content with organizational relationships to corresponding entities. Then, digital twin entities are paired up, and a candidate set of related edges is generated based on organizational relationship results, co-occurrence results, action transmission results, and state linkage results. After multi-dimensional gating constraint processing, a digital twin association graph is formed. The resulting graph structure is no longer a simple object connection, but an association graph with business constraints and state constraints.

[0103] In the graph search phase, this invention extracts an initial search state set from the digital twin association graph and constructs a Bron-Kerbosch maximal clique recursive search process. Based on the search state units, a compatible propagation recursive expansion is performed. Then, candidate expansion priority control is applied to the expanded node set. The expansion order is constructed by calculating candidate priority values, prioritizing the expansion of nodes more likely to form high-quality candidate cliques, thereby reducing invalid branches. After candidate cliques are formed, a clique formation semantic locking process is performed. Candidate cliques satisfying the structural connectivity condition are further judged for business semantic closure, retaining only target candidate cliques that simultaneously satisfy both structural and business semantic constraints, and updating the maximal clique recursive search process. This avoids obtaining business-meaningless clique results solely from the graph structure.

[0104] During the dynamic operation phase, the system continuously monitors the state of digital twin entities. When a state transition is detected, it further identifies the inversion source that triggers the associated structure. Then, it reconstructs the associated influence domain starting from the inversion source and performs local maximal clique identification within the associated influence domain. In this way, when a local anomaly occurs, it is not necessary to re-perform a complete search of the entire graph. Instead, a local reanalysis is completed within the affected area, which can more quickly locate the anomaly propagation range and highly coupled influence groups. Finally, the recursive search results of maximal cliques and the results of local maximal clique identification are uniformly collected, and clique verification and clique convergence processing are performed to obtain the set of associated core cliques, and the association analysis results are output. This result can be used to assist in anomaly localization, collaborative scheduling, and risk assessment, and can more accurately characterize the association relationships, event transmission relationships, and state linkage relationships between business objects.

[0105] To further illustrate the effectiveness of this invention, the solution of this invention is compared with the traditional method. The specific comparison data is shown in Table 1:

[0106] Table 1. Performance comparison results of the method of the present invention and traditional correlation analysis methods

[0107] Comparison indicators traditional methods The present invention Total amount of multi-source business data 1,586,240 1,586,240 Number of digital twin entities formed 41,835 44,912 Final number of valid associated edges 227,408 149,275 Number of intermediate recursive search states 88,437 53206 Average global search time 17.4 seconds 10.5 seconds Explanatory power of candidate group business 60.7% 89.6% Time consumption of a single round of dynamic update 11.8 seconds 2.3 seconds Business personnel's result adoption rate 69.5% 91.2%

[0108] As shown in Table 1, under the same scale of multi-source business data, the method of this invention forms more digital twin entities, indicating that the invention has stronger structural expression capabilities in business object merging and association organization. Meanwhile, the method of this invention obtains fewer effective associated edges, indicating that redundant relationships are effectively compressed after multi-dimensional gating constraints. In the search phase, the number of intermediate recursive search states decreased from 88,437 to 53,206, and the average global search time decreased from 17.4 seconds to 10.5 seconds, indicating that the candidate expansion priority control in the maximal clique recursive search process of this invention can effectively reduce invalid expansions. In terms of result quality, the business interpretability rate of candidate cliques increased from 60.7% to 89.6%, and the result adoption rate by business personnel increased from 69.5% to 91.2%, indicating that the invention improves the business usability of results through clique formation semantic locking and clique verification convergence processing. In terms of dynamic processing, the time for a single round of dynamic updates decreased from 11.8 seconds to 2.3 seconds, indicating that the invention improves the local response capability in state change scenarios through inversion source determination and association influence domain reconstruction.

[0109] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for multi-source business data correlation analysis based on digital twins, characterized in that, Includes the following steps: Acquire multi-source business data from the target business scenario and construct a digital twin coupling based on the multi-source business data; Entity mapping is performed based on the digital twin coupler to generate a candidate set of associated edges, and multi-dimensional gating constraint processing is performed on the candidate set of associated edges to generate a digital twin association graph. An initial search state set is extracted from the digital twin association graph, and a Bron-Kerbosch maximal clique recursive search process is constructed based on the initial search state set; In the recursive search process of the Bron-Kerbosch maximal clique, candidate expansion priority control is performed on the expansion node set, and recursive expansion is performed according to the expansion order of candidate nodes to form a candidate clique set; Perform semantic locking processing on the candidate cluster set. While maintaining the pairwise connectivity of nodes within the candidate cluster set, filter target candidate clusters that meet the preset business semantic closure conditions, update the maximal cluster recursive search process, and form the maximal cluster recursive search results. The state of digital twin entities in the digital twin entity set is monitored to determine the inversion source of the state transition. The associated influence domain is reconstructed according to the inversion source. Local maxima clique identification is performed within the associated influence domain to form the local maxima clique identification result. Perform clique verification convergence processing on the recursive search results of maximal cliques and the identification results of local maximal cliques, and output the correlation analysis results.

2. The method for multi-source business data correlation analysis based on digital twins according to claim 1, characterized in that, The construction of the digital twin coupler specifically includes: The system performs access processing on multi-source business data in the target business scenario, converts each data record entering the processing flow into a unified record structure, and performs normalization processing on the record element set in the unified record structure to form the original dataset. The original dataset is parsed and processed. Each data record is split, classified, and associated according to the record element set in the unified record structure to obtain a set of coupled components. Coupled organizational relationships are established according to the correspondence between each data record before and after splitting, forming coupled organizational data. Based on the coupled organization data, a coupled body construction process is performed. A digital twin coupled association is established according to the coupled component set. A combination mapping is performed on each coupled component according to the coupled organization relationship to form a digital twin coupled body.

3. The method for multi-source business data correlation analysis based on digital twins according to claim 1, characterized in that, The specific steps involved in obtaining the digital twin association graph are as follows: Entity mapping is performed based on the digital twin coupling. Corresponding retrieval, cross-comparison, and merging and positioning are performed on each coupling component in the digital twin coupling along the coupling organizational relationship. Coupling components representing the same business object are mapped to the same digital twin entity, and coupling components with organizational relationships are mapped to the corresponding digital twin entities to form a set of digital twin entities. The digital twin entities in the digital twin entity set are paired up in pairs. For each entity pair, the organizational association results, co-occurrence results, action transmission results, and state linkage results between the corresponding coupling components are read. The association strength is calculated according to the preset association judgment rules. Entity pairs that meet the candidate edge construction conditions are determined as candidate association edges, forming a candidate association edge set. Multidimensional gating constraints are applied to the candidate association edge set. For each candidate association edge, gating judgments are performed from the perspectives of organizational association consistency, relationship transmission continuity, event effect closure, and state change coordination. The results of each gating judgment are then jointly screened. Candidate association edges that meet the retention conditions in the joint screening are determined as valid association edges, forming a set of valid association edges. The digital twin entities in the digital twin entity set are used as graph nodes, and the set of effective associated edges are used as connecting edges between nodes. The node connection relationship is established according to the entity pairing results corresponding to the effective associated edges, thus forming a digital twin association graph.

4. The method for multi-source business data correlation analysis based on digital twins according to claim 1, characterized in that, The construction of the Bron-Kerbosch maximal clique recursive search process specifically includes: Based on the digital twin association graph, the graph structure is read, and node adjacency relationships are established between each digital twin entity node and the digital twin entity nodes that maintain effective association edges, thus obtaining node adjacency relationship data; Node order is constructed based on node adjacency relationship data. The node index order is formed according to the registration order of digital twin entity nodes in the node adjacency relationship data. For each digital twin entity node in the node index order, the corresponding search primitive is extracted to form node search primitive data. Based on the node search primitive data, the initial search state set is constructed. Each digital twin entity node in the node index order is written into the corresponding search state unit to form an initial search state that corresponds one-to-one with each digital twin entity node. All initial search states are then organized into an initial search state set. Based on the initial search state set, a compatibility propagation recursive expansion is performed. From the range of nodes to be expanded corresponding to any initial search state, a digital twin entity node to be expanded is selected. The selected digital twin entity node to be expanded is incorporated into the current clique primitive to form an expanded clique node combination. The selected digital twin entity node to be expanded is used to propagate and update the edge compatibility records in the current search state to form an expanded search state. The expanded search state includes the expanded clique node combination, the continued expansion range, the continued exclusion range, and the updated edge compatibility records. The expanded search state is recursively continued. The digital twin entity nodes in the continued expansion range are repeatedly treated as digital twin entity nodes to be expanded and compatibility propagation recursive expansion is performed. When the continued expansion range is empty, there are no nodes in the continued exclusion range that can be incorporated into the current expansion clique combination, and the nodes maintain common edge compatibility with all digital twin entity nodes, the current expansion clique node combination is determined as a maximal clique. The recursive search states that have completed expansion and are formed corresponding to the initial search state set are organized into a Bron-Kerbosch maximal clique recursive search process.

5. The method for multi-source business data correlation analysis based on digital twins according to claim 1, characterized in that, The formation of the candidate group set specifically includes: For each recursive search state in the Bron-Kerbosch maximal clique recursive search process, the extended range is read, and the continued extended range in each recursive search state is read as the corresponding extended node set to form the candidate extended control input; Based on the candidate expansion control input, the candidate priority value is calculated. After the digital twin entity node to be expanded is merged into the expansion clique node combination, the number of nodes that still maintain common edge compatibility with the continued expansion range, the continued exclusion range, and the updated edge compatibility record are counted. The expansion retention value, exclusion constraint value, and compatibility propagation value are formed. The candidate priority value corresponding to each digital twin entity node to be expanded in the expansion node set is obtained according to the preset calculation method. The candidate node expansion order is constructed based on the candidate priority value. The candidate nodes to be expanded are arranged in order of priority from high to low to form the candidate node expansion order. The first candidate node to be expanded in the candidate node expansion order is determined as the priority expansion node in this round. The candidate node expansion order is written into the order control record of the corresponding recursive search state to form the priority control search state. Based on the priority control search state, the recursive expansion is performed in sequence. Based on the digital twin entity node to be expanded, the scope of continued expansion, the scope of continued exclusion, and the updated edge compatibility record are synchronously shrunk and updated to obtain the next layer recursive search state corresponding to the currently read digital twin entity node to be expanded. The remaining digital twin entity nodes to be expanded in the candidate node expansion order are retained as nodes to be processed in the current layer. Recursive continuation control is executed on each of the next-level recursive search states formed by the sequential recursive expansion process. The candidate priority value calculation process, the candidate node expansion order construction process, and the sequential recursive expansion process are repeatedly executed until the preset termination condition is reached. The current expansion clique node combination is recorded as a candidate clique, and all candidate cliques formed by the Bron-Kerbosch maximal clique recursive search process are organized into a candidate clique set.

6. The method for multi-source business data correlation analysis based on digital twins according to claim 1, characterized in that, The formation of the maximal clique recursive search result specifically includes: Perform cluster screening on the candidate cluster set, perform pairwise connectivity verification according to the connection status between the digital twin entity nodes in each candidate cluster, and retain the candidate clusters that meet the pairwise connectivity conditions as candidate clusters to be judged, forming a set of candidate clusters to be judged; Based on the set of candidate clusters to be determined, a business semantic closure determination is performed. For each candidate cluster in the set of candidate clusters to be determined, the associated content of the digital twin coupler corresponding to the candidate cluster is read to form semantic closure determination information. The semantic closure determination information is then matched with the preset business semantic closure conditions. Candidate clusters whose matching results meet the preset business semantic closure conditions are determined as target candidate clusters, thus forming a target candidate cluster set. The maximal clique recursive search process is updated based on the target candidate clique set. Each target candidate clique in the target candidate clique set is written back to the search state unit of the corresponding recursive search state. The state locking update is performed according to the search state corresponding to the target candidate clique. The extension branches that meet the preset business semantic closure conditions are retained in the maximal clique recursive search process, and the extension branches that do not meet the preset business semantic closure conditions are removed from the maximal clique recursive search process, forming the maximal clique recursive search result.

7. The method for multi-source business data correlation analysis based on digital twins according to claim 1, characterized in that, The formation of the local maxima identification result specifically includes: The state of digital twin entities in the digital twin entity set is monitored, the current state is compared with the historical state, digital twin entities whose state representation has changed are selected to form a state transition entity set, and the inversion source is determined based on the triggering effect of the state transition entities on the associated structure to form an inversion source set. Based on the set of inversion sources, the associated influence domain is reconstructed. Starting from each inversion source, the influence transmission is tracked along the node connection relationship in the digital twin association graph. The associated range affected by state transition is extracted and organized into associated influence domains corresponding to each inversion source, forming a set of associated influence domains. Local maximal clique identification is performed based on the set of associated influence domains. Local recursive expansion is performed in each associated influence domain according to the node connection relationship. The current expanded node combination is determined as a local maximal clique, forming the local maximal clique identification result.

8. The method for multi-source business data correlation analysis based on digital twins according to claim 1, characterized in that, The output of the correlation analysis results specifically includes: The clique results in the recursive search results of maximal cliques and the clique results in the local maximal clique identification results are written into the clique result set to form the clique set to be verified. Clique verification is performed on the clique set to be verified. Correspondence comparison, consistency screening, and conflict identification are performed on the clique results in the clique set to be verified. Clique results that meet the retention conditions are determined as valid clique results, and clique results with overlapping, cross or conflicting relationships are determined as clique results to be converged, thus forming a clique verification result set. Based on the clique verification result set, clique convergence processing is performed. The clique results to be converged are merged, overlaps are resolved, and conflicts are resolved. The converged clique results and the valid clique results are organized together into a set of associated core cliques, and the association analysis results are output based on the set of associated core cliques.

Images