Enterprise management risk monitoring method and system based on data transmission analysis

Abstract

The invention relates to the technical field of data transmission analysis, and particularly discloses an enterprise management risk monitoring method and system based on data transmission analysis. The method comprises the following steps: acquiring enterprise management element data in data transmission to construct a manifold space; analyzing the traceability task nodes, then carrying out compatibility analysis to identify misbidding nodes, determining whether a manifold space meets differential homeomorphism requirements or not through homeomorphism judgment coefficients, and if yes, constructing a manifold task chain; collecting Riemannian metric tensors of manifold task chain nodes to perform joint unbalance analysis, identifying an unbalance effect and calculating a deviation degree, and judging whether a tensor deformation alarm is triggered or not; positioning an initial disturbance point and a root responsible party, and outputting a fusing operation instruction set; the system comprises a manifold construction module, a homeomorphic analysis module, a deformation analysis module and an instruction fusing module. The method is beneficial to identifying the responsibility gap and redundancy, and improves the responsibility management standardability in enterprise data transmission.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data transmission analysis, and particularly relates to an enterprise management risk monitoring method and system based on data transmission analysis. BACKGROUND

[0002] In the process of enterprise digital management, the demand for management risk monitoring of enterprises in the data transmission scene is urgent, especially in the scene of data cross-department circulation and external sharing, it is necessary to ensure clear responsibility attribution and traceable risk to reduce data leakage, resource waste, compliance violations and other problems caused by chaotic responsibility.

[0003] However, the prior art lacks a dynamic structure carrier capable of depicting the correlation between multiple elements. In traditional management systems, department hierarchy, data sensitivity, and approval timeliness are stored in different modules, and only basic business information is carried during data transmission, and the responsibility information cannot be updated and expanded synchronously with the data flow, and the local correlation characteristics (such as the aggregation of high-sensitivity data and high-level departments) between elements cannot be preserved, making it difficult to identify the mismatch between responsibilities (such as the redundancy of high-level departments handling low-sensitivity internal log data) through structured analysis.

[0004] Therefore, the present application provides an enterprise management risk monitoring method and system based on data transmission analysis. SUMMARY

[0005] The present application aims to provide an enterprise management risk monitoring method and system based on data transmission analysis to solve the above background problems.

[0006] The purpose of the present application can be achieved by the following technical solutions: An enterprise management risk monitoring method based on data transmission analysis, comprising the following steps: Obtain enterprise management element data in data transmission and construct a data feature vector, construct a manifold space containing the data feature vector based on a manifold learning algorithm, and output a data stream of responsibility manifold labels; Obtain a core traceability hierarchy and analyze the traceability task nodes of the responsibility manifold labels; perform compatibility analysis on the traceability task nodes to obtain node compatibility curvature and identify missing nodes; adjust the responsibility of the missing nodes and perform homeomorphism judgment analysis to obtain a homeomorphism judgment coefficient to determine whether the manifold space meets the differential homeomorphism requirement, and if so, construct a manifold task chain; Collect the Riemann metric tensor of the nodes in the manifold task chain and perform joint imbalance analysis to identify whether the nodes have an imbalance effect, and if so, extract the deviation degree and determine whether to trigger a tensor deformation alarm; The proximate coordinate set triggering the deformation alarm is extracted, root cause positioning processing is performed based on the proximate coordinate set, an initial disturbance point is determined and a root cause responsible party is associated, demand analysis is performed on the initial disturbance point, it is judged whether a fuse instruction needs to be triggered, and if so, a fuse operation instruction set is output.

[0007] Further, the manner of outputting the responsibility manifold label is: A manifold space containing data feature vectors is constructed based on a manifold learning algorithm, node coordinates and Riemann metric tensors corresponding to the data feature vectors in the manifold space are extracted; A hash equation is constructed, the node coordinates and Riemann metric tensors are input into the hash equation, and a topological signature corresponding to the data feature vector is obtained; the topological signature and the node coordinates are combined to construct the responsibility manifold label of the element data.

[0008] Further, the manner of performing the compatibility analysis is: The investigation frequency of the data flow of different department levels is obtained from historical tracing and audit investigation, and the department level with the highest investigation frequency is taken as the core tracing level; The current data flow of the core tracing level is obtained, the coordinate points of the manifold space corresponding to the responsibility manifold label in the data flow are parsed, and the node Riemann metric tensors in the manifold space are parsed; The nodes in the manifold space involved in the core tracing level are taken as tracing task nodes, the coordinate points of the manifold space and the node Riemann metric tensors in the manifold space are input into the compatibility curvature equation, and the node compatibility curvature is obtained.

[0009] Further, the manner of performing the homeomorphism determination analysis is: The manifold space reconstructed after the responsibility adjustment of the missing target node is obtained, the edges commonly existing in the generated trees in the manifold spaces before and after reconstruction are extracted as coincident edges; The proportion of the coincident edges in the total number of edges in the manifold space after reconstruction is calculated as the edge coincidence rate; A single-edge difference equation is constructed, the weights of the edges before and after reconstruction of the manifold space are obtained, and the weights of the edges before and after reconstruction are input into the single-edge difference equation to obtain a single-edge difference rate; Based on the single-edge difference rate, an edge weight similarity analysis is performed to obtain a weight similarity degree, and the edge weight similarity degree and the edge coincidence rate are summed to obtain a homeomorphism determination coefficient; A differential structure criterion is established, and the homeomorphism determination coefficient is input into the differential structure criterion to obtain a determination result of whether the differential homeomorphism is satisfied.

[0010] Further, the manner of obtaining the missing target node is: The compatibility curvature of all the tracing task nodes is obtained, the missing target node is determined based on the compatibility curvature, and the missing target node of the tracing task node is obtained.

[0011] Further, the way of judging whether to trigger the deformation alarm of the tensor is: If the imbalance effect exists, the Riemann metric tensor of the nodes in the current manifold task chain is obtained, and a Riemann metric tensor matrix is constructed; The Riemann metric tensor matrix is input into the constructed Christoffel symbol equation to obtain the real-time connection coefficient of each node; The reference connection coefficient is obtained, a deviation degree equation is constructed, and the reference connection coefficient and the real-time connection coefficient are input into the deviation degree equation to obtain the deviation degree; a deviation degree range is established, and if the deviation degree is not in the deviation degree range, the deformation alarm of the tensor exceeding the threshold is triggered.

[0012] Further, the identification method of the imbalance effect is: All nodes of the core trace level in the data transmission manifold task chain are obtained, the Riemann metric tensor of each node in the monitoring period is collected, and the change rate of the Riemann metric tensor in the monitoring period is calculated as the management time derivative; If the management time derivative of the node is positive, the node is marked as a right and responsibility enhancement node, and if it is negative, it is marked as a right and responsibility weakening node; the enhancement intensity of the right and responsibility enhancement node is analyzed, and the weakening intensity of the right and responsibility weakening node is analyzed to obtain the enhancement intensity sum and the weakening intensity sum; The enhancement intensity sum and the weakening intensity sum are calculated by difference to obtain the joint imbalance index of the right and responsibility; Based on the joint imbalance index, comparison analysis is performed to obtain the result of whether the imbalance effect exists.

[0013] Further, the way of analyzing the initial disturbance point fuse demand is: The prevalent curvature scalar of the initial disturbance point is obtained through the Riemann curvature tensor contraction algorithm equation; The core nodes of the manifold task chain are screened, the geodesic distance between the core nodes and the initial disturbance node is calculated to obtain the core node distance; the risk radiation intensity is obtained by multiplying the core node distance and the curvature scalar; If the risk radiation intensity is in the highest level, the Gaussian curvature of the initial disturbance point is extracted, and whether the fuse instruction needs to be triggered is judged based on the Gaussian curvature.

[0014] Further, the way of obtaining the initial disturbance point is: The trigger alarm node and the adjacent node of the manifold space in the data transmission are obtained, and the geodesic line path from the trigger alarm node to the adjacent node is calculated through the geodesic line algorithm; The path deviation amount of the actual node connection path and different geodesic line paths is obtained, the geodesic line path with the highest path deviation amount is selected as the disturbance propagation main path; Based on the main path of disturbance propagation, the time stamps of different nodes joining the manifold space are obtained, and the nodes are extracted in reverse order of the time stamps as the backtracking nodes; The deviation of the current and initial state Riemann metric tensor of the backtracking node is calculated as the initial disturbance deviation, the norm of the initial disturbance deviation is calculated as the disturbance intensity, and the zero point mutation analysis is performed based on the extraction order of the backtracking node and combined with the disturbance intensity to extract the initial disturbance point.

[0015] An enterprise management risk monitoring system based on data transmission analysis, comprising the following modules: A manifold construction module is used to obtain enterprise management element data in data transmission and construct a data feature vector, construct a manifold space containing the data feature vector based on a manifold learning algorithm, and output a data stream of a responsibility manifold label; An isomorphism analysis module is used to obtain a core traceability level of historical traceability and analyze traceability task nodes of the responsibility manifold label, perform compatibility analysis on the traceability task nodes to obtain node compatibility curvature and identify missing label nodes, adjust the responsibility of the missing label nodes, and perform isomorphism determination analysis to obtain an isomorphism determination coefficient to determine whether the manifold space meets the differential isomorphism requirement, and if it meets the requirement, a manifold task chain is constructed; A deformation analysis module is used to collect node Riemann metric tensors in the manifold task chain and perform joint imbalance analysis to identify whether there is an imbalance effect, and if there is, extract a deviation and determine whether a deformation alarm of the tensor is triggered; An instruction fuse module is used to extract a nearby coordinate set triggering the deformation alarm, perform root cause positioning processing based on the nearby coordinate set, determine an initial disturbance point and associate a root cause responsible party, perform fuse demand analysis on the initial disturbance point, and determine whether a fuse instruction needs to be triggered, and if it needs to be triggered, output a fuse operation instruction set.

[0016] The beneficial effects of the present application are: The department level with the highest investigation frequency in historical traceability and audit investigation is taken as the core traceability level, the geometric features of the responsibility manifold label are analyzed to determine the traceability task nodes, the responsibility gaps and responsibility redundant nodes are identified through compatibility curvature, the missing label nodes are fine-tuned according to the minimum disturbance rule, and the isomorphism determination coefficient is used to realize that the manifold topological structure is not damaged, and finally a manifold task chain with consistent curvature is constructed. It is beneficial to realize the problems of poor level focus in traditional risk traceability, easy damage to management system in missing label node rectification, and inconsistent task flow responsibility, and reduces the compliance risks and resource waste caused by mismatched responsibility.

[0017] The Riemann metric tensor of the acquisition manifold task chain node is collected, the nodes are enhanced and weakened by calculating the management time derivative label responsibility, the joint imbalance index is obtained by the difference between the enhanced intensity sum and the weakened intensity sum to identify the imbalance effect, the deviation degree is calculated based on the Christoffel symbol, and whether the tensor deformation alarm is triggered is judged by combining the historical stable period range. It is beneficial to capture the change trend of the department level, data sensitivity and other responsibility characteristics, quantify the imbalance degree and deviation risk, early warning of tensor threshold deformation, reduce the management loopholes or data security risks caused by the accumulation of responsibility imbalance, and improve the real-time and accuracy of risk monitoring.

[0018] The adjacent coordinate set of the alarm triggering node is extracted, the disturbance propagation main path is determined by the geodesic algorithm, the initial disturbance point and the root responsibility party are located by combining the node timestamp backtracking and the Riemann metric tensor deviation, and the risk radiation intensity is calculated by the manifold curvature scalar, the fusing demand is judged by the Gaussian curvature, and the operation instruction set is output when fusing is needed. It is beneficial to realize root location, risk diffusion blocking and responsibility attribution, accurately decide whether to fuse based on the risk radiation range and local manifold mutation, block the risk diffusion path in time, reduce the influence range of the risk on the enterprise management system, and improve the efficiency and pertinence of risk disposal. BRIEF DESCRIPTION OF DRAWINGS

[0019] The application will be further described below with reference to the drawings.

[0020] Figure 1 is a flowchart of an enterprise management risk monitoring method based on data transmission analysis of the application; Figure 2 is a logic judgment diagram of whether the deformation alarm is triggered; Figure 3 is a module diagram of an enterprise management risk monitoring method based on data transmission analysis of the application. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the application.

[0022] Embodiment 1: Please refer to Figure 1 The application is an enterprise management risk monitoring method based on data transmission analysis, which comprises the following steps: S1. Obtain enterprise management element data in the data transmission and construct data feature vectors. Based on the manifold learning algorithm, construct a manifold space containing data feature vectors and output the data stream with responsibility manifold labels. The method for obtaining enterprise management element data during data transmission and constructing data feature vectors is as follows: Preferably, the enterprise management element data acquired during data transmission includes: personnel in charge at the department level, data sensitivity under different transmission scenarios, and data approval timeliness; Data feature vectors of enterprise management element data are constructed by performing data feature analysis on enterprise management element data. For example, the method of data feature generation for enterprise management element data is as follows: By using a pre-defined feature weight mapping table, the data sensitivity of department-level personnel in different transmission scenarios is characterized according to the weight mapping table. For example, the data feature weight of a certain person in charge of a department in the finance department is 0.85, while that of the operations and maintenance department is 0.3; the feature weight (sensitivity) of log data is 0.2, the feature weight of PII (Personal Privacy Data) data transmitted internally across departments is 1.0, and the feature weight of data shared with partners is 1.2. To address the timeliness of data approval, a time decay factor equation is established: Time decay factor for obtaining data approval timeliness The time decay factor is used as the timeliness of data approval; in, These represent the acceptance time and the data approval time, respectively, with the data approval time being later than the acceptance time. T1 is the standardized time. The data feature weights of department-level personnel, data sensitivity under different transmission scenarios, and data approval timeliness are integrated using min-max normalization to construct data feature vectors for enterprise management element data. Establish a feature vector set by acquiring the data feature vectors of all transmission tasks within the monitoring period; The method for constructing a manifold space containing data feature vectors based on the manifold learning algorithm and outputting a data stream with weighted manifold labels is as follows: The t-SNE dimensionality reducer performs dimensionality reduction on the feature vector set and maps the data feature vectors to each coordinate point in three-dimensional space. ; Those skilled in the art will understand that the dimensionality reduction parameters of the t-SNE dimensionality reducer are set as follows: perplexity is set to 30 (to balance local and global structures), and the number of iterations is 1000, to ensure that the local similarity of the original feature vectors is preserved after dimensionality reduction (such as the aggregation of highly sensitive and high-level vectors in three-dimensional space). For each coordinate point of the data feature vector in three-dimensional space Geometric feature extraction is performed to obtain the Riemann metric tensor for each coordinate point i; It should be noted that the method for geometric feature extraction is as follows: for each coordinate point Select the coordinates of the k nearest neighbors (e.g., k=10, adjusted according to the size of the vector set) to construct a local covariance matrix; Calculating the Riemannian metric tensor based on the covariance matrix Tensor elements reflect the degree of local stretching of coordinate points in the three dimensions of x, y, and z (e.g., if a certain region is significantly stretched in the x dimension, it indicates that the departmental level has a greater impact on the region's authority and responsibility). Riemannian metric tensor based on each coordinate point Through the formula: Get from coordinate point To coordinate point Spatial distance in three-dimensional space ; Where i and j are the coordinate point numbers; Based on the spatial distance between adjacent feature vectors in the feature vector set Construct a manifold space; Those skilled in the art will understand that the method of constructing a manifold space based on the spatial distance of the feature vector set is as follows: using each coordinate point in the coordinate set Q as a node and the distance as the edge weight, a weighted undirected graph is constructed; the minimum spanning tree algorithm is used to retain the core connections (removing redundant edges) to form a manifold topology that reflects the inherent relationship between rights and responsibilities (such as forming a dense subgraph for the transmission task of highly sensitive data), thereby realizing the construction of the manifold space; Based on node coordinates and the Riemann metric tensor, using the SM3 hash equation: Obtain the topological signature corresponding to the data feature vector; Combine topological signatures and node coordinates to construct a manifold label for the rights and responsibilities of the feature data; It is understandable that the purpose of constructing the rights and responsibilities manifold label is: Function 1: Enables the synchronous carrying and dynamic expansion of responsibility information during data transmission. The responsibility manifold label is constructed from a combination of topological signatures and node coordinates, and packaged with the transmitted data to form a data stream. When data flows between different departmental levels within an enterprise, the label expands synchronously with the information of the department involved. This helps solve the problem of scattered storage of responsibility elements such as data sensitivity and approval timeliness, which cannot be updated synchronously with data flow, ensuring that data carries complete responsibility association information when transmitted across departments and scenarios. Secondly, it provides core geometric features for the analysis and compatibility analysis of traceability task nodes. By analyzing the weight and responsibility manifold labels, geometric features such as coordinate points and node Riemannian metric tensors in the manifold space are extracted to determine the traceability task nodes. Subsequently, these features need to be input into the compatibility curvature equation to calculate the node compatibility curvature to identify missing nodes. The weight and responsibility manifold labels are the key data source for traceability task location and missing node identification, and the geometric features they carry directly determine the accuracy of traceability analysis.

[0023] Packaging is performed based on the rights and responsibilities manifold labels and the corresponding manifold labels to output a data stream containing rights and responsibilities manifold labels during data transmission; It should be noted that the data flow containing the authority and responsibility flow label is transferred between different departmental levels within the enterprise. As the information of the transferring department is updated, the authority and responsibility flow label is expanded accordingly. S2. Obtain the core tracing level of historical tracing and parse the tracing task nodes of the responsibility manifold labels; perform compatibility analysis on the tracing task nodes to obtain the node compatibility curvature and identify the missing labels; adjust the responsibility of the missing labels and perform homeomorphism judgment analysis to obtain the homeomorphism judgment coefficient to determine whether the manifold space meets the differential homeomorphism requirement. If it does, construct the manifold task chain. The method for obtaining the core tracing level of historical traceability and parsing the responsibility manifold labels of the tracing task nodes is as follows: Preferably, the frequency of data flow from different departmental levels is obtained from historical tracing and audit investigation, and the departmental level with the highest investigation frequency is taken as the core tracing level; Obtain the current data stream of the core traceability level and parse the geometric features corresponding to the rights and responsibilities manifold labels in the data stream; Among them, the geometric features include: the coordinates of the points in the manifold space, the Riemannian tensor of the nodes in the manifold space, the numbering of adjacent nodes in the manifold space, and the spatial distance between them; The nodes in the manifold space involved in the core tracing hierarchy are used as tracing task nodes; The method for performing compatibility analysis on tracing task nodes, obtaining node compatibility curvature, and identifying out-of-mark nodes is as follows: By constructing a compatible curvature equation: Obtain each traceability task node Node-compatible curvature; Understandably, node compatibility curvature quantifies the degree of geometric compatibility deviation between traceability task nodes and compliant nodes in the manifold space; node compatibility curvature serves as the core criterion for determining the rights and responsibilities of non-compliant nodes, while also verifying the effectiveness of adjusting the rights and responsibilities of non-compliant nodes, thus supporting the construction of the manifold task chain. in, To trace task nodes The three-dimensional coordinate vector, This is the mean vector of coordinates of normal nodes in the corresponding scenario. For normal nodes, the Riemannian metric tensor for The inverse matrix, where T is the transpose; It should be noted that a normal node is a normal node within the manifold space that complies with the rights and responsibilities (such as a node where a high-level department processes highly sensitive external data, a node where the approval timeliness meets the standard, or a node where the rights and responsibilities are successfully traced). Among them, adjusting the weights and responsibilities of the unmarked nodes and performing homeomorphism analysis to obtain the homeomorphism determination coefficient is the method for determining whether the manifold space satisfies the differential homeomorphism requirement: Obtain the compatibility curvature of all traceability task nodes, and determine the missing nodes based on the compatibility curvature; For example, the method for determining missing nodes is as follows: like (like =0.5), mark the traceability task node as a responsibility gap node (responsibility requirements are not met, such as low-level positions handling highly sensitive data); like (like =0.3), marking the traceability task node as a redundant node in terms of authority and responsibility (waste of authority and responsibility resources, such as high-level departments processing low-sensitivity internal data); in, These are the mean and variance of the compatibility curvature of normal nodes, respectively. Based on the minimum perturbation adaptation rule for injecting missing nodes, this method is used to fine-tune the weight and responsibility characteristics of missing nodes without destroying the topology of the manifold space, so that their curvature returns to the normal range. Understandably, the method for injecting the minimum perturbation adaptation rule is as follows: For nodes with positive misalignment: If the deviation is caused by insufficient job weight, the job weight of the node will be corrected to the original job weight plus the supplementary signature job weight, while retaining the original initiation record. If the deviation is caused by a mismatch in data sensitivity scenarios (e.g., using an external sensitivity weight of 1.2 for an internal transmission scenario), the minimum disturbance adaptation rule will correct the sensitivity weight to the standard value for the corresponding scenario (e.g., correcting it to 1.0 for an internal scenario). For nodes with negative misalignment: if the deviation is caused by a waste of high-level resources (e.g., a headquarters department processing log data with a sensitivity of 0.2), the adapter triggers the responsibility decentralization logic: correcting the departmental level weight to the original level weight × 0.8; setting a maximum disturbance threshold ensures that the core responsibility attributes of the node remain unchanged after fine-tuning. For nodes that have been injected with the minimum perturbation adaptation rule, the policy compatibility curvature is recalculated. fall into[ , If the adaptation is successful, then the adaptation rules are adjusted repeatedly (e.g., adding a supplementary signing level) until the curvature meets the requirements. The manifold space is reconstructed by injecting the unmarked node with the minimum perturbation adaptation rule and other normal nodes with the same period, and the edges that exist in the spanning tree in the manifold space before and after reconstruction are obtained as overlapping edges. Calculate the proportion of overlapping edges in the total number of edges in the reconstructed manifold space, and use it as the edge overlap rate; Using Equation 1: One-sided Difference Rate = Obtain the one-sided difference rate; in, , These are the weights before and after the reconstruction of the manifold space, respectively. Edge weight similarity can be obtained using Equation 1: Edge weight similarity = 1 - E; Where E is the mean of the one-sided difference rate of all overlapping edges in the overlapping edge; The similarity of edge weights and the overlap rate of edges are summed to obtain the homeomorphism determination coefficient; The homeomorphism determination coefficient is compared with the preset homeomorphism determination threshold. If the homeomorphism determination coefficient is higher than or equal to the preset homeomorphism determination threshold, the topological structure is determined to be intact. Then, a differential structure criterion is established to obtain the determination result of whether the differential homeomorphism is satisfied. It should be noted that the method for establishing the differential structure criterion is as follows: for each node of the reconstructed manifold, select k nearest neighbors (k=10, adjusted according to the node size) to construct a neighborhood mapping; calculate the Jacobian matrix J of the neighborhood mapping, find the absolute value of the determinant of the Jacobian matrix |det(J)|, and take the mean of |det(J)| for all nodes; if the mean of det(J)| is in [0.9, 1.1], then the tangent space isomorphism requirement of the differential homeomorphism is satisfied, denoted as D. diff =1; otherwise D diff =0; The homeomorphism criterion coefficient is compared with D. diff Product processing; if the product result is higher than the preset 1.6, the manifold space is determined to satisfy the differential homeomorphism requirement; The purpose of determining whether a manifold satisfies the requirement of differential homeomorphism is as follows: Objective 1: To ensure the consistency of the responsibility and accountability logic within the manifold task chain, minimizing disruption to the original management system caused by adjustments to missing nodes. Differential homeomorphism requires that the manifold space, after adjustment, maintains both topological similarity and differential structural compatibility. The core objective of this determination is to ensure that fine-tuning of missing nodes does not disrupt the inherent relationship logic between responsibility and accountability elements in the manifold space. This ensures that the final constructed manifold task chain possesses consistent curvature and stable responsibility and accountability relationships, reducing fragmentation of the management system due to node adjustments and providing a unified benchmark for subsequent imbalance analysis.

[0024] Objective 2: To ensure the accuracy of quantitative indicators for subsequent joint imbalance analysis and root cause localization. Both joint imbalance analysis and root cause localization rely on the geometric characteristics of the manifold space. Differential homeomorphism ensures the stability of the local geometric properties of the manifold space, meaning that the differential structure and local distance relationships of the manifold are not distorted due to node adjustments. If differential homeomorphism is not satisfied, the geometric characteristics of the manifold will be distorted, leading to errors in the calculation of indicators such as management time derivative, deviation, and risk radiation intensity. This, in turn, can cause misjudgments of imbalance effects and deviations in the location of initial disturbance points, ultimately affecting the accuracy of risk monitoring.

[0025] If the homeomorphism determination coefficient is lower than the preset homeomorphism determination threshold, then the differential homeomorphism requirement is not met. If the requirement of differential homeomorphism is met, then the core tracing hierarchy constructs a manifold task chain with consistent curvature; Example 2: like Figure 1 As shown, the present invention is an enterprise management risk monitoring method based on data transmission analysis, which further includes the following steps: S3. Obtain all nodes of the core traceability hierarchical manifold task chain, collect the Riemannian metric tensor of the nodes in the manifold task chain and perform joint imbalance analysis, identify whether there is an imbalance effect in the nodes, and if so, extract the deviation and determine whether to trigger the deformation alarm of the tensor. The method for identifying whether nodes exhibit imbalance effects is as follows: This involves acquiring all nodes of the core tracing hierarchy manifold task chain, collecting the Riemannian metric tensors of the nodes in the manifold task chain, performing joint imbalance analysis, and identifying the nodes' imbalance effects. Preferably, all nodes of the core traceability level in the manifold task chain during data transmission are obtained, and the Riemann metric tensor of each node is collected within the monitoring period. The rate of change of the Riemann metric tensor during the monitoring period is calculated as the management time derivative. It should be noted that the larger the absolute value of the management time derivative of each node, the faster the change in the weight and responsibility characteristics such as department level and data sensitivity (e.g., a management time derivative of 0.2 / hour means that the weight of a certain dimension increases by 0.2 per hour). If the derivative of the management time of a node is positive, the node is marked as a node with enhanced authority and responsibility (e.g., the data sensitivity weight increases); if it is negative, it is marked as a node with weakened authority and responsibility (e.g., the approval timeliness factor decreases). For each responsibility enhancement node, obtain the corresponding data feature vector and calculate the enhancement strength coefficient of each dimension in the data feature vector; The reinforcement strength coefficients of all the responsibility-enhancing nodes are summed to obtain the reinforcement strength sum; It should be noted that the enhancement intensity coefficient = the absolute value of the management time derivative × the weight of the management time derivative dimension (preset weights: department level management time derivative = management time derivative 0.4, data sensitivity management time derivative = management time derivative 0.3, approval timeliness management time derivative = management time derivative 0.3). Weakening strength coefficient = absolute value of the derivative of node management time × corresponding dimension weight; For example: if the time derivative of the data sensitivity management of a certain node is +0.2 (enhanced), then the enhancement strength coefficient = 0.2 × 0.3 = 0.06; if the time derivative of the approval timeliness management of a certain node is -0.1 (weakened), then the weakening strength coefficient = 0.1 × 0.3 = 0.03. For each responsibility weakening node, obtain the corresponding data feature vector, calculate the weakening strength coefficient of each dimension in the data feature vector, and sum the weakening strength coefficients of the full responsibility weakening nodes to obtain the weakening strength sum. The difference between the strengthening and weakening strengths is calculated to obtain the joint imbalance index of weights and responsibilities; If the joint imbalance index of rights and responsibilities is not within the preset imbalance range, then there is an imbalance effect in the nodes of the manifold task chain; If the joint imbalance index of rights and responsibilities is within the preset imbalance range, the changes in the joint imbalance index will be continuously monitored. The method for extracting the deviation and determining whether to trigger a tensor deformation alarm, if present, is as follows: If an imbalance effect exists, obtain the Riemann metric tensor of the node in the current manifold task chain and construct the Riemann metric tensor matrix; Through Christofel's symbolic equation: Obtain the real-time communication coefficient for each node. ; Understandably, the real-time communication coefficient is a core geometric quantity that reflects the rate of change of local associations between the dimensions of authority and responsibility elements in the manifold space. By combining the partial derivatives of the Riemann metric tensor, it quantifies the local geometric communication relationships between the three-dimensional coordinate dimensions corresponding to authority and responsibility elements such as departmental level, data sensitivity, and approval timeliness in the manifold space, and reflects whether the current node's authority and responsibility association status conforms to normal local geometric rules. in, , , The elements of the Riemannian metric tensor G reflect the local stretching and correlation strength of the dimensions of authority and responsibility elements such as departmental hierarchy, data sensitivity, and approval timeliness in the manifold space. Let be the inverse of the Riemannian metric tensor matrix. , , Let I be the coordinate variable in the manifold space, and let I be the contravariant index, with subscripts a and b being covariant indices. The sign of the partial derivative. A dummy index (traversing 1, 2, 3, corresponding to the x, y, z axes); By establishing the deviation equation: Obtain the deviation D; in, The norm of the preset baseline connection coefficient, with the norm type being L2; Obtain the deviation from historical stable periods and establish the deviation range. ; The comparison and analysis based on the deviation range are used to determine whether the node's tensor over-threshold deformation alarm is triggered.

[0026] Preferably, the comparative analysis and judgment method is as follows: if the deviation is within the deviation range, the deformation alarm of tensor exceeding the threshold is not triggered, and the change of deviation is continuously monitored; if the deviation is not within the deviation range, the deformation alarm of tensor exceeding the threshold is triggered.

[0027] S4. Extract the set of nearby coordinates that trigger the deformation alarm, perform root cause localization processing based on the set of nearby coordinates, determine the initial disturbance point and associate it with the root cause responsible party, analyze the circuit breaker requirement of the initial disturbance point, determine whether it is necessary to trigger the circuit breaker command, and if so, output the set of circuit breaker operation commands. The method for extracting the nearest coordinate set that triggers the deformation alarm and performing root cause localization based on the nearest coordinate set is as follows: Preferably, the triggering alarm node and neighboring nodes in the manifold space during data transmission are obtained, and a neighboring coordinate set containing the triggering alarm node and neighboring nodes is constructed; The geodesic algorithm is used to calculate the geodesic path from the alarm trigger node to the neighboring nodes. Understandably, the geodesic algorithm uses Dijkstra's algorithm; Obtain the path deviation between the actual node connection path and different geodesic paths, and select the geodesic path with the highest path deviation as the main path for disturbance propagation. Based on the main path of perturbation propagation, the timestamps of different nodes joining the manifold space are obtained and the nodes are extracted in reverse order of the timestamps as backtracking nodes. Calculate the deviation between the current and initial states of the backtracking node's Riemann metric tensor, and use it as the initial perturbation deviation; The norm of the initial perturbation deviation is calculated as the perturbation intensity. Based on the extraction order of the backtracking nodes, the node closest to the point where the perturbation intensity abruptly changes from non-zero to zero is obtained as the initial perturbation point. Extract the data feature vector corresponding to the initial disturbance point, identify the responsible personnel at the department level corresponding to the data feature vector, and realize the responsibility of the root cause of the failure of the management element; The method for analyzing the initial disturbance point circuit breaker demand and determining whether a circuit breaker command needs to be triggered is as follows: Using the Riemann curvature tensor shrinkage algorithm equation: Obtain the popular curvature scalar R of the initial perturbation point; in, For the Ricci curvature tensor, the larger the value of the popular curvature scalar, the more severe the local bending of the node, and the easier it is for the risk to spread; Select the core nodes of the manifold task chain, calculate the geodesic distance between the core nodes and the initial perturbation nodes, and obtain the core node distance; The risk radiation intensity is obtained by multiplying the core node distance by the curvature scalar. Different circuit breaker decision levels are divided based on the magnitude of the risk radiation intensity. If it is at the highest level, the Gaussian curvature of the initial disturbance point is extracted, and the circuit breaker command is determined based on the Gaussian curvature. For example, the different circuit breaker decision levels are determined as follows: The top 5% of core nodes in the manifold task chain connectivity are selected. The geodesic distance between the initial disturbance point and the core node (e.g., 8) and the manifold curvature scalar of the disturbance point (e.g., 7.5) are calculated. The product of these two values ​​yields a risk radiation intensity of 60. Based on historical cases, a maximum threshold of >50 is set, with 60 falling into the highest level. The Gaussian curvature of the disturbance point's neighborhood is then calculated to be 0.65. Comparing this to the historical threshold of 0.6, if 0.65 ≥ 0.6, the highest level circuit breaker is triggered, and an operation command is output. If a circuit breaker instruction needs to be triggered, a set of circuit breaker operation instructions will be generated and output to prevent the spread of enterprise management risks. The method for extracting the Gaussian curvature of the initial perturbation point is as follows: Extract the local manifold in the neighborhood of the initial perturbation point, and extract the Gaussian curvature using the Gaussian curvature algorithm; It should be noted that the selection rule for the neighborhood of the initial perturbation point is based on the coordinates of the initial perturbation point. Centered on the x-axis, select one adjacent node in each of the positive and negative directions of the x, y, and z axes, forming a local manifold (including the central node) with a total of 7 nodes. Extract historical circuit breaker cases from enterprise management, set a Gaussian curvature threshold, and if the Gaussian curvature is higher than or equal to the Gaussian curvature threshold, it indicates that there is a fold-like abrupt change in the local manifold and triggers a circuit breaker command. If the Gaussian curvature is lower than the Gaussian curvature threshold, the circuit breaker will not be tripped and enhanced monitoring will be initiated.

[0028] Example 3: like Figure 3 As shown, this invention is an enterprise management risk monitoring system based on data transmission analysis, comprising the following modules: Manifold Construction Module: Used to acquire enterprise management element data in data transmission and construct data feature vectors, construct a manifold space containing data feature vectors based on manifold learning algorithm, and output data stream with weight and responsibility manifold labels; Homeomorphism Analysis Module: Used to obtain the core tracing level of historical tracing and parse the tracing task nodes of the responsibility manifold labels; perform compatibility analysis on the tracing task nodes to obtain the node compatibility curvature and identify the missing labels; adjust the responsibility of the missing labels and perform homeomorphism judgment analysis to obtain the homeomorphism judgment coefficient to determine whether the manifold space meets the differential homeomorphism requirement. If it does, construct the manifold task chain. Deformation Analysis Module: Used to obtain all nodes of the core traceability hierarchical manifold task chain, collect the Riemannian metric tensors of the nodes in the manifold task chain and perform joint imbalance analysis, identify whether there is an imbalance effect in the nodes, and if so, extract the deviation and determine whether to trigger the deformation alarm of the tensor. Command Circuit Breaker Module: Used to extract the nearest coordinate set that triggers deformation alarm, perform root cause localization processing based on the nearest coordinate set, determine the initial disturbance point and associate it with the root cause responsible party, analyze the circuit breaker requirement of the initial disturbance point, determine whether a circuit breaker command needs to be triggered, and if so, output a set of circuit breaker operation commands.

[0029] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A method for monitoring enterprise management risks based on data transmission analysis, characterized in that: Includes the following steps: The system acquires enterprise management element data from data transmission and constructs data feature vectors. Based on the manifold learning algorithm, it constructs a manifold space containing the data feature vectors and outputs a data stream with weight and responsibility manifold labels. The core traceability hierarchy is obtained and the traceability task nodes with the rights and responsibilities manifold labels are parsed; the traceability task nodes are subjected to compatibility analysis to obtain the node compatibility curvature and identify the missing nodes; the rights and responsibilities of the missing nodes are adjusted and homeomorphism analysis is performed to obtain the homeomorphism determination coefficient to determine whether the manifold space meets the differential homeomorphism requirement. If it does, the manifold task chain is constructed. Collect the Riemannian metric tensors of nodes in the manifold task chain and perform joint imbalance analysis to identify whether there is an imbalance effect in the nodes. If there is, extract the deviation and determine whether to trigger the deformation alarm of the tensor. Extract the set of nearest coordinates that trigger the deformation alarm, perform root cause localization processing based on the set of nearest coordinates, determine the initial disturbance point and associate it with the root cause responsible party, analyze the circuit breaker requirement of the initial disturbance point, determine whether it is necessary to trigger the circuit breaker command, and if so, output the set of circuit breaker operation commands.

2. The enterprise management risk monitoring method based on data transmission analysis according to claim 1, characterized in that: The method for outputting the responsibility manifold labels is as follows: Based on the manifold learning algorithm, a manifold space containing data feature vectors is constructed, and the node coordinates and Riemann metric tensors corresponding to the data feature vectors in the manifold space are extracted. Construct a hash equation, input the node coordinates and Riemannian metric tensor into the hash equation, and obtain the topological signature corresponding to the data feature vector; combine the topological signature and node coordinates to construct the weight and responsibility manifold label of the feature data.

3. The enterprise management risk monitoring method based on data transmission analysis according to claim 1, characterized in that: The compatibility analysis is performed as follows: The frequency of data flow from different departmental levels is obtained from historical tracing and audit investigations, and the departmental level with the highest investigation frequency is taken as the core tracing level; Obtain the current data stream of the core traceability level, and parse the coordinates of the manifold space corresponding to the rights and responsibilities manifold labels in the data stream, as well as the Riemann tensor of the nodes in the manifold space; The nodes in the manifold space involved in the core tracing level are taken as tracing task nodes. By constructing a compatible curvature equation, the coordinate points of the manifold space and the Riemann tensor of the nodes in the manifold space are input into the compatible curvature equation to obtain the node compatible curvature.

4. The enterprise management risk monitoring method based on data transmission analysis according to claim 1, characterized in that: The method for performing the homeomorphism determination analysis is as follows: Obtain the reconstructed manifold space after adjusting the weights and responsibilities of the missing nodes, and extract the edges that exist in the spanning trees in the manifold space before and after reconstruction as overlapping edges; Calculate the proportion of overlapping edges in the total number of edges in the reconstructed manifold space, and use it as the edge overlap rate; Construct a one-sided difference equation, obtain the weights before and after reconstruction of the manifold space, input the weights before and after reconstruction into the one-sided difference equation, and obtain the one-sided difference rate. Based on the unilateral difference rate, the weight similarity is obtained through edge weight similarity analysis. The edge weight similarity and edge overlap rate are summed to obtain the homeomorphism determination coefficient. Establish a differential structure criterion, and input the homeomorphism criterion to obtain the determination result of whether the differential homeomorphism is satisfied.

5. The enterprise management risk monitoring method based on data transmission analysis according to claim 4, characterized in that: The method for obtaining the missing nodes is as follows: Obtain the compatibility curvature of all traceability task nodes, and determine the missing nodes based on the compatibility curvature to obtain the missing nodes of the traceability task nodes.

6. The enterprise management risk monitoring method based on data transmission analysis according to claim 1, characterized in that: The method for determining whether the tensor deformation alarm has been triggered is as follows: If an imbalance effect is identified in the node, the Riemann metric tensor of the node in the current manifold task chain is obtained, and the Riemann metric tensor matrix is ​​constructed. By inputting the Riemann metric tensor matrix into the constructed Christofel symbolic equation, the real-time communication coefficient of each node is obtained. Obtain the baseline connection coefficient, construct the deviation equation, and input the baseline connection coefficient and real-time connection coefficient into the deviation equation to obtain the deviation. Establish the deviation range. If the deviation is not within the deviation range, trigger the deformation alarm of the tensor exceeding the threshold.

7. The enterprise management risk monitoring method based on data transmission analysis according to claim 6, characterized in that: The method for identifying the existence of the aforementioned imbalance effect is as follows: Obtain all nodes in the manifold task chain of the core traceability level in data transmission, collect the Riemann metric tensor of each node within the monitoring period, and calculate the rate of change of the Riemann metric tensor within the monitoring period as the management time derivative. If the derivative of a node's management time is positive, the node is marked as a node with enhanced responsibility; if it is negative, it is marked as a node with weakened responsibility. Strengthening analysis is performed on nodes that enhance rights and responsibilities, and weakening analysis is performed on nodes that weaken rights and responsibilities, resulting in the sum of strengthening and weakening strengths; The difference between the strengthening and weakening strengths is calculated to obtain the joint imbalance index of weights and responsibilities; A comparative analysis based on the joint imbalance index was conducted to determine whether the aforementioned imbalance effect exists.

8. The enterprise management risk monitoring method based on data transmission analysis according to claim 1, characterized in that: The method for analyzing the circuit breaker demand at the initial disturbance point is as follows: The popular curvature scalar of the initial perturbation point is obtained through the Riemann curvature tensor shrinkage algorithm equation. The core nodes of the manifold task chain are selected, and the geodesic distance between the core nodes and the initial perturbation nodes is calculated to obtain the core node distance. The core node distance is then multiplied by the curvature scalar to obtain the risk radiation intensity. If the risk radiation intensity is at the highest level, the Gaussian curvature of the initial disturbance point is extracted, and the decision on whether to trigger the circuit breaker command is based on the Gaussian curvature.

9. The enterprise management risk monitoring method based on data transmission analysis according to claim 8, characterized in that: The method for obtaining the initial perturbation point is as follows: Obtain the trigger alarm node and neighboring nodes in the manifold space during data transmission, and calculate the geodesic path from the trigger alarm node to the neighboring node using the geodesic algorithm; Obtain the path deviation between the actual node connection path and different geodesic paths, and select the geodesic path with the highest path deviation as the main path for disturbance propagation. Based on the main path of perturbation propagation, the timestamps of different nodes joining the manifold space are obtained and the nodes are extracted in reverse order of the timestamps as backtracking nodes. Calculate the deviation between the current and initial states of the backtracking node's Riemann metric tensor, and use it as the initial perturbation deviation; The norm of the initial disturbance deviation is calculated as the disturbance intensity. Based on the extraction order of the backtracking nodes and combined with the disturbance intensity, zero-point mutation analysis is performed to extract the initial disturbance point.

10. An enterprise management risk monitoring system based on data transmission analysis, used to implement any one of the enterprise management risk monitoring methods based on data transmission analysis as described in claims 1-9, characterized in that: Includes the following modules: Manifold Construction Module: Used to acquire enterprise management element data in data transmission and construct data feature vectors, construct a manifold space containing data feature vectors based on manifold learning algorithm, and output data stream with weight and responsibility manifold labels; Homeomorphism Analysis Module: Used to obtain the core tracing level of historical tracing and parse the tracing task nodes of the responsibility manifold labels; perform compatibility analysis on the tracing task nodes to obtain the node compatibility curvature and identify the missing labels; adjust the responsibility of the missing labels and perform homeomorphism judgment analysis to obtain the homeomorphism judgment coefficient to determine whether the manifold space meets the differential homeomorphism requirement. If it does, construct the manifold task chain. Deformation Analysis Module: Used to collect Riemannian metric tensors of nodes in the manifold task chain and perform joint imbalance analysis to identify whether there is an imbalance effect in the node. If there is, the deviation is extracted and it is determined whether to trigger the deformation alarm of the tensor. Command Circuit Breaker Module: Used to extract the nearest coordinate set that triggers deformation alarm, perform root cause localization processing based on the nearest coordinate set, determine the initial disturbance point and associate it with the root cause responsible party, analyze the circuit breaker requirement of the initial disturbance point, determine whether a circuit breaker command needs to be triggered, and if so, output a set of circuit breaker operation commands.

Images