Multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection

Abstract

The application relates to the technical field of industrial internet, and discloses a multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection, which comprises the following steps: establishing a reliability verification matrix of multi-dimensional procurement data, and extracting abnormal data points; identifying the cross-correlation mode of the abnormal data points, and constructing a procurement behavior interaction graph of a target enterprise; analyzing the risk evolution path of a procurement risk factor, and calculating a risk transmission entropy value; generating a risk classification label of the procurement risk factor according to the risk transmission entropy value, so as to output a procurement anomaly detection report of the target enterprise. The application can improve the accuracy of enterprise procurement anomaly detection.

Description

TECHNICAL FIELD

[0001] The application relates to a multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection, and belongs to the technical field of industrial internet. BACKGROUND

[0002] Enterprise procurement anomaly detection refers to a process of identifying abnormal procurement behaviors by collecting various data (such as procurement price, procurement quantity, supplier qualification, delivery time, etc.) in the enterprise procurement process in real time, combining enterprise procurement systems and historical transaction rules. The process covers price deviation analysis, quantity abnormality monitoring, supplier risk assessment and other measures, which can effectively reduce procurement costs, prevent procurement risks and ensure the compliance of procurement processes.

[0003] However, the traditional enterprise procurement anomaly detection method mainly uses fixed risk lists and static threshold standards for risk judgment. Although this method can realize basic data aggregation and simple risk identification, it cannot realize automatic association and integration of cross-business data and dynamic optimization of risk assessment thresholds, resulting in insufficient accuracy of enterprise procurement anomaly detection.

[0004] Therefore, a solution is needed to improve the accuracy of enterprise procurement anomaly detection. SUMMARY

[0005] The application provides a multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection, which mainly aims to improve the accuracy of enterprise procurement anomaly detection.

[0006] To achieve the above purpose, the application provides a multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection, which comprises:

[0007] Collecting multi-dimensional procurement data of a target enterprise, establishing a reliability verification matrix of the multi-dimensional procurement data, and extracting abnormal data points in the reliability verification matrix, wherein the multi-dimensional procurement data includes internal procurement process data, supplier transaction data and market trend data;

[0008] Identifying the cross-correlation mode of the abnormal data points, and constructing a procurement behavior interaction graph of the target enterprise based on the cross-correlation mode;

[0009] According to the procurement behavior interaction graph, identifying the procurement risk hotspots of the target enterprise, and detecting the cross-dimensional abnormal mode of the target enterprise;

[0010] Based on the procurement risk hotspots and the cross-dimension abnormal patterns, a procurement risk factor of the target enterprise is extracted, and an evolution path of the procurement risk factor is analyzed, and according to the evolution path, a risk transmission entropy value of the procurement risk factor is calculated;

[0011] According to the risk transmission entropy value, a risk classification label of the procurement risk factor is generated, and a procurement abnormality detection report of the target enterprise is output.

[0012] Optionally, the establishment of the reliability verification matrix of the multi-dimension procurement data comprises:

[0013] A target procurement subject is determined through the multi-dimension procurement data;

[0014] Price data, supplier qualification data and delivery time limit data are extracted from the multi-dimension procurement data, and real-time execution data of the target procurement subject is obtained;

[0015] The price data is used to generate a price fluctuation index of the target procurement subject;

[0016] The credit rating index in the supplier qualification data and the price fluctuation index are subjected to nonlinear fitting processing to output a supplier basic credit score;

[0017] Market supply and demand data of an industry in which the target procurement subject is located are collected, and the delivery fluctuation coefficient corresponding to the target procurement subject is calculated in combination with the delivery time limit data and the market supply and demand data;

[0018] Based on the real-time execution data, an execution deviation rate of the target procurement subject is identified;

[0019] According to the delivery fluctuation coefficient and the execution deviation rate, a dynamic execution risk coefficient of the target procurement subject is calculated;

[0020] The supplier basic credit score, the dynamic execution risk coefficient and the price fluctuation index are combined to establish the reliability verification matrix of the multi-dimension procurement data.

[0021] Optionally, the extraction of the abnormal data points in the reliability verification matrix comprises:

[0022] Adjacent index items and feature dimensions in the reliability verification matrix are extracted;

[0023] A normalized data difference degree of the adjacent index items is calculated;

[0024] A logical association degree between the feature dimensions is identified, and a collaborative deviation degree between the feature dimensions is calculated based on the logical association degree;

[0025] calculating an abnormal point density of the feature dimension in the reliability verification matrix;

[0026] determining an abnormal accumulation index of the feature dimension based on the normalized data difference degree, the cooperative deviation degree and the abnormal point density;

[0027] setting an abnormal threshold of the abnormal accumulation index, and determining a potential abnormal data area of the multi-dimensional procurement data by using the abnormal threshold;

[0028] identifying a benchmark abnormal point of the multi-dimensional procurement data from the potential abnormal data area;

[0029] identifying a pseudo abnormal point of the benchmark abnormal point, and performing a rejection processing of the pseudo abnormal point to obtain an abnormal data point of the multi-dimensional procurement data.

[0030] Optionally, the constructing the procurement behavior interaction graph of the target enterprise based on the cross-correlation pattern comprises:

[0031] locating a core interaction node corresponding to the cross-correlation pattern, and identifying an associated dimension and its upstream and downstream dimensions of the core interaction node;

[0032] calculating an associated strength value and a conduction time limit of the associated dimension and the upstream and downstream dimensions;

[0033] determining an associated influence weight of the associated dimension according to the associated strength value and the conduction time limit;

[0034] identifying a key conduction path corresponding to the core interaction node based on the associated influence weight;

[0035] setting a visual association line of the cross-correlation pattern according to the key conduction path;

[0036] constructing the procurement behavior interaction graph of the target enterprise in combination with the core interaction node, the associated influence weight and the visual association line.

[0037] Optionally, the identifying a procurement risk hotspot of the target enterprise according to the procurement behavior interaction graph comprises:

[0038] analyzing an interaction feature of the procurement behavior interaction graph;

[0039] identifying a node association weight and a historical abnormal record corresponding to the interaction feature;

[0040] extracting a key path distribution and a business dependence strength in the node association weight;

[0041] analyze a trigger event type of the historical abnormal record, and identify a risk transmission path corresponding to the trigger event type;

[0042] determine a risk hotspot density of the procurement behavior interaction graph based on the critical path distribution and the risk transmission path;

[0043] identify a potential impact range of the risk transmission path;

[0044] determine a procurement risk hotspot of the target enterprise in combination with the risk hotspot density, the potential impact range, and the business dependency strength.

[0045] Optionally, the detecting a cross-dimension abnormal pattern of the target enterprise according to the procurement behavior interaction graph comprises:

[0046] identifying a cross-business domain abnormal transmission link and a normal business link in the procurement behavior interaction graph;

[0047] extracting a business activity sequence corresponding to the cross-business domain abnormal transmission link and a business domain feature thereof;

[0048] calculating a business coupling deviation and a risk diffusion delay degree between the cross-business domain abnormal transmission link and the normal business link;

[0049] determining a risk transmission direction of the cross-business domain abnormal transmission link based on the business coupling deviation and the risk diffusion delay degree;

[0050] detecting the cross-dimension abnormal pattern of the target enterprise in combination with the risk transmission direction, the business activity sequence, and the business domain feature.

[0051] Optionally, the extracting a procurement risk factor of the target enterprise based on the procurement risk hotspot and the cross-dimension abnormal pattern comprises:

[0052] extracting a risk aggregation strength of the procurement risk hotspot, and performing an impact range definition process of the cross-dimension abnormal pattern to obtain a cross-domain impact radius;

[0053] defining a risk feature matrix of the procurement risk factor according to the risk aggregation strength and the cross-domain impact radius;

[0054] generating a risk factor candidate pool of the target enterprise based on the risk feature matrix;

[0055] calculating a coupling coefficient between factors in the risk factor candidate pool and a business weight of the factors;

[0056] filtering out a core risk factor from the risk factor candidate pool through the coupling coefficient and the business weight;

[0057] constructing a risk assessment framework of the target enterprise based on the coupling coefficient and the core risk factor;

[0058] setting an optimization trigger threshold of the risk assessment framework according to the hotspot migration rate of the procurement risk hotspot and the mode variation frequency of the cross-dimension abnormal mode;

[0059] extracting a procurement risk factor of the target enterprise based on the risk assessment framework and the optimization trigger threshold.

[0060] Optionally, the risk evolution path of the procurement risk factor is analyzed based on the procurement risk hotspot and the cross-dimension abnormal mode, including:

[0061] identifying a risk transmission sensitive area of the cross-dimension abnormal mode based on the procurement risk hotspot;

[0062] setting a dynamic tracking module of the procurement risk factor in the risk transmission sensitive area;

[0063] capturing real-time state data of the procurement risk factor in different business links through the dynamic tracking module;

[0064] configuring a risk diffusion monitor of the procurement risk factor according to the dynamic tracking module and the real-time state data;

[0065] analyzing a risk evolution trend of the procurement risk factor based on the risk diffusion monitor;

[0066] outputting a risk evolution path of the procurement risk factor according to the risk evolution trend.

[0067] Optionally, the risk transmission entropy value of the procurement risk factor is calculated according to the risk evolution path, including:

[0068] locating a path starting point of the procurement risk factor in the risk evolution path and obtaining a risk trigger time stamp of the path starting point;

[0069] determining a risk exposure duration of the procurement risk factor based on the risk trigger time stamp;

[0070] identifying a transmission link of the procurement risk factor in the risk evolution path according to the path starting point;

[0071] extracting a risk quantification index of the procurement risk factor and calculating an index deviation rate of the risk quantification index in the transmission link;

[0072] calculating a link correlation coefficient between the transmission links;

[0073] The risk transmission entropy value of the risk evolution path is calculated by the following formula in combination with the risk exposure duration, the link correlation coefficient and the index deviation rate:

[0074]

[0075] wherein, represents the risk transmission entropy value of the procurement risk factor, represents the reference entropy value of the procurement risk factor, represents the weight of the i-th risk quantization index, represents the index deviation rate of the i-th risk quantization index, represents the maximum index deviation rate, n represents the number of risk quantization indexes, and i represents the index of the risk quantization index, represents the risk growth index, represents the link correlation coefficient, represents the link transmission efficiency, represents the reference link correlation coefficient, represents the time sensitivity coefficient, represents the risk exposure duration, represents the risk half-life.

[0076] Optionally, the risk classification label of the procurement risk factor is generated according to the risk transmission entropy value, comprising:

[0077] According to the risk transmission entropy value, a dynamic risk assessment threshold of the procurement risk factor is calculated;

[0078] The real-time fluctuation characteristics of the dynamic risk assessment threshold are analyzed;

[0079] Based on the real-time fluctuation characteristics, a classification risk interval corresponding to the dynamic risk assessment threshold is divided, wherein the classification risk interval includes a risk warning interval, a risk intervention interval and a risk disposal interval;

[0080] The risk diffusion acceleration of the classification risk interval is calculated;

[0081] In combination with the risk diffusion acceleration and the change slope of the risk transmission entropy value, a risk classification evaluation matrix of the procurement risk factor is constructed;

[0082] Based on the risk classification evaluation matrix, a risk classification label of the procurement risk factor is generated.

[0083] Compared with the problems described in the background art, the embodiments of the present application can accurately present the authenticity, consistency and correlation characteristics of enterprise procurement data by establishing a reliability verification matrix of the multi-dimensional procurement data, ensuring the accurate empowerment of the reliability verification matrix for multi-dimensional data integration and dynamic risk assessment of enterprise procurement anomaly detection; further, the embodiments of the present application can ensure the correlation of multi-dimensional data integration and the timeliness of dynamic risk assessment in enterprise procurement anomaly detection by extracting abnormal data points in the reliability verification matrix; the embodiments of the present application can reveal the linkage law of abnormal data points of different characteristic dimensions in complex procurement scenarios by constructing a procurement behavior interaction graph of the target enterprise based on the cross-correlation mode, improving the global recognition and tracing ability of multi-dimensional procurement anomalies; further, the embodiments of the present application can improve the effectiveness of multi-dimensional data integration and the timeliness of dynamic risk assessment in enterprise procurement anomaly detection by identifying procurement risk hotspots of the target enterprise according to the procurement behavior interaction graph, and enhance the forward warning capability of the procurement behavior interaction graph for enterprise procurement whole-process risk hidden dangers; the embodiments of the present application can dynamically quantify the conduction path of the cross-dimensional abnormal mode and the real-time deviation state of the normal boundary of the procurement process by detecting the cross-dimensional abnormal mode of the target enterprise according to the procurement behavior interaction graph, improving the accurate capture ability of cross-dimensional correlation risks in enterprise procurement anomaly detection; the embodiments of the present application can clarify the core direction of multi-dimensional data integration by extracting procurement risk factors of the target enterprise based on the procurement risk hotspots and the cross-dimensional abnormal mode, improving the efficiency and accuracy of dynamic risk assessment in enterprise procurement anomaly detection; further, the embodiments of the present application can improve the prediction ability of the enterprise for the development trend of procurement risks and the pertinence of the optimization of risk intervention measures by analyzing the risk evolution path of the procurement risk factors based on the procurement risk hotspots and the cross-dimensional abnormal mode; the embodiments of the present application can adapt to the complex characteristics of different risk evolution scenarios by calculating the risk conduction entropy value of the procurement risk factors according to the risk evolution path, forming a precise and dynamic procurement risk assessment mechanism, ensuring efficient identification of multi-dimensional risk correlation, and improving the decision support effect of the target enterprise procurement anomaly detection report; finally, the embodiments of the present application can generate risk classification labels of the procurement risk factors according to the risk conduction entropy value to output the procurement anomaly detection report of the target enterprise, which not only can significantly improve the accuracy of procurement anomaly detection and reduce the misjudgment and omission, but also can dynamically adjust the threshold standard according to the conduction characteristics and evolution trend of the risk factors, thereby effectively improving the adaptability and timeliness of the target enterprise procurement risk assessment, and can optimize the risk control process through dynamic threshold and classification labels, reduce the loss caused by procurement risks, improve the stability and efficiency of enterprise procurement business, and enhance the market competitiveness of the enterprise.Therefore, the multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection provided by the embodiment of the present application can improve the accuracy of enterprise procurement anomaly detection. BRIEF DESCRIPTION OF DRAWINGS

[0084] Figure 1 FIG. 1 shows a flowchart of a multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection provided by an embodiment of the present application.

[0085] Figure 2 FIG. 2 shows a procurement flowchart for implementing the multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection provided by an embodiment of the present application.

[0086] Figure 3 FIG. 3 shows a module diagram of a system for implementing the multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection provided by an embodiment of the present application.

[0087] The object implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0088] It should be understood that the specific embodiments described herein are merely intended to explain the present application and not to limit the present application.

[0089] The embodiment of the present application provides a multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection. The execution subject of the multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection includes but is not limited to at least one of electronic devices such as a server, a terminal, etc. which can be configured to execute the method provided by the embodiment of the present application. In other words, the multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection can be executed by software or hardware installed in a terminal device or a server device. The server includes but is not limited to a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0090] Referring to Figure 1 FIG. 1 shows a flowchart of a multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection provided by an embodiment of the present application. In this embodiment, the multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection includes:

[0091] S1, collecting multi-dimensional procurement data of a target enterprise, establishing a reliability verification matrix of the multi-dimensional procurement data, and extracting abnormal data points in the reliability verification matrix.

[0092] The embodiment of the present application can provide multi-modal data support for subsequent precise identification of procurement abnormal behavior and prevention of procurement risk by collecting multi-dimensional procurement data of target enterprises, wherein the target enterprise refers to an organization or unit that initiates procurement activities, and the multi-dimensional procurement data refers to various data sets reflecting procurement activities of the target enterprise from multiple different angles and levels, including internal procurement process data, supplier transaction data, market quotation data, etc., such as the name, qualification, historical cooperation record, and supply cycle of the supplier, as well as the application time, approval node, and approval personnel in the procurement process.

[0093] Further, the embodiment of the present application can accurately present the authenticity, consistency, and correlation characteristics of enterprise procurement data by establishing a reliability verification matrix of the multi-dimensional procurement data, thereby ensuring the accurate empowerment of the reliability verification matrix to the multi-dimensional data integration and dynamic risk assessment of enterprise procurement anomaly detection, wherein the reliability verification matrix refers to a structured analysis tool for verifying the reliability of multi-dimensional procurement data, which takes each specific index of multi-dimensional procurement data as the horizontal axis and the verification dimension of data reliability as the vertical axis, sets clear verification rules and judgment standards at each intersection point of the matrix, for example, at the intersection point of "procurement price" and "authenticity", the deviation threshold from the market average price can be set as the verification standard; at the intersection point of "procurement quantity" and "consistency", the matching degree verification rule can be associated with inventory data and production plan.

[0094] As an embodiment of the present application, the establishment of the reliability verification matrix of the multi-dimensional procurement data comprises:

[0095] determining a target procurement subject through the multi-dimensional procurement data;

[0096] extracting price data, supplier qualification data, and delivery time limit data from the multi-dimensional procurement data, and obtaining real-time execution data of the target procurement subject;

[0097] generating a price fluctuation index of the target procurement subject using the price data;

[0098] performing nonlinear fitting processing on the credit rating index in the supplier qualification data and the price fluctuation index to output a supplier basic credit score;

[0099] collecting market supply and demand data of the industry in which the target procurement subject is located, and combining the delivery time limit data and the market supply and demand data to calculate a delivery fluctuation coefficient corresponding to the target procurement subject;

[0100] identifying an execution deviation rate of the target procurement subject based on the real-time execution data;

[0101] According to the delivery fluctuation coefficient and the execution deviation rate, a dynamic execution risk coefficient of the target procurement subject is calculated;

[0102] In combination with the supplier basic credit score, the dynamic execution risk coefficient and the price fluctuation index, a reliability verification matrix of the multi-dimensional procurement data is established.

[0103] Wherein, the target procurement subject refers to the specific procurement behavior or procurement object pointed by multi-dimensional procurement data and needs to be verified for reliability, the price data refers to various data sets related to the target procurement subject and reflecting the price characteristics of the procurement subject, including but not limited to historical transaction price of procurement subject, real-time quotation, price fluctuation range, comparison value of quotations of different suppliers, and deviation rate of price and market benchmark price, etc., the supplier qualification data refers to various data used to represent the supplier's performance capability and compliance, including but not limited to supplier's credit rating, qualification certification documents (such as industry access qualification, quality system certification, etc.), historical cooperation violation records (such as the number of times of breach of contract, reasons for breach of contract), compliance inspection results of industry supervision department and annual score of third-party credit evaluation agency, etc., the delivery time data refers to data reflecting the time characteristics of the procurement subject from order confirmation to actual delivery, including but not limited to delivery cycle agreed in the contract, actual delivery time, number of delivery delays, delay time and delay reason tracing records, etc., which can be obtained by matching procurement contract and logistics receipt records, the real-time execution data refers to data reflecting the procurement progress and status generated in real time during the execution of the target procurement subject, including but not limited to production progress of procurement order, logistics transportation tracking data, warehousing acceptance progress, etc., which can be collected in real time through Internet of Things devices or supply chain management system, the price fluctuation index refers to a quantitative indicator reflecting the price stability of the procurement subject calculated based on the price data, with a value range of [0, 1], the credit rating indicator refers to a core quantitative indicator directly reflecting the credit level of the supplier extracted from the supplier qualification data, including but not limited to annual credit score (full score 100) of third-party credit evaluation agency, industry supervision compliance rate (compliance times / total inspection times), historical cooperation breach rate (breach times / total cooperation times), etc., the nonlinear fitting processing refers to the process of correlating the credit rating indicator and the price fluctuation index through a nonlinear mathematical model, such as BP neural network, for example, using a 3-layer BP neural network (input layer is credit score and compliance rate, hidden layer is 5 neurons, output layer is price fluctuation index correlation degree), fitting the credit rating indicator (92 points, 100%) and the price fluctuation index (0.32) of supplier A, and obtaining the nonlinear correlation degree of 0.85 (indicating a strong correlation between the two), the supplier basic credit score refers to a quantitative score obtained by nonlinear fitting processing, comprehensively reflecting the correlation degree between the supplier credit level and the price data reliability, and the value range is [0, 100], the market supply and demand data refers to the market supply and demand state data of the industry to which the target procurement subject belongs, including but not limited to market inventory of procurement target, industry demand growth rate, supply chain interruption risk level (such as low / medium / high), raw material price index, etc., which can be obtained through industry reports or supply chain monitoring platform, the delivery fluctuation coefficient refers to a quantitative index comprehensively reflecting the influence degree of market supply and demand on delivery timeliness data, which is calculated by the formula "delivery fluctuation coefficient = (actual delivery time - contract delivery cycle) / contract delivery cycle x market supply and demand tension index", the execution deviation rate refers to a quantitative index reflecting the deviation degree of real-time execution data from the planned progress, which is calculated by the formula "execution deviation rate = (|actual progress - planned progress| / planned progress) / n" (wherein n is the dimension number of real-time execution data, such as production progress, logistics progress, etc.), and the value range is [0, 1], the dynamic execution risk coefficient refers to a quantitative index comprehensively reflecting the influence of delivery fluctuation coefficient and execution deviation rate on the reliability of procurement data, which is calculated by the weighted summation formula "dynamic execution risk coefficient =. × delivery fluctuation coefficient + × execution deviation rate" (wherein 、 are weights, which are set according to the importance of the procurement target, ), and the value range is [0, 1].

[0104] The embodiment of the application can ensure the correlation of multi-dimensional data integration and the timeliness of dynamic risk assessment in enterprise procurement anomaly detection by extracting the abnormal data points in the reliability verification matrix. The abnormal data points refer to the data points that truly reflect the abnormal characteristics of multi-dimensional procurement data after eliminating pseudo abnormal points in the potential abnormal data area of the reliability verification matrix. For example, the "precision instrument quotation data" of a supplier is marked as a benchmark abnormal point in the potential abnormal data area. After eliminating the processing (verifying that there is no input error, and it is not caused by special market factors), it is confirmed that the quotation is significantly higher than the industry average level and there is an inexplicable contradiction with the supplier's capacity and historical cooperation price. Therefore, the data point is an abnormal data point.

[0105] As an embodiment of the application, the extraction of the abnormal data points in the reliability verification matrix comprises:

[0106] extracting adjacent index items and feature dimensions in the reliability verification matrix;

[0107] calculating the normalized data difference degree of the adjacent index items;

[0108] identify a degree of logical association between the characteristic dimensions, and calculate a degree of collaborative bias between the characteristic dimensions based on the degree of logical association;

[0109] calculate a density of abnormal points of the characteristic dimensions within the reliability verification matrix;

[0110] determine an abnormal accumulation index of the characteristic dimensions based on the normalized data difference degree, the degree of collaborative bias, and the density of abnormal points;

[0111] set an abnormal threshold of the abnormal accumulation index, and determine a potential abnormal data region of the multi-dimensional procurement data using the abnormal threshold;

[0112] identify a benchmark abnormal point of the multi-dimensional procurement data from the potential abnormal data region;

[0113] identify a pseudo abnormal point of the benchmark abnormal point, and perform a rejection processing of the pseudo abnormal point to obtain an abnormal data point of the multi-dimensional procurement data.

[0114] The adjacent index item refers to the adjacent core index entries in the reliability verification matrix that have logical correlation or data connection relationship, and is usually an index reflecting different aspects of the same procurement link or the associated characteristics of different links, such as price fluctuation index and supplier basic credit score, dynamic execution risk coefficient and delivery time deviation rate, etc. The characteristic dimension refers to the core data dimension constituting the reliability verification matrix, including but not limited to price dimension, supplier dimension, delivery dimension, quality dimension, etc. The normalized data difference degree refers to the quantitative index obtained by normalizing the numerical difference of adjacent index items, which is used to eliminate the influence of different index dimensions and objectively reflect the fluctuation deviation degree of adjacent data. The calculation formula is "normalized data difference degree = |index item A value - index item B value| / max (index item A value, index item B value)", and the value range is [0, 1]. For example, the normalized data difference degree of the adjacent index items "steel price fluctuation index (0.6)" and "supplier steel supply stability score (0.2)" is |0.6-0.2| / max(0.6,0.2)=0.4 / 0.6≈0.67, indicating that the data difference is significant. The logical correlation degree refers to the inherent correlation strength between characteristic dimensions based on the logic of procurement business, such as the strong correlation between price dimension and supplier dimension due to "cost-credit correlation", and the moderate correlation between delivery dimension and quality dimension due to "urgent order quality risk". The collaborative deviation degree refers to the degree of deviation of data in different characteristic dimensions from the normal collaborative relationship based on the logical correlation degree between characteristic dimensions. The abnormal point density refers to the number of abnormal data points contained in a single characteristic dimension in the reliability verification matrix per unit data volume. The calculation formula is "abnormal point density = number of abnormal data points in the characteristic dimension / total data points in the dimension", and the value range is [0, 1]. The abnormal accumulation index refers to the quantitative index value of the abnormal accumulation of each characteristic dimension, which is calculated based on the normalized data difference degree, the collaborative deviation degree and the abnormal point density. For example, the abnormal accumulation index of the "real-time price of raw material A" in the price dimension is 0.75, and the abnormal accumulation index of the "supplier B qualification data" in the supplier dimension is 0.62. The abnormal threshold refers to the critical value used to determine whether the abnormal accumulation index meets the abnormal standard, which is set based on the abnormal distribution characteristics of historical procurement data. For example, by analyzing the procurement abnormal data of an industry in the past three years, it is determined that the upper limit of the 95% confidence interval of the abnormal accumulation index is 0.65, and then 0.65 is set as the abnormal threshold. The procurement data with an abnormal accumulation index greater than or equal to 0.65 is marked as potential abnormal. The potential abnormal data region refers to the region covered by the data set with an abnormal accumulation index exceeding the abnormal threshold in the reliability verification matrix. The reference abnormal point refers to the core abnormal data point with the highest abnormal accumulation index and representative in the potential abnormal data region. For example, in the potential abnormal data region, the "imported equipment price of a certain supplier" has an abnormal accumulation index of 0.92) "significantly higher than other data points, and the quotation is inconsistent with the supplier's qualification and historical transaction price, so it is identified as a benchmark abnormal point, the pseudo abnormal point refers to the non-real abnormal data point in the benchmark abnormal point caused by objective and reasonable factors (such as data entry error, special market environment, temporary policy adjustment, etc.), for example, the quotation data of a supplier is mistakenly input as "10000 yuan per unit" by the input personnel, which leads to the abnormal accumulation index of the abnormal point being marked as a benchmark abnormal point, and the data point is a pseudo abnormal point after verification, the elimination process refers to the operation process of identifying and removing the pseudo abnormal point in the benchmark abnormal point through manual review, system verification or algorithm verification, etc., the purpose is to exclude interference factors and ensure that the finally extracted abnormal data point is real and effective.

[0115] Optionally, the logical correlation degree between the feature dimensions can be identified by CCA correlation analysis method, the abnormal accumulation index of the feature dimension can be determined by entropy weight method, and the potential abnormal data region of the multi-dimensional procurement data can be determined by Peak-over-Threshold method.

[0116] S2, identify the cross-correlation mode of the abnormal data point, and construct a procurement behavior interaction graph of the target enterprise based on the cross-correlation mode.

[0117] The embodiment of the application can break the data island, construct a global abnormal correlation network, reveal the risk transmission path of enterprise procurement anomaly, realize real-time early warning and root positioning by identifying the cross-correlation mode of the abnormal data point, and the cross-correlation mode refers to the internal correlation rule or combination characteristics between abnormal data points of different feature dimensions in multi-dimensional procurement data based on business logic, for example, in procurement anomaly detection, when the supplier qualification is fake, 80% of the probability will be accompanied by abnormal high price quotation, and 60% of the cases will appear short delivery, and this mode can be obtained by Apriori association rule algorithm.

[0118] Further, the embodiment of the application can reveal the linkage rule of abnormal data points of different feature dimensions in complex procurement scenarios, improve the global identification and tracing ability of multi-dimensional procurement anomaly by constructing the procurement behavior interaction graph of the target enterprise based on the cross-correlation mode, and the procurement behavior interaction graph refers to a topological graph constructed based on the cross-correlation mode, which visually presents the correlation relationship between multi-dimensional abnormal data points in the procurement process of the target enterprise, and is composed of nodes (abnormal data points), edges (correlation strength) and attribute labels (such as correlation type, risk level).

[0119] As an embodiment of the present application, the constructing the procurement behavior interaction graph of the target enterprise based on the cross correlation mode comprises:

[0120] locating the core interaction node corresponding to the cross correlation mode, and identifying the correlation dimension and its upstream and downstream dimensions of the core interaction node;

[0121] calculating the correlation strength value and the conduction time of the correlation dimension and the upstream and downstream dimensions;

[0122] determining the correlation influence weight of the correlation dimension according to the correlation strength value and the conduction time;

[0123] identifying the key conduction path corresponding to the core interaction node based on the correlation influence weight;

[0124] setting the visual correlation line of the cross correlation mode according to the key conduction path;

[0125] constructing the procurement behavior interaction graph of the target enterprise in combination with the core interaction node, the correlation influence weight and the visual correlation line.

[0126] The core interaction node refers to a key data node in an abnormal linkage core position in the cross-correlation mode, which has a triggering or leading role on other correlation dimensions, for example, in the cross-correlation mode of “supplier qualification abnormality→bid price high→delivery delay”, the “supplier qualification abnormality” node can be determined as the core interaction node because the correlation influence weight (0.82) is higher than that of other nodes. The correlation dimension refers to a procurement behavior characteristic classification dimension directly correlated with the core interaction node and participating in abnormal linkage, such as a supplier management dimension, a price management dimension, a contract execution dimension, etc. The upstream dimension refers to a dimension that has a sequential connection relationship with the correlation dimension based on the logic of the procurement business process, wherein the upstream dimension is a preceding link dimension of the correlation dimension, such as the “procurement demand dimension” being the upstream of the “supplier screening dimension”. The downstream dimension is a subsequent link dimension of the correlation dimension, such as the “contract execution dimension” being the downstream of the “supplier screening dimension”. The correlation strength value refers to an index quantifying the abnormal linkage closeness between the correlation dimension and the upstream and downstream dimensions, which is calculated by weighting the co-occurrence frequency (the number of times that two dimensions are abnormal at the same time / total sample number) and the correlation confidence (the probability that the downstream dimension is abnormal when the upstream dimension is abnormal), and the value range is [0, 1]. For example, the co-occurrence frequency of the “supplier qualification dimension” and the “delivery delay dimension” is 0.65, the correlation confidence is 0.92, and the correlation strength value is calculated by weighting (weighting each 0.5) to be 0.5*0.65+0.5*0.92=0.785, indicating that the linkage is close. The conduction time lag refers to a time interval parameter from the occurrence of the abnormality of the correlation dimension to the conduction to the upstream and downstream dimensions, which is usually in units of hours or days. The correlation influence weight refers to the influence degree of the correlation dimension on the upstream and downstream dimensions, which is calculated by the formula “correlation influence weight=correlation strength value*(1 / normalized value of conduction time lag)”, and the value range is [0, 1]. The key conduction path refers to a conduction route that plays a decisive role in the abnormal linkage of procurement behavior. The visual correlation line refers to a graphical element in the procurement behavior interaction diagram, which visually displays the key conduction path with the core interaction node as the center. The line thickness represents the correlation strength value, the color represents the conduction time lag, and the arrow direction represents the conduction direction from the core node outward.

[0127] Optionally, the core interaction node corresponding to the cross-correlation mode can be located by using a k-shell decomposition algorithm, and the key conduction path corresponding to the core interaction node can be identified by using Yen's algorithm.

[0128] To clearly show the interaction logic, correlation strength and conduction path of each link in the procurement business process of the target enterprise and assist in analyzing the complex relationship between procurement behaviors, refer to Figure 2As shown, a procurement process diagram of the multi-dimensional data integration and dynamic risk assessment method for realizing the enterprise procurement anomaly detection provided by an embodiment of the present application is provided, which clearly presents the step framework of the whole procurement process from "procurement decision making" to "contract signing", then to "procurement order establishment and procurement warehousing", and finally "procurement settlement", which can be used as the basic context of the "key conduction path", and in combination with the associated influence weight, the visual line can be more accurately set to restore the real correlation logic of the procurement behavior of the interactive diagram, and the construction efficiency and accuracy are improved.

[0129] S3, according to the procurement behavior interactive diagram, identifying the procurement risk hot spot of the target enterprise, and detecting the cross-dimension anomaly mode of the target enterprise.

[0130] According to the procurement behavior interactive diagram, the embodiment of the present application can improve the effectiveness of multi-dimensional data integration and the timeliness of dynamic risk assessment in enterprise procurement anomaly detection, and enhance the forward warning capability of the procurement behavior interactive diagram for the risk hidden danger of the whole procurement process of the enterprise. The procurement risk hot spot refers to the abnormal aggregation area with high risk correlation strength and rapid conduction characteristics identified based on the procurement behavior interactive diagram in the whole procurement process of the target enterprise.

[0131] As an embodiment of the present application, according to the procurement behavior interactive diagram, identifying the procurement risk hot spot of the target enterprise, comprising:

[0132] Analyzing the interactive features of the procurement behavior interactive diagram;

[0133] Identifying the node correlation weight and the historical anomaly record corresponding to the interactive features;

[0134] Extracting the key path distribution and business dependence strength in the node correlation weight;

[0135] Analyzing the trigger event type of the historical anomaly record, and identifying the risk conduction path corresponding to the trigger event type;

[0136] Based on the key path distribution and the risk conduction path, determining the risk hot spot density of the procurement behavior interactive diagram;

[0137] Identifying the potential influence range of the risk conduction path;

[0138] In combination with the risk hot spot density, the potential influence range and the business dependence strength, the procurement risk hot spot of the target enterprise is determined.

[0139] The interaction feature refers to the comprehensive attribute of the correlation between the core interaction nodes and the associated dimensions in the procurement behavior interaction graph, including interaction frequency, interaction direction, and interaction time sequence features, etc. For example, the interaction feature of the "supplier evaluation" node and the "order allocation" node is: interaction frequency is 15 times per month, interaction direction is one-way (from evaluation to allocation), and interaction time sequence feature is concentrated in the first week of each month, reflecting the stable periodic correlation between the two. The node correlation weight refers to the numerical parameter that quantifies the correlation strength between nodes in the interaction feature, which is calculated based on interaction frequency, historical abnormal co-occurrence probability, and business importance weighting, with a value range of [0, 1]. The historical abnormal record refers to the abnormal event archives that occurred in the procurement process of the target enterprise in the past, including contract breach, delivery delay, price fraud, etc. The key path distribution refers to the spatial distribution state of high-weight (such as ≥0.7) interaction paths selected from the node correlation weight in the procurement behavior interaction graph, including path number, covered business links, and concentrated areas, etc. The business dependency strength refers to the degree of mutual dependence between different business links due to process logic, which can be divided into three levels: strong (probability ≥70%), medium (30%-70%), and weak (≤30%). For example, the strong dependency relationship between the contract approval node and the supplier qualification audit node. The trigger event type refers to the initial event category in the historical abnormal record that triggers subsequent risk transmission, which is divided into: supplier violation category (such as providing false qualifications), process operation category (such as abuse of approval authority), external environment category (such as sudden rise in raw material prices), etc. The risk transmission path refers to the specific route of abnormality spread from the initial node to the associated dimensions triggered by the trigger event type, which is described by the time sequence and link sequence of node correlation, such as "supplier credit risk → payment delay → cash flow pressure". The risk hotspot density refers to the number of overlapping areas of key path distribution and risk transmission path in unit business link, which is calculated by "overlapping area number / total business link number". The higher the value, the higher the risk aggregation degree in the area. For example, a certain enterprise's procurement process includes 10 business links, among which "supplier evaluation", "contract signing", and "payment review" are the overlapping areas of key path and risk transmission path. Risk hotspot density = 3 / 10 = 0.3, indicating that the risk aggregation in this area is strong. The potential impact range refers to the procurement-related business fields and external associated links that may be affected by the risk transmission path, which is predicted based on the historical risk maximum spread range and the current business dependency strength, including direct impact links (such as within the procurement department) and indirect impact links (such as finance and production departments).

[0140] Optionally, the node association weight corresponding to the interaction feature can be identified by a PageRank algorithm, the risk transmission path corresponding to the trigger event type can be identified by a causal inference algorithm, the risk hotspot density of the procurement behavior interaction graph can be determined by a kernel density estimation (KDE) algorithm, and the potential influence range of the risk transmission path can be identified by a graph propagation model.

[0141] Further, by detecting the cross-dimension abnormal mode of the target enterprise according to the procurement behavior interaction graph, the real-time deviation state of the transmission path of the cross-dimension abnormal mode from the normal boundary of the procurement process can be dynamically quantified, the precision capturing ability of cross-dimension associated risks in enterprise procurement anomaly detection can be improved, and the cross-dimension abnormal mode refers to an abnormal combination mode with correlation and transmission that appears across two or more independent business dimensions in the procurement process of the target enterprise, for example, a false certificate in the supplier qualification link (which should be audited for production qualification) leads to a false high price in the price link (which should be reasonably priced), and further leads to a short delivery in the contract execution link (which should be delivered according to the agreement).

[0142] As an embodiment of the present application, the detecting the cross-dimension abnormal mode of the target enterprise according to the procurement behavior interaction graph comprises:

[0143] identifying cross-business domain abnormal transmission links and normal business links in the procurement behavior interaction graph;

[0144] extracting business activity sequences and business domain features corresponding to the cross-business domain abnormal transmission links;

[0145] calculating business coupling deviation and risk diffusion delay between the cross-business domain abnormal transmission links and the normal business links;

[0146] determining a risk transmission direction of the cross-business domain abnormal transmission links based on the business coupling deviation and the risk diffusion delay;

[0147] detecting the cross-dimension abnormal mode of the target enterprise in combination with the risk transmission direction, the business activity sequences and the business domain features.

[0148] The cross-business domain abnormal conduction link refers to an abnormal association path connecting two or more business domains in the procurement behavior interaction graph, and the interaction characteristics between nodes significantly deviate from the historical normal mode. The normal business link refers to a baseline conduction path reflecting the standard business process established based on historical compliance data, and the node association strength and time sequence characteristics conform to the established operation specification of the enterprise. The business activity sequence refers to the specific operation steps triggered in time sequence by the nodes corresponding to the cross-business domain abnormal conduction link in the procurement behavior interaction graph, which can be restored through the time sequence marking (such as timestamp) and line direction relationship of the nodes in the graph. For example, in the procurement behavior interaction graph, the node time sequence marking of the cross-business domain abnormal conduction link is “supplier qualification fraud (T1) → procurement order irregular issuance (T2) → financial excessive payment (T3) → warehouse receipt of inferior goods (T4)”, and the corresponding business activity sequence is: T1 audits qualification documents, T2 enters orders, T3 generates payment instructions, and T4 inspects goods. The business domain feature refers to the inherent attribute label of different business domain nodes in the procurement behavior interaction graph, including node function label, association rule description and normal threshold range. The business coupling deviation refers to the difference between the line association strength value of the cross-business domain abnormal conduction link and the standard association strength value of the corresponding line of the normal business link in the procurement behavior interaction graph, which can be directly calculated through the numerical value marked on the line in the graph. For example, in the procurement behavior interaction graph, the line marked association strength of the abnormal link “abnormal procurement order → abnormal warehouse inspection” is 0.9, the standard value marked on the corresponding line of the normal link is 0.4, and the business coupling deviation = 0.9-0.4 = 0.5. The risk diffusion delay degree refers to the deviation rate of the line conduction time value of the cross-business domain abnormal conduction link and the standard conduction time value of the corresponding line of the normal business link in the procurement behavior interaction graph, which is calculated through the time parameter marked on the line in the graph. For example, in the procurement behavior interaction graph, the line marked time of the abnormal link “abnormal procurement contract terms → abnormal financial payment” is 48 hours, and the standard time marked on the corresponding line of the normal link is 120 hours, and the risk diffusion delay degree = (48-120) / 120x100%=-60%. The risk conduction direction refers to the line direction of the cross-business domain abnormal conduction link in the procurement behavior interaction graph, which is determined by the arrow marking of the line. The arrow points from the upstream node to the downstream node, reflecting the path of risk diffusion in the graph.

[0149] Optionally, the business activity sequence corresponding to the cross-business domain abnormal conduction link can be extracted using a graph traversal algorithm. The cross-business domain abnormal conduction link and the normal business link in the procurement behavior interaction graph can be identified using an abnormal detection algorithm based on a graph neural network. The cross-dimensional abnormal mode of the target enterprise can be detected by a DBSCAN clustering algorithm.

[0150] S4, extracting a procurement risk factor of the target enterprise based on the procurement risk hot spot and the cross-dimension abnormal mode, and analyzing a risk evolution path of the procurement risk factor, and calculating a risk transmission entropy value of the procurement risk factor according to the risk evolution path.

[0151] The embodiment of the application can determine the core direction of multi-dimensional data integration by extracting the procurement risk factor of the target enterprise based on the procurement risk hot spot and the cross-dimension abnormal mode, and improve the efficiency and accuracy of dynamic risk assessment in enterprise procurement anomaly detection. The procurement risk factor refers to a core element or quantitative index that has a decisive influence on the risk status of procurement activities and is extracted from the procurement risk hot spot and the cross-dimension abnormal mode of the target enterprise.

[0152] As an embodiment of the application, the extraction of the procurement risk factor of the target enterprise based on the procurement risk hot spot and the cross-dimension abnormal mode comprises:

[0153] extracting the risk aggregation intensity of the procurement risk hot spot, and performing influence range definition processing of the cross-dimension abnormal mode to obtain a cross-domain influence radius;

[0154] defining a risk feature matrix of the procurement risk factor according to the risk aggregation intensity and the cross-domain influence radius;

[0155] generating a risk factor candidate pool of the target enterprise based on the risk feature matrix;

[0156] calculating the coupling coefficient between factors in the risk factor candidate pool and the business weight of the factors;

[0157] filtering out a core risk factor from the risk factor candidate pool through the coupling coefficient and the business weight;

[0158] constructing a risk assessment framework of the target enterprise based on the coupling coefficient and the core risk factor;

[0159] setting an optimization trigger threshold of the risk assessment framework according to the hot spot migration rate of the procurement risk hot spot and the mode variation frequency of the cross-dimension abnormal mode;

[0160] extracting the procurement risk factor of the target enterprise based on the risk assessment framework and the optimization trigger threshold.

[0161] The risk aggregation intensity refers to the concentration degree of risk factors in the procurement risk hotspot area, which is calculated by weighting the occurrence frequency and impact degree of high-risk events in unit business links, and the value range is [0, 1]. The cross-domain influence radius refers to the range size of the influence of the cross-dimension abnormal mode on different business domains, which is represented by the ratio of the number of affected business domains to the total number of business domains, and the value range is [0, 1]. For example, enterprise procurement involves supplier management, contract management, financial management, and inventory management, and a certain cross-dimension abnormal mode affects supplier management, contract management, and financial management. Therefore, the cross-domain influence radius is 3 / 4=0.75. The risk characteristic matrix refers to an m*n dimensional feature matrix that describes the multi-dimensional characteristic attributes of procurement risk factors, where m represents the risk characteristic dimension (such as risk occurrence probability, impact range, duration, etc.), n represents the number of risk factors extracted from the procurement risk hotspot and the cross-dimension abnormal mode, and the element value in the matrix represents the quantitative performance of the corresponding risk factor in the characteristic dimension. The risk factor candidate pool refers to a set containing all possible procurement risk factors generated based on the risk characteristic matrix, which is stored by business dimension. The risk factor candidate pool is stored by business dimension, including: supplier class: credit rating, on-time delivery rate, etc.; contract class: clause risk index, change frequency, etc.; financial class: payment anomaly rate, fund occupation ratio, etc. The coupling coefficient refers to the degree of mutual association and influence between different factors in the risk factor candidate pool. The business weight refers to the weight value assigned according to the importance of the risk factor in the procurement business process, which is set by the enterprise according to its business characteristics and risk preferences, and the value range is [0, 1]. The sum of the business weights of all factors is 1. The core risk factor refers to the risk factor selected from the risk factor candidate pool, which has a decisive influence on procurement risk and is representative. It is usually a factor with a high coupling coefficient and a large business weight. For example, after screening, the credit rating of the supplier, the clause risk index of the contract, and the payment anomaly rate of the finance are determined as core risk factors, and their overall influence on procurement risk exceeds 60%. The risk assessment framework refers to a structured system for assessing procurement risk, including data layer, analysis layer, and application layer. The data layer is responsible for real-time collection of core factor indicators, such as real-time collection of supplier credit rating change data and contract clause modification records. The analysis layer uses a coupling network model, where nodes represent factors and edge weights are corrected coupling coefficients, which are used to analyze the association between factors. The application layer contains a risk warning rule engine that triggers a warning when the core factor indicator reaches the warning threshold, for example, when the supplier credit rating decreases by more than 20% and the coupling network edge weight between the credit rating and the payment anomaly rate reaches 0.7 o'clock, the rule engine issues a warning, the hotspot migration rate refers to the rate of procurement risk hotspot transfer from one business link or area to other business links or areas, which is calculated by the ratio of the migration distance of the hotspot in a unit of time to the original hotspot influence range, for example, a procurement risk hotspot originally concentrates in the supplier screening link, and migrates to the contract signing link after 1 month, the migration distance corresponds to 2 business links, and the original hotspot influence range is 3 business links, so the hotspot migration rate is 2 / 3≈0.67, the mode variation frequency refers to the number of changes of the cross-dimension abnormal mode in a certain time, for example, in half a year, a cross-dimension abnormal mode changes from “supplier qualification fraud→price virtual high” to “supplier qualification fraud→delivery delay→price virtual high”, and then changes to “supplier qualification fraud→contract clause loophole→delivery delay”, so the mode variation frequency is 2 times, and the optimization trigger threshold refers to a step threshold for triggering optimization of the risk assessment framework, including a primary threshold and a high-level threshold, wherein the primary threshold triggers a warning notification to remind relevant personnel to pay attention to the running state of the risk assessment framework; when the high-level threshold is triggered, the structure and parameters of the risk assessment framework are adjusted.

[0162] Optionally, the coupling coefficient between the factors in the risk factor candidate pool can be calculated by using the Pearson correlation coefficient, the business weight of the factor can be determined by using the AHP hierarchical analysis method, and the risk assessment framework of the target enterprise can be constructed by using a graph neural network model.

[0163] Further, by analyzing the risk evolution path of the procurement risk factor based on the procurement risk hotspot and the cross-dimension abnormal mode, the embodiment of the present application can improve the prediction ability of the enterprise to the development trend of procurement risk and the pertinence of optimizing risk intervention measures, and the risk evolution path refers to a dynamic process and specific route in which the procurement risk factor gradually develops, diffuses from the initial state, and causes a series of chain risk reactions under the influence of the procurement risk hotspot combined with the conduction characteristics of the cross-dimension abnormal mode.

[0164] As an embodiment of the present application, the analysis of the risk evolution path of the procurement risk factor based on the procurement risk hotspot and the cross-dimension abnormal mode comprises:

[0165] Based on the procurement risk hotspot, identifying the risk conduction sensitive area of the cross-dimension abnormal mode;

[0166] In the risk conduction sensitive area, setting a dynamic tracking module of the procurement risk factor;

[0167] Capturing real-time state data of the procurement risk factor in different business links through the dynamic tracking module;

[0168] According to the dynamic tracking module and real-time state data, a risk diffusion monitor of the procurement risk factor is configured;

[0169] Based on the risk diffusion monitor, a risk evolution trend of the procurement risk factor is analyzed;

[0170] According to the risk evolution trend, a risk evolution path of the procurement risk factor is output.

[0171] The risk transmission sensitive area refers to a specific business area or link in the cross-dimension abnormal mode that has high sensitivity to the transmission of the procurement risk factor, is easily affected, and is easy to become a key node of risk diffusion. For example, in the cross-dimension abnormal mode of “supplier qualification abnormality → procurement price excessively high → contract execution default”, the “contract approval link” is a cross node connecting supplier management and financial payment, and once there is a mistake, it is easy to accelerate risk transmission, so it is marked as a risk transmission sensitive area. The dynamic tracking module refers to a software and hardware combined functional module deployed in the risk transmission sensitive area, which is used to capture and record the dynamic changes of the procurement risk factor in different business links in real time. This module can realize continuous monitoring and data collection of key indicators of risk factors by connecting data sources such as enterprise procurement management systems and supply chain management platforms. For example, for the procurement risk factor of “supplier credit rating”, the dynamic tracking module can connect the third-party credit rating platform and the enterprise internal cooperation record system in real time, update the credit score and default record of the supplier every 24 hours, and store the data in the database synchronously. The real-time state data refers to the quantitative information reflecting the specific state of the procurement risk factor at a specific time point or time period collected by the dynamic tracking module. For example, the real-time state data of the risk factor of “raw material procurement price” can include: the current procurement unit price (8500 yuan / ton), the price increase compared with yesterday (3%), the procurement link (contract signing stage), the corresponding supplier's offer validity period (remaining 3 days), etc. The risk diffusion monitor refers to a tool that analyzes and warns the diffusion speed, influence range and evolution direction of the procurement risk factor in different business links in real time based on the real-time state data collected by the dynamic tracking module. The risk evolution trend refers to the characteristic description of the development direction, change amplitude and possible business scope affected by the procurement risk factor in the future period of time obtained by analyzing the real-time state data and historical change law of the risk factor under the action of the dynamic tracking module and the risk diffusion monitor. For example, by analyzing the real-time state data (the qualified rates in the past 7 days are 98%, 95%, 90%, and 88% respectively) and historical trend of the risk factor of “supplier delivery qualified rate”, it is concluded that the risk evolution trend is “continuous decline”, and it may fall from “qualified” (≥90%) to “unqualified” (<85%) in the next month, and may affect the “production and assembly link”.

[0172] Optionally, the risk conduction sensitive area of the cross-dimension abnormal pattern can be identified by a heat map clustering algorithm, such as a K-means algorithm, and the risk diffusion monitor of the procurement risk factor can be constructed in combination with a directed acyclic graph (DAG) and a Bayesian network dynamic inference engine.

[0173] According to the risk evolution path, the embodiment of the present application can adapt to the complex characteristics of different risk evolution scenarios, form a precise and dynamic procurement risk assessment mechanism, ensure efficient identification of multi-dimensional risk correlation, and improve the decision support effect of the target enterprise procurement abnormal detection report. The risk conduction entropy value refers to an index for quantifying the disorder degree and interaction complexity of the risk factor state distribution at each link in the risk evolution path.

[0174] As an embodiment of the present application, the risk conduction entropy value of the procurement risk factor is calculated according to the risk evolution path, which includes:

[0175] Positioning the procurement risk factor at the path starting point of the risk evolution path and obtaining the risk trigger timestamp of the path starting point;

[0176] Based on the risk trigger timestamp, the risk exposure duration of the procurement risk factor is determined;

[0177] According to the path starting point, the conduction link of the procurement risk factor in the risk evolution path is identified;

[0178] Extracting the risk quantification index of the procurement risk factor and calculating the index deviation rate of the risk quantification index in the conduction link;

[0179] Calculating the link correlation coefficient between the conduction links;

[0180] In combination with the risk exposure duration, the link correlation coefficient and the index deviation rate, the risk conduction entropy value of the procurement risk factor is calculated.

[0181] The path starting point refers to a specific business link where the procurement risk factor first appears in the risk evolution path. For example, in the risk evolution path of "supplier material supply interruption → production plan stagnation → order delivery default", the "supplier material supply interruption" link is the path starting point. The risk trigger timestamp refers to the specific time record when the procurement risk factor first reaches the risk warning threshold at the path starting point. The risk exposure duration refers to the duration of the procurement risk factor in the risk state from the risk trigger timestamp to the current time (or the time when the risk is transmitted to the next link). It is calculated in hours. For example, the risk trigger timestamp is August 10, 2025, 9:30:15, and the current time is August 10, 2025, 15:30:15. The risk exposure duration is 6 hours. The transmission link refers to the connection sequence between the business links through which the procurement risk factor passes in the risk evolution path, reflecting the transmission path of the risk factor and the logical relationship between the links. For example, the sequence of "supplier material supply interruption → procurement department emergency coordination → production department plan adjustment → logistics department distribution change" is the transmission link of the risk factor. The risk quantification index refers to a measurable parameter for quantifying the severity of the procurement risk factor. Different types of risk factors correspond to different quantification indexes. For example, for the "supplier material supply interruption" risk factor, its risk quantification index can include the proportion of interrupted material (such as 30%) and the number of affected production batches (such as 5 batches). The index deviation rate refers to the deviation between the actual measurement value of the risk quantification index and the preset standard value. It is calculated by the formula "index deviation rate = (actual value - standard value) / standard value x 100%". The value can be positive or negative. A positive value indicates that the actual value is higher than the standard value, and a negative value indicates that the actual value is lower than the standard value. The link correlation coefficient refers to an index for quantifying the correlation between adjacent business links in the transmission link. It is calculated based on the business interaction frequency, data sharing degree, and process dependence strength between the links. The value range is [0, 1].

[0182] As another embodiment of the present application, the risk transmission entropy value of the risk evolution path is calculated by the following formula:

[0183]

[0184] wherein, represents the risk transmission entropy value of the procurement risk factor, represents the baseline entropy value of the procurement risk factor, represents the weight of the i-th risk quantification index, represents the index deviation rate of the i-th risk quantification index, represents the maximum index deviation rate, n represents the number of risk quantification indexes, and i represents the index of the risk quantification index. represents a risk growth index, represents a link correlation coefficient, represents a link conduction efficiency, represents a benchmark link correlation coefficient, represents a time sensitivity coefficient, represents a risk exposure time, represents a risk half-life.

[0185] It should be noted that the formula is used to quantify the relative risk intensity of each risk quantification index, can reflect the nonlinear amplification effect of risk, which can be determined by a nonlinear regression model, is used to simulate the energy dissipation of risk in conduction, can combine the time difference and information distortion rate (such as missing key data, error rate) of the transmission of procurement risk factors from the upstream link to the downstream link to calculate, for example, if the time difference accounts for 20% and the information distortion rate is 10%, = 0.6, = 0.4, then = 1-0.6x0.2-0.4x0.1=0.84, wherein, , is the weight, which is determined by historical data regression and satisfies , can reflect the cumulative effect of risk with exposure time, is used to quantify the effect of risk decay over time, wherein, represents the time required for the risk to naturally decay by half, which can be determined by calculating the average time required for different procurement risk factor types to naturally decay to 50% of the initial value.

[0186] S5, according to the risk conduction entropy value, generating the risk classification label of the procurement risk factor, to output the procurement anomaly detection report of the target enterprise.

[0187] The embodiment of the present application can not only significantly improve the accuracy of procurement anomaly detection, reduce misjudgment and missed judgment, but also dynamically adjust the threshold standard according to the transmission characteristics and evolution trend of the risk factor, thereby effectively improving the adaptability and timeliness of the procurement risk assessment of the target enterprise, and optimizing the risk control process through dynamic threshold and grading label, reducing the loss caused by procurement risk, improving the stability and efficiency of enterprise procurement business, enhancing the market competitiveness of the enterprise, the risk grading label refers to the identification label assigned after dividing the procurement risk level according to the size of the dynamic threshold of the procurement risk and the risk transmission entropy value, for example, the procurement risk can be divided into three levels of 'low risk','medium risk' and 'high risk', and the corresponding grading labels are 'R1', 'R2' and 'R3', respectively, when the risk transmission entropy value of a certain procurement risk factor is less than 0.5 and does not exceed the dynamic threshold, it is marked as 'R1'; when the entropy value is between 0.5-1.5 and close to the dynamic threshold, it is marked as 'R2'; when the entropy value exceeds 1.5 and exceeds the dynamic threshold, it is marked as 'R3', the procurement anomaly detection report refers to a comprehensive report about the abnormal situation in the procurement business of the target enterprise based on the analysis results of the dynamic threshold of the procurement risk, the risk grading label and the risk transmission entropy value, etc., including the specific information of the procurement risk factor, the risk evolution path, the risk level, the abnormal performance, the possible influence and the corresponding suggestions, etc., for example, a certain procurement anomaly detection report may point out: 'Supplier A's raw material delivery delay risk factor, risk transmission entropy value is 2.0, exceeds the dynamic threshold 1.5, corresponding grading label is 'R3', its abnormal performance is that the delivery delay time of the last 3 times exceeds 5 days, which may cause production plan delay, and it is suggested to replace the standby supplier urgently.

[0188] As an embodiment of the present application, the risk grading label of the procurement risk factor is generated according to the risk transmission entropy value, comprising:

[0189] According to the risk transmission entropy value, the dynamic risk assessment threshold of the procurement risk factor is calculated;

[0190] The real-time fluctuation characteristics of the dynamic risk assessment threshold are analyzed;

[0191] Based on the real-time fluctuation characteristics, the grading risk interval corresponding to the dynamic risk assessment threshold is divided, wherein the grading risk interval includes a risk warning interval, a risk intervention interval and a risk disposal interval;

[0192] The risk diffusion acceleration of the grading risk interval is calculated;

[0193] construct a risk grading evaluation matrix of the procurement risk factor based on the change slope of the risk diffusion acceleration and the risk conduction entropy value;

[0194] generate a risk grading label of the procurement risk factor based on the risk grading evaluation matrix.

[0195] The dynamic risk assessment threshold refers to a critical value for defining different risk levels of the procurement risk factor, which is calculated in real time and dynamically adjusted based on the risk conduction entropy value, and the specific calculation formula is: , wherein, represents the dynamic risk assessment threshold, represents the mean value of the risk conduction entropy value, represents the standard deviation of the risk conduction entropy value, represents the industry adjustment coefficient corresponding to the procurement risk factor, the real-time fluctuation characteristic refers to the numerical change rule of the dynamic risk assessment threshold within a unit time (such as every hour), including quantifiable characteristic parameters such as fluctuation amplitude, frequency, trend direction, etc., the grading risk interval refers to a numerical range with clear risk level boundaries divided based on the real-time fluctuation characteristic of the dynamic risk assessment threshold, which is used to intuitively distinguish the risk severity of the procurement risk factor, including a risk warning interval, a risk intervention interval, and a risk disposal interval, for example, the dynamic risk assessment threshold range of the “procurement contract breach risk” is 0-2.0, wherein the risk warning interval is [0, 0.6), the risk intervention interval is [0.6, 1.2), and the risk disposal interval is [1.2, 2.0], the risk diffusion acceleration refers to the change rate of the diffusion speed of the procurement risk factor in the grading risk interval with time, which is used to quantify the intensification or mitigation trend of risk diffusion, and the calculation formula is “risk diffusion acceleration = (current diffusion speed - previous time diffusion speed) / time interval”, the change slope refers to the curve slope of the risk conduction entropy value with time, which is used to reflect the change rate and direction of the risk conduction entropy value, a positive number indicates an increase in the entropy value (an increase in risk uncertainty), and a negative number indicates a decrease in the entropy value (a decrease in risk uncertainty), the risk grading evaluation matrix refers to a two-dimensional matrix constructed with the risk diffusion acceleration as the vertical axis and the change slope of the risk conduction entropy value as the horizontal axis, each cell in the matrix corresponds to a specific risk level, which is used to comprehensively evaluate the risk state of the procurement risk factor, for example, in a certain matrix, when the risk diffusion acceleration ≥ 2% / hour² and the change slope ≥ 0.3 / hour, the corresponding cell is marked as “high risk-quick deterioration”; when the risk diffusion acceleration < 1% / hour² and the change slope ≤ 0.1 / hour, the corresponding cell is marked as “low risk-slow mitigation”.

[0196] Optionally, the real-time fluctuation feature of the dynamic risk assessment threshold can be analyzed in combination with a Hurst index and a wavelet transform, and the risk classification label of the procurement risk factor can be generated using a decision tree algorithm.

[0197] Compared with the problems described in the background art, the embodiments of the present application can accurately present the authenticity, consistency and correlation characteristics of enterprise procurement data by establishing a reliability verification matrix of the multi-dimensional procurement data, ensuring the accurate empowerment of the reliability verification matrix for multi-dimensional data integration and dynamic risk assessment of enterprise procurement anomaly detection; further, the embodiments of the present application can ensure the correlation of multi-dimensional data integration and the timeliness of dynamic risk assessment in enterprise procurement anomaly detection by extracting abnormal data points in the reliability verification matrix; the embodiments of the present application can reveal the linkage law of abnormal data points of different characteristic dimensions in complex procurement scenarios by constructing a procurement behavior interaction graph of the target enterprise based on the cross-correlation mode, improving the global recognition and tracing ability of multi-dimensional procurement anomalies; further, the embodiments of the present application can improve the effectiveness of multi-dimensional data integration and the timeliness of dynamic risk assessment in enterprise procurement anomaly detection by identifying procurement risk hotspots of the target enterprise according to the procurement behavior interaction graph, and enhance the forward warning capability of the procurement behavior interaction graph for enterprise procurement whole-process risk hidden dangers; the embodiments of the present application can dynamically quantify the conduction path of the cross-dimensional abnormal mode and the real-time deviation state of the normal boundary of the procurement process by detecting the cross-dimensional abnormal mode of the target enterprise according to the procurement behavior interaction graph, improving the accurate capture ability of cross-dimensional correlation risks in enterprise procurement anomaly detection; the embodiments of the present application can clarify the core direction of multi-dimensional data integration by extracting procurement risk factors of the target enterprise based on the procurement risk hotspots and the cross-dimensional abnormal mode, improving the efficiency and accuracy of dynamic risk assessment in enterprise procurement anomaly detection; further, the embodiments of the present application can improve the prediction ability of the enterprise for the development trend of procurement risks and the pertinence of the optimization of risk intervention measures by analyzing the risk evolution path of the procurement risk factors based on the procurement risk hotspots and the cross-dimensional abnormal mode; the embodiments of the present application can adapt to the complex characteristics of different risk evolution scenarios by calculating the risk conduction entropy value of the procurement risk factors according to the risk evolution path, forming a precise and dynamic procurement risk assessment mechanism, ensuring efficient identification of multi-dimensional risk correlation, and improving the decision support effect of the target enterprise procurement anomaly detection report; finally, the embodiments of the present application can generate risk classification labels of the procurement risk factors according to the risk conduction entropy value to output the procurement anomaly detection report of the target enterprise, which not only can significantly improve the accuracy of procurement anomaly detection and reduce the misjudgment and omission, but also can dynamically adjust the threshold standard according to the conduction characteristics and evolution trend of the risk factors, thereby effectively improving the adaptability and timeliness of the target enterprise procurement risk assessment, and can optimize the risk control process through dynamic threshold and classification labels, reduce the loss caused by procurement risks, improve the stability and efficiency of enterprise procurement business, and enhance the market competitiveness of the enterprise.Therefore, the multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection provided by the embodiment of the application can improve the accuracy of enterprise procurement anomaly detection.

[0198] As shown in Figure 3 is a functional module diagram of a multi-dimensional data integration and dynamic risk assessment system for enterprise procurement anomaly detection.

[0199] The multi-dimensional data integration and dynamic risk assessment system 200 for enterprise procurement anomaly detection can be installed in an electronic device. According to the functions implemented, the multi-dimensional data integration and dynamic risk assessment system for enterprise procurement anomaly detection can include a data verification module 201, a procurement behavior analysis module 202, an anomaly identification module 203, a risk assessment module 204, and a result output module 205. The modules of the application can also be referred to as units, which refer to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, and are stored in the memory of the electronic device.

[0200] In the embodiment of the application, the functions of each module / unit are as follows:

[0201] The data verification module 201 is configured to collect multi-dimensional procurement data of a target enterprise, establish a reliability verification matrix of the multi-dimensional procurement data, and extract abnormal data points in the reliability verification matrix, wherein the multi-dimensional procurement data includes internal procurement process data, supplier transaction data, and market trend data.

[0202] The procurement behavior analysis module 202 is configured to identify cross-correlation patterns of the abnormal data points, and construct a procurement behavior interaction graph of the target enterprise based on the cross-correlation patterns.

[0203] The anomaly identification module 203 is configured to identify procurement risk hotspots of the target enterprise according to the procurement behavior interaction graph, and detect cross-dimensional anomaly patterns of the target enterprise.

[0204] The risk assessment module 204 is configured to extract procurement risk factors of the target enterprise based on the procurement risk hotspots and the cross-dimensional anomaly patterns, analyze risk evolution paths of the procurement risk factors, calculate risk transmission entropy values of the procurement risk factors according to the risk evolution paths, and generate risk classification labels of the procurement risk factors according to the risk transmission entropy values.

[0205] The result output module 205 is configured to generate risk classification labels of the procurement risk factors according to the risk transmission entropy values, and output a procurement anomaly detection report of the target enterprise.

[0206] In detail, the modules in the multi-dimensional data integration and dynamic risk assessment system 200 for enterprise procurement anomaly detection in the embodiments of the present application adopt the same technical means as the multi-dimensional data integration and dynamic risk assessment method for enterprise procurement anomaly detection in the above-mentioned Figure 1

[0207] It is obvious for those skilled in the art that the present application is not limited to the details of the above-mentioned exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0208] Finally, it should be noted that in the above-mentioned embodiments, each embodiment can be combined with or independent of each other, and deleting any one of them does not affect the technical implementation of the other embodiments. The above embodiments are only used to illustrate the technical solutions of the present application but not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application.​

Claims

1. A method for multi-dimensional data integration and dynamic risk assessment of enterprise procurement anomaly detection, characterized in that, The method comprises: Collecting multi-dimensional procurement data of a target enterprise, establishing a reliability verification matrix of the multi-dimensional procurement data, and extracting abnormal data points in the reliability verification matrix, wherein the multi-dimensional procurement data comprises internal procurement process data, supplier transaction data and market quotation data; Identifying a cross-correlation pattern of the abnormal data points, and constructing a procurement behavior interaction graph of the target enterprise based on the cross-correlation pattern; According to the procurement behavior interaction graph, identifying procurement risk hotspots of the target enterprise, and detecting cross-dimension abnormal patterns of the target enterprise; Based on the procurement risk hotspots and the cross-dimension abnormal patterns, extracting procurement risk factors of the target enterprise, analyzing risk evolution paths of the procurement risk factors, and calculating risk transmission entropy values of the procurement risk factors according to the risk evolution paths, wherein the calculation of the risk transmission entropy values of the procurement risk factors according to the risk evolution paths comprises: Locating the procurement risk factor at the starting point of the risk evolution path, and obtaining a risk trigger timestamp of the starting point; Based on the risk trigger timestamp, determining the risk exposure duration of the procurement risk factor; According to the starting point, identifying the transmission link of the procurement risk factor in the risk evolution path; Extracting risk quantification indicators of the procurement risk factor, and calculating index deviation rates of the risk quantification indicators in the transmission link; Calculating the link correlation coefficient between the transmission links; Combining the risk exposure duration, the link correlation coefficient and the index deviation rate, the risk transmission entropy value of the risk evolution path is calculated by the following formula: wherein, represents a risk transmission entropy value of a procurement risk factor, represents a benchmark entropy value of a procurement risk factor, represents a weight of the i-th risk quantification indicator, represents an indicator deviation rate of the i-th risk quantification indicator, represents a maximum indicator deviation rate, n represents the number of risk quantification indicators, and i represents an index of the risk quantification indicator, represents a risk growth index, represents a link correlation coefficient, represents a link transmission efficiency, represents a benchmark link correlation coefficient, represents a time sensitivity coefficient, represents a risk exposure duration, represents a risk half-life. According to the risk transmission entropy value, generating a risk classification label of the procurement risk factor, and outputting a procurement anomaly detection report of the target enterprise.

2. The method of claim 1, wherein the method further comprises: identifying a plurality of dimensions of data associated with the enterprise procurement process; and determining a plurality of risk factors associated with the enterprise procurement process based on the plurality of dimensions of data. The establishment of the reliability verification matrix of the multi-dimensional procurement data comprises: Determining a target procurement subject through the multi-dimensional procurement data; Extracting price data, supplier qualification data and delivery time limit data from the multi-dimensional procurement data, and obtaining real-time execution data of the target procurement subject; Generating a price fluctuation index of the target procurement subject using the price data; Performing nonlinear fitting processing on the credit rating indicators in the supplier qualification data and the price fluctuation index to output a supplier basic credit score; Collecting market supply and demand data of the industry in which the target procurement subject is located, and combining the delivery time limit data and the market supply and demand data to calculate a delivery fluctuation coefficient corresponding to the target procurement subject; Identifying an execution deviation rate of the target procurement subject based on the real-time execution data; According to the delivery fluctuation coefficient and the execution deviation rate, calculating a dynamic execution risk coefficient of the target procurement subject; Combining the supplier basic credit score, the dynamic execution risk coefficient and the price fluctuation index, the reliability verification matrix of the multi-dimensional procurement data is established.

3. The multi-dimensional data integration and dynamic risk assessment method for detecting anomalies in enterprise procurement as described in claim 1, characterized in that, The extraction of the abnormal data points in the reliability verification matrix comprises: Extracting adjacent index items and feature dimensions in the reliability verification matrix; calculating a normalized data difference degree of the adjacent index items; identifying a logical correlation degree between the feature dimensions and calculating a collaborative bias degree between the feature dimensions based on the logical correlation degree; calculating an abnormal point density of the feature dimensions in the reliability verification matrix; determining an abnormal accumulation index of the feature dimensions based on the normalized data difference degree, the collaborative bias degree and the abnormal point density; setting an abnormal threshold of the abnormal accumulation index and determining a potential abnormal data region of the multi-dimensional procurement data using the abnormal threshold; identifying a benchmark abnormal point of the multi-dimensional procurement data from the potential abnormal data region; identifying pseudo abnormal points of the benchmark abnormal point and performing a rejection processing of the pseudo abnormal points to obtain an abnormal data point of the multi-dimensional procurement data.

4. The multi-dimensional data integration and dynamic risk assessment method for detecting anomalies in enterprise procurement as described in claim 1, characterized in that, The constructing of the procurement behavior interaction graph of the target enterprise based on the cross-correlation pattern comprises: locating a core interaction node corresponding to the cross-correlation pattern and identifying an associated dimension and its upstream and downstream dimensions of the core interaction node; calculating an associated strength value and a conduction time of the associated dimension and the upstream and downstream dimensions; determining an associated influence weight of the associated dimension according to the associated strength value and the conduction time; identifying a key conduction path corresponding to the core interaction node based on the associated influence weight; setting a visual association line of the cross-correlation pattern according to the key conduction path; constructing the procurement behavior interaction graph of the target enterprise in combination with the core interaction node, the associated influence weight and the visual association line.

5. The method for multi-dimensional data integration and dynamic risk assessment of enterprise procurement anomaly detection of claim 1, wherein, The identifying of the procurement risk hotspots of the target enterprise according to the procurement behavior interaction graph comprises: analyzing an interaction feature of the procurement behavior interaction graph; identifying a node association weight and a historical abnormal record corresponding to the interaction feature; extracting a key path distribution and a business dependence strength in the node association weight; analyzing a trigger event type of the historical abnormal record and identifying a risk conduction path corresponding to the trigger event type; determining a risk hotspot density of the procurement behavior interaction graph based on the key path distribution and the risk conduction path; identifying a potential impact range of the risk conduction path; determining the procurement risk hotspots of the target enterprise in combination with the risk hotspot density, the potential impact range and the business dependence strength.

6. The method for multi-dimensional data integration and dynamic risk assessment of enterprise procurement anomaly detection of claim 1, wherein, The detecting of the cross-dimensional abnormal pattern of the target enterprise according to the procurement behavior interaction graph comprises: identifying a cross-business domain abnormal conduction link and a normal business link in the procurement behavior interaction graph; extracting a business activity sequence and a business domain feature corresponding to the cross-business domain abnormal conduction link; calculating a business coupling bias and a risk diffusion delay degree between the cross-business domain abnormal conduction link and the normal business link; determining a risk conduction direction of the cross-business domain abnormal conduction link based on the business coupling bias and the risk diffusion delay degree; detecting the cross-dimensional abnormal pattern of the target enterprise in combination with the risk conduction direction, the business activity sequence and the business domain feature.

7. The method for multi-dimensional data integration and dynamic risk assessment of enterprise procurement anomaly detection of claim 1, wherein, The extraction of the procurement risk factor of the target enterprise based on the procurement risk hot spot and the cross-dimension abnormal mode comprises: extracting the risk aggregation intensity of the procurement risk hot spot and performing influence range definition processing of the cross-dimension abnormal mode to obtain a cross-domain influence radius; defining a risk feature matrix of the procurement risk factor according to the risk aggregation intensity and the cross-domain influence radius; generating a risk factor candidate pool of the target enterprise based on the risk feature matrix; calculating a coupling coefficient between factors in the risk factor candidate pool and a business weight of the factor; screening a core risk factor from the risk factor candidate pool through the coupling coefficient and the business weight; constructing a risk assessment framework of the target enterprise based on the coupling coefficient and the core risk factor; setting an optimization trigger threshold of the risk assessment framework according to a hot spot migration rate of the procurement risk hot spot and a mode variation frequency of the cross-dimension abnormal mode; extracting the procurement risk factor of the target enterprise based on the risk assessment framework and the optimization trigger threshold.

8. The method for multi-dimensional data integration and dynamic risk assessment of enterprise procurement anomaly detection of claim 1, wherein, The analysis of the risk evolution path of the procurement risk factor based on the procurement risk hot spot and the cross-dimension abnormal mode comprises: identifying a risk transmission sensitive area of the cross-dimension abnormal mode based on the procurement risk hot spot; setting a dynamic tracking module of the procurement risk factor in the risk transmission sensitive area; capturing real-time state data of the procurement risk factor at different business links through the dynamic tracking module; configuring a risk diffusion monitor of the procurement risk factor according to the dynamic tracking module and the real-time state data; analyzing a risk evolution trend of the procurement risk factor based on the risk diffusion monitor; outputting the risk evolution path of the procurement risk factor according to the risk evolution trend.

9. The method for multi-dimensional data integration and dynamic risk assessment of enterprise procurement anomaly detection of claim 1, wherein, The generation of the risk classification label of the procurement risk factor according to the risk transmission entropy value comprises: calculating a dynamic risk assessment threshold of the procurement risk factor according to the risk transmission entropy value; analyzing real-time fluctuation characteristics of the dynamic risk assessment threshold; dividing a classification risk interval corresponding to the dynamic risk assessment threshold based on the real-time fluctuation characteristics, wherein the classification risk interval comprises a risk early warning interval, a risk intervention interval and a risk disposal interval; calculating a risk diffusion acceleration of the classification risk interval; constructing a risk classification assessment matrix of the procurement risk factor in combination with the risk diffusion acceleration and a change slope of the risk transmission entropy value; generating the risk classification label of the procurement risk factor based on the risk classification assessment matrix.

Images