A method and system for identifying and dynamically controlling the liability risks of bidding entities by integrating multi-source data.
By integrating multi-source data and constructing a heterogeneous information network, combined with dynamic adaptive thresholds, the problems of low efficiency in identifying risks to bidding entities and delayed risk warnings have been solved, thus realizing intelligent and refined supervision of the bidding market.
Patent Information
- Application Number
- CN202610967177.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-01
- Publication Date
- 2026-11-10
AI Technical Summary
In existing technologies, the identification of risks to the responsibility of bidding entities is inefficient, manual review is easily affected by subjective factors, the identification coverage of a single data source is limited and cannot discover hidden relationships, and risk identification is mostly based on static scoring, which leads to delayed risk warnings and makes it impossible to effectively intercept violations before the bid opening.
By integrating multi-source data (business and commercial, judicial, transaction, and behavioral logs) to resolve data conflicts, constructing a heterogeneous information network, extracting device fingerprints and equity control characteristics, quantifying risk scores, and employing dynamic adaptive thresholds and hierarchical prevention and control, dynamic risk identification and interception can be achieved.
It enables accurate identification and dynamic control of risks of bidding entities, reduces the probability of missed or incorrect judgments, can automatically intercept high-risk entities before the bid opening, and generates structured reports of traceability information, thereby improving regulatory efficiency and accuracy.
Smart Images

Figure CN123070673A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of big data, graph computing, and intelligent supervision of bidding and tendering, specifically to a method and system for identifying and dynamically controlling the liability risks of bidding and tendering entities by integrating multi-source data. Background Technology
[0002] As an important means of resource allocation, the fairness and compliance of bidding activities are directly related to market order and public interest. However, in bidding practice, behaviors that pose risks to the bidding entities, such as bid rigging, collusion, qualification fraud, and the transfer of benefits, are rampant and seriously disrupt the normal order of the bidding market.
[0003] Currently, the identification and prevention of risks related to the liability of bidding entities mainly adopt the following two methods: one is to rely on manual review, that is, to have bidding experts or supervisors check the bid documents and enterprise qualifications one by one; the other is to use a rule-based early warning system based on a single data source, such as identifying suspected bid rigging by detecting whether the bidder's IP address and MAC address are the same.
[0004] However, existing technologies have the following drawbacks: manual review is inefficient, difficult to handle large-scale bidding projects, and easily affected by subjective factors; the rule warning coverage of a single data source is limited and cannot detect hidden relationships; existing systems lack the ability to integrate and process multi-source heterogeneous data, with data sources such as business, judicial, transaction, and behavior logs being isolated from each other, and data conflicts cannot be automatically resolved; risk identification is mostly static scoring or post-event statistics, which cannot achieve dynamic tracking and real-time prevention and control, resulting in delayed risk warnings and the inability to effectively intercept violations before the bid opening. Therefore, we propose a method and system for risk identification and dynamic prevention and control of bidding entity responsibilities that integrates multi-source data to solve the above-mentioned problems. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides a method and system for identifying and dynamically controlling the liability risks of bidding entities by integrating multi-source data. This solves the problems of existing technologies, such as data fragmentation, difficulty in handling information conflicts, inability to identify hidden violations, rigid risk assessment standards, lagging supervision, low efficiency of manual review, and inconvenience in obtaining evidence and tracing the source.
[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for identifying and dynamically controlling the liability risks of bidding entities by integrating multi-source data. The method collects heterogeneous data from multiple sources, including business registration data, judicial data, transaction data, and behavioral logs, and integrates this data. Combining a heterogeneous information network model with risk quantification calculations, it achieves risk identification of bidding entities and interception of violating entities. Specifically, the method includes the following steps:
[0007] S1. Perform data conflict detection and fusion resolution on the same attribute items of the same bidding entity, remove contradictory data, and output a reliable fused dataset;
[0008] S2. Based on the trusted fusion dataset, construct a heterogeneous information network containing three types of nodes: tenderer, bidder, and agent. Extract device fingerprint association features and equity control features and form a composite risk evidence chain to quantify the initial risk atom value corresponding to the bidding node.
[0009] S3. Calculate the correlation strength between any two bidding nodes according to the preset meta-path, and solve the final risk score of the bidding node by combining the correlation strength with the initial risk atom value.
[0010] S4. Compare the final risk score with a preset risk threshold. When the final risk score exceeds the threshold, generate an interception command to block the corresponding bidding entity from participating in this bidding activity before the bid opening.
[0011] This invention provides a system for identifying and dynamically preventing risks to bidding entities' responsibilities by integrating multi-source data. The system is used to execute the method described above and includes a data conflict resolution engine, a heterogeneous graph association calculation unit, and a dynamic scoring and execution module.
[0012] The data conflict resolution engine is configured to receive heterogeneous data from multiple sources, such as business, judicial, transaction, and behavior logs, and to complete data conflict detection, fusion and resolution processing, and output a trusted fusion dataset.
[0013] The heterogeneous graph association calculation unit is configured to store a heterogeneous information network consisting of three types of main nodes: the tenderer, the bidder, and the agency, as well as the association relationships between the nodes. It uses a frequent itemset mining algorithm to statistically analyze the co-occurrence pattern of device fingerprints and completes the calculation of equity control features and node association strength.
[0014] The dynamic scoring and execution module is configured to perform risk quantification calculation, adaptive threshold determination, and violation interception logic, and it is connected to the standard interface of the electronic bidding platform to intercept high-risk bidding entities before the bid opening.
[0015] Beneficial effects
[0016] This invention provides a method and system for identifying and dynamically controlling the liability risks of bidding entities by integrating multi-source data. Compared with existing technologies, it has the following advantages:
[0017] This method and system for identifying and dynamically controlling the liability risks of bidding entities by integrating multi-source data integrates four types of heterogeneous data: industrial and commercial, judicial, transaction, and behavioral logs. It is equipped with comprehensive data conflict resolution rules, effectively breaking down the data silos that are prevalent in the industry. It automatically identifies and handles contradictions in multi-source information, ensuring the authenticity and reliability of the analyzed data and laying a solid data foundation for risk identification. Simultaneously, this invention constructs a heterogeneous information network to deeply mine implicit interest relationships such as equity connections, equipment sharing, and business dealings. Combined with risk propagation algorithms and quantitative models, it can accurately identify hidden violations such as bid rigging and collusion, overcoming the shortcomings of traditional single-data source identification dimensions and insufficient depth of mining.
[0018] The system adopts a control model that combines dynamic adaptive thresholds with tiered prevention and control, overcoming the shortcomings of traditional fixed judgment standards and static risk scoring. It can dynamically adjust judgment rules based on the overall market risk situation, effectively reducing the probability of missed or incorrect risk assessments. The system can automatically intercept high-risk entities before bidding, achieving pre-emptive risk prevention and completely improving the problem of traditional supervision being primarily reactive and lagging in its control.
[0019] It can automatically generate structured risk reports with complete traceability information, meeting the compliance requirements for regulatory record keeping and judicial evidence collection. Combined with network graph visualization capabilities, it can intuitively present the risk transmission path, significantly reducing the workload of manual review while balancing regulatory efficiency and control precision. This comprehensively improves the intelligence and refinement of bidding supervision, effectively maintaining a fair and orderly operating environment in the bidding market. Attached Figure Description
[0020] Figure 1 This is a flowchart of a method for identifying and dynamically controlling the liability risks of bidding entities that integrates multi-source data, according to the present invention.
[0021] Figure 2 This is a connection block diagram of a bidding entity liability risk identification and dynamic prevention and control system that integrates multi-source data according to the present invention. Detailed Implementation
[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0023] like Figure 1-2 As shown:
[0024] A method for identifying and dynamically controlling the liability risks of bidding entities by integrating multi-source data is proposed. This method collects and integrates heterogeneous data from multiple sources, including business registration data, judicial data, transaction data, and behavioral logs. Combining a heterogeneous information network model with risk quantification calculation, it achieves risk identification of bidding entities and interception of violators. Specifically, it includes six core steps: data fusion and resolution, feature extraction and initial risk assignment, node association strength and final risk score calculation, threshold comparison and interception execution, dynamic risk propagation, and hierarchical adaptive prevention and control. The detailed implementation process is as follows:
[0025] S1. The system uniformly collects multi-source heterogeneous data from the entire bidding and tendering process. The data sources specifically include: business registration data output by the market supervision and management department (company name, unified social credit code, business address, shareholder information, legal representative, company status, etc.), judicial data output by the judicial authorities (company litigation records, administrative penalties, dishonest persons subject to enforcement, violation penalty timestamps, etc.), transaction data generated by the electronic bidding and tendering platform (bid registration information, bid price, bid section information, number of bidders, price details, etc.), and behavioral log data collected by the platform front-end (user login records, file upload records, device operating environment, device identification, login geographical location, etc.).
[0026] Data conflict detection is performed on identical attribute items for the same bidding entity, and the data is fused and resolved according to preset rules to remove contradictory and invalid data. Finally, a standardized, highly reliable fused dataset is output. The specific conflict detection and fusion resolution rules are as follows:
[0027] The judicial data priority rule stipulates that if the update timestamp of judicial data for the same attribute item of the same entity is later than that of business registration data, the judicial data content will be given priority, and the attribute will be marked as a legal priority attribute. Because judicial data records legally mandated violations such as penalties, litigation, and breaches of trust by enterprises, its timeliness and legal validity are higher than those of regular business registration information. Therefore, this priority rule is established to ensure the legality and accuracy of risk data.
[0028] The geographic information logical verification rules extract the real-time geographic location corresponding to the device fingerprint from the behavior log, convert it to the geographic code of the business address registered in the business registration data, and calculate the straight-line distance between the two addresses. If the distance exceeds the preset distance threshold, it is determined that there is a logical contradiction between the two types of data. The system automatically triggers a real-time information verification request, and the entity’s real business and office location is manually reviewed or verified a second time to eliminate the risk of false addresses and illegal operations in other locations.
[0029] S2. Based on the trusted fusion dataset generated in S1, a heterogeneous information network is constructed. The network nodes are divided into three categories: tenderer nodes, bidder nodes, and agency nodes. Network edges are established between nodes based on business associations, personnel associations, equipment associations, and equity associations. At the same time, equity control features and equipment fingerprint association features are extracted from the dataset. The two types of features are combined to form a composite risk evidence chain. Combined with typical violation scenarios such as bidding collusion, bid rigging, and bid rigging, the initial risk atomic value is quantitatively calculated for each bidding node.
[0030] Equity control features are extracted by calculating the shortest equity connection path between different bidding companies based on shareholder and legal representative information from business registration data. If two companies have direct shareholding, are controlled by the same natural person, or are parent and subsidiary companies, the shorter the shortest equity connection path, the higher the degree of interest linkage between the companies and the greater the risk of bid rigging. This is used to generate equity control features.
[0031] Device fingerprint association feature extraction involves collecting front-end operating environment information from key stages such as bidders logging into the bidding platform and uploading bid documents. A hash algorithm is used to encrypt and calculate device hardware parameters and operating environment parameters, generating a globally unique device identifier (i.e., device fingerprint). If multiple different bidder accounts / enterprises share the same device identifier within the same bidding collaboration cycle, it is determined that shared device operation behavior exists, and device fingerprint association features are generated.
[0032] The initial risk atomic value quantification assignment rules combine historical bidding violation case statistics to set quantification thresholds, and complete risk assignment for typical violation scenarios. Specific rules are as follows:
[0033] If the number of times the device fingerprint association feature co-occurs is ≥3, it means that multiple companies have been using the same device for bidding for a long time, which poses an extremely high risk of collusion. The corresponding risk value is set at 0.8.
[0034] In the equity control feature, the shortest connection path length of equity is 1, which means that the enterprise is a directly related entity with obvious interest ties, and the corresponding risk value is set to 0.9;
[0035] Anomaly chains in bid prices are identified. When the number of valid bidders within the same bid section is less than 5 and the bid price dispersion coefficient is greater than 0.2, it indicates abnormal bidding and suspicion of malicious bid control. The corresponding risk value is set at 0.6.
[0036] The above three types of risk values serve as the initial risk atomic values for bidding nodes. If the same node triggers multiple risk rules, the highest risk value is taken as the final initial risk atomic value for that node.
[0037] S3. Based on the preset meta-path of the heterogeneous information network, calculate the correlation strength between any two bidding nodes, and then combine the correlation strength with the initial risk atom value obtained in S2 to solve the final risk score of each bidding node.
[0038] Formula for calculating correlation strength:
[0039]
[0040] in For edge weights, the meta-path must include at least "bidder-tenderer-bidder" and "bidder-agent-bidder";
[0041] Final risk score calculation formula:
[0042]
[0043] in, For bidding nodes The initial risk atom value, For bidding nodes The final risk score, The propagation coefficient is preset and 0 ≤ ≤1, To be in sync with the bidding nodes A set of other related bidding nodes;
[0044] This formula achieves a comprehensive calculation of both the risk of a single point of origin and the risk transmitted from related entities, fully taking into account the risk linkage effect between related enterprises;
[0045] S4. Compare the final risk score of each bidding node with the system's preset risk threshold. When the final risk score exceeds the threshold, the system immediately generates an interception command to block the corresponding bidding entity from participating in this bidding activity before the bidding process officially begins, thus preventing violations from the source.
[0046] This invention employs a dynamic adaptive threshold, which automatically adjusts according to the overall risk situation within a time window. The dynamic threshold calculation formula is as follows:
[0047]
[0048] in, This is the average of the risk scores of all entities within the current time window. Standard deviation, This is the adjustment coefficient;
[0049] When the interception rate of the main body of M consecutive bidding projects is lower than the preset interception rate threshold, the adjustment coefficient is automatically reduced.
[0050] At the same time, an adaptive threshold optimization rule is set: when the interception rate of the main body of M consecutive bidding projects is lower than the preset interception rate threshold, the system automatically reduces the adjustment coefficient k, tightens the risk judgment standard, and improves the risk identification coverage.
[0051] S5. Introduce dynamic risk propagation algorithm and time decay factor to realize the full-domain propagation and simulation of risk in heterogeneous information network, take into account the timeliness of historical risks and the characteristics of risk transmission, and avoid risk omission caused by single static scoring.
[0052] Time decay factor is introduced. ,in >0 represents the preset risk decay rate, used to characterize the degree to which historical risks weaken over time; For the current moment, It is the time when the historical risk event occurred, and it meets the following conditions. ≥ Historical risks that are further back in time have a higher degree of attenuation and a lower weight, which is in line with the actual business logic of risk supervision.
[0053] The dynamic risk propagation algorithm is as follows: for the neighboring nodes of the risk seed node... ,node Risk value after transmission The calculation formula is:
[0054]
[0055] in, For nodes The set of precursor risk nodes, Predecessor node With nodes Edge weights between them Predecessor node The risk value, Predecessor node The latest update time of the risk data.
[0056] S6. The system pre-sets a first threshold and a second threshold, and if the first threshold > the second threshold, it divides the risk into three risk levels based on the subject's final risk score and implements a differentiated hierarchical prevention and control strategy:
[0057] If the final risk score of an entity exceeds the first threshold, it is determined to be an extremely high-risk entity, and real-time blocking action is executed to prohibit it from participating in this bidding process.
[0058] When the second threshold < the final risk score of the entity ≤ the first threshold: the entity is judged to be of medium to high risk, and enhanced manual review will be carried out, with regulatory personnel assigned to conduct key verification of the entity's qualifications, tender documents, and related relationships.
[0059] When the final risk score of an entity is less than or equal to the second threshold, it is determined to be a low-risk entity. Routine risk monitoring actions are performed, including data recording and routine monitoring, without any additional intervention.
[0060] This invention provides a system for identifying and dynamically controlling the liability risks of bidding entities by integrating multi-source data. It is used to execute any of the aforementioned risk identification and dynamic control methods. The overall system architecture consists of four functional modules: a data conflict resolution engine, a heterogeneous graph association calculation unit, a dynamic scoring and execution module, and a risk situation visualization module. These modules collaborate and share data. The specific structure and functions are as follows:
[0061] As the system's data entry layer, the data conflict resolution engine is responsible for the unified access of four types of multi-source heterogeneous data: industrial and commercial data, judicial data, transaction data, and behavior logs. It performs data verification, conflict resolution, and attribute labeling according to the S1 conflict detection rules, and finally outputs a standardized and reliable fused dataset, and synchronizes the data with downstream modules.
[0062] The heterogeneous graph association calculation unit, as the core layer of the system for feature calculation and network modeling, mainly includes the following functions: storing the heterogeneous information network composed of three types of nodes—tenderers, bidders, and agents—and their interrelationships; using frequent itemset mining algorithms to statistically analyze the co-occurrence patterns of device fingerprints; completing the extraction of equity control features, calculation of the association strength between nodes, and dynamic risk propagation calculation, and pushing the calculation results to the dynamic scoring and execution module;
[0063] The dynamic scoring and execution module serves as the system's decision-making and execution layer. Its core functions include: running the initial risk atomic value and quantitatively calculating the final risk score; executing dynamic adaptive threshold determination and hierarchical prevention and control logic; connecting to the standard interface of the electronic bidding platform to issue interception instructions to the platform before the bid opening to block high-risk entities from bidding; and simultaneously outputting a structured risk report while generating the interception instructions.
[0064] The structured risk report includes fragments of original risk evidence, data generation timestamps, and unique fingerprints of source data, enabling the solidification of electronic data and full traceability, fully meeting the compliance requirements for judicial evidence collection and regulatory record keeping;
[0065] The risk situation visualization module serves as the system's human-computer interaction and monitoring display layer. It uses a network graph format to visualize heterogeneous information networks and risk propagation paths between bidding entities. It automatically identifies and highlights high-risk entity nodes and risk transmission network edges. It supports panoramic monitoring of the overall risk situation, single-entity association graph query, and historical risk backtracking, providing regulators with an intuitive and visual regulatory interface.
[0066] In this implementation plan: The system first uses a data conflict resolution engine as a unified data entry point, connecting in real time to external systems to centrally collect four types of multi-source heterogeneous data: industrial and commercial data, judicial data, transaction data, and behavior logs. It then performs data conflict detection, contradiction resolution, and attribute labeling according to preset rules, removing invalid contradictory data to generate a standardized and reliable fused dataset, which is then synchronized to the heterogeneous graph association calculation unit. The heterogeneous graph association calculation unit, based on the received fused dataset, builds a heterogeneous information network containing three types of nodes: tenderers, bidders, and agents. Using frequent itemset mining algorithms to statistically analyze device fingerprint co-occurrence patterns, it completes equity control feature extraction, node association strength calculation, and dynamic risk propagation calculation with the introduction of a time decay factor. All feature data, association results, and risk propagation data are then uniformly pushed to the dynamic scoring and execution module. The dynamic scoring and execution module quantifies the initial risk atomic value and the final risk score according to predetermined rules. The system calculates and combines dynamic adaptive thresholds and two-level tiered thresholds to conduct risk assessment and tiered prevention and control. It generates interception instructions for high-risk entities and sends them to the electronic bidding platform through a standard interface to block non-compliant entities from participating in bidding activities before the bidding opens. At the same time, it extracts risk evidence, timestamps, and source data fingerprints to generate structured risk reports, achieving electronic data solidification, full traceability, and compliant evidence collection. The risk situation visualization module synchronizes all data such as network structure, risk scores, and risk propagation paths in real time, intuitively displaying bidding entities and risk transmission relationships in the form of a network graph. It automatically highlights high-risk nodes and risk links, supports panoramic monitoring, single-entity queries, and historical risk retrospectives, and provides a visual operation interface for regulators. The entire system operates automatically from data access, fusion processing, intelligent calculation, risk decision-making to situation display, realizing intelligent identification, dynamic prevention and control, and full-process supervision of the responsibility risks of bidding entities.
[0067] This solution integrates four types of heterogeneous data from multiple sources: business registration, judicial records, transaction logs, and behavioral logs. It is accompanied by comprehensive data conflict resolution rules, effectively breaking down the data silos prevalent in the industry. It automatically identifies and handles contradictions in multi-source information, ensuring the authenticity and reliability of the analyzed data and laying a solid data foundation for risk identification. Simultaneously, this invention constructs a heterogeneous information network to deeply mine hidden interest relationships such as equity connections, shared equipment, and business dealings. Combined with risk propagation algorithms and quantitative models, it can accurately identify hidden violations such as bid rigging and collusion, overcoming the shortcomings of traditional single-data source identification dimensions and insufficient depth of data mining.
[0068] The system adopts a control model that combines dynamic adaptive thresholds with tiered prevention and control, overcoming the shortcomings of traditional fixed judgment standards and static risk scoring. It can dynamically adjust judgment rules based on the overall market risk situation, effectively reducing the probability of missed or incorrect risk assessments. The system can automatically intercept high-risk entities before bidding, achieving pre-emptive risk prevention and completely improving the problem of traditional supervision being primarily reactive and lagging in its control.
[0069] It can automatically generate structured risk reports with complete traceability information, meeting the compliance requirements for regulatory record keeping and judicial evidence collection. Combined with network graph visualization capabilities, it can intuitively present the risk transmission path, significantly reducing the workload of manual review while balancing regulatory efficiency and control precision. This comprehensively improves the intelligence and refinement of bidding supervision, effectively maintaining a fair and orderly operating environment in the bidding market.
[0070] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for identifying and dynamically controlling the liability risks of bidding entities by integrating multi-source data, characterized in that: The method described above collects and integrates heterogeneous data from multiple sources, including business registration data, judicial data, transaction data, and behavioral logs. It then combines this data with a heterogeneous information network model and risk quantification calculations to identify risks associated with bidding entities and intercept those violating regulations. Specifically, the method includes the following steps: S1. Perform data conflict detection and fusion resolution on the same attribute items of the same bidding entity, remove contradictory data, and output a reliable fused dataset; S2. Based on the trusted fusion dataset, construct a heterogeneous information network containing three types of nodes: tenderer, bidder, and agent. Extract device fingerprint association features and equity control features and form a composite risk evidence chain to quantify the initial risk atom value corresponding to the bidding node. S3. Calculate the correlation strength between any two bidding nodes according to the preset meta-path, and solve the final risk score of the bidding node by combining the correlation strength with the initial risk atom value. S4. Compare the final risk score with a preset risk threshold. When the final risk score exceeds the threshold, generate an interception command to block the corresponding bidding entity from participating in this bidding activity before the bid opening.
2. The method for identifying and dynamically controlling the liability risks of bidding entities by integrating multi-source data as described in claim 1, characterized in that: In S1, the conflict detection and fusion resolution rules are specifically as follows: If the judicial data update timestamp corresponding to the same attribute item is later than the business data update timestamp, the judicial data will be accepted and the attribute will be marked as a legal priority attribute. If the straight-line distance between the geographical location corresponding to the device fingerprint extracted from the behavior log and the geocode of the business address registered with the industry and commerce administration exceeds a preset distance threshold, it is determined that there is a logical contradiction between the two and a real-time information verification request is triggered.
3. The method for identifying and dynamically controlling the liability risks of bidding entities by integrating multi-source data as described in claim 1, characterized in that: In S2, Based on shareholder and legal representative information in business registration data, calculate the shortest equity connection path between different bidding companies, and generate equity control features based on the shortest equity connection path; Collect front-end operating environment information of bidders when logging into the bidding platform and uploading bid documents, and generate unique device identifiers using a hash algorithm; if multiple bidders share the same device identifier within the same bidding collaboration cycle, generate device fingerprint association features.
4. The method for identifying and dynamically controlling the liability risks of bidding entities by integrating multi-source data as described in claim 1, characterized in that: In step S3, the correlation strength is calculated as follows: in As an edge weight, the meta-path must include at least "bidder-tenderer-bidder" and "bidder-agent-bidder"; The final risk score is calculated as described in S3 as follows: in, For bidding nodes The initial risk atom value, For bidding nodes The final risk score, The propagation coefficient is preset and 0 ≤ ≤1, To be in sync with the bidding nodes A set of other related bidding nodes.
5. The method for identifying and dynamically controlling the liability risks of bidding entities by integrating multi-source data as described in claim 1, characterized in that: The initial risk atomic value in S2 is quantified and assigned a value in conjunction with the bidding collusion scenario, and the assignment rule is as follows: When the number of times the associated fingerprint features of a device co-occurs is greater than or equal to 3, the corresponding risk value is set to 0.8; When the shortest connection path length of equity in the equity control feature is equal to 1, the corresponding risk value is set to 0.9; Anomaly chains in bid prices are extracted from the trusted fusion dataset. When the number of valid bidders in the same bid section is less than 5 and the coefficient of variation of bid prices is greater than 0.2, the corresponding risk value is set to 0.
6. The above quantitative thresholds were all determined based on statistical analysis of historical bidding and tendering violations.
6. The method for identifying and dynamically controlling the liability risks of bidding entities by integrating multi-source data as described in claim 1, characterized in that: The preset threshold adopts a dynamic adaptive threshold: in, This is the average of the risk scores of all entities within the current time window. Standard deviation, This is the adjustment coefficient; When the interception rate of the main body of M consecutive bidding projects is lower than the preset interception rate threshold, the adjustment coefficient is automatically reduced. The method also includes a tiered prevention and control step: a first threshold and a second threshold are preset, with the first threshold being greater than the second threshold; when the subject's final risk score exceeds the first threshold, a real-time interception action is performed; when the subject's final risk score is greater than the second threshold but less than or equal to the first threshold, an enhanced manual review action is performed; when the subject's final risk score is less than or equal to the second threshold, a routine risk monitoring action is performed.
7. The method for identifying and dynamically controlling the liability risks of bidding entities by integrating multi-source data as described in claim 1, characterized in that: The method also includes a risk propagation step within a heterogeneous information network: A dynamic risk propagation algorithm is employed to propagate the risk value of the initial risk seed node along the network edges in a heterogeneous information network; a time decay factor is introduced. Historical risk events are attenuated and accumulated, among which >0 represents the preset risk attenuation rate. The moment when historical risk events occurred, the current moment. satisfy ≥ ; The dynamic risk propagation algorithm specifically involves: for the adjacent nodes of the risk seed node... ,node Risk value after transmission The calculation formula is: in, For nodes The set of precursor risk nodes, Predecessor node With nodes Edge weights between them Predecessor node The risk value, Predecessor node The latest update time of the risk data.
8. A system for identifying and dynamically controlling the liability risks of bidding entities by integrating multi-source data, used to execute the method described in any one of claims 1 to 7, characterized in that, This includes a data conflict resolution engine, a heterogeneous graph association calculation unit, and a dynamic scoring and execution module; The data conflict resolution engine is configured to receive heterogeneous data from multiple sources, such as business, judicial, transaction, and behavior logs, and to complete data conflict detection, fusion and resolution processing, and output a trusted fusion dataset. The heterogeneous graph association calculation unit is configured to store a heterogeneous information network consisting of three types of main nodes: the tenderer, the bidder, and the agency, as well as the association relationships between the nodes. It uses a frequent itemset mining algorithm to statistically analyze the co-occurrence pattern of device fingerprints and completes the calculation of equity control features and node association strength. The dynamic scoring and execution module is configured to perform risk quantification calculation, adaptive threshold determination, and violation interception logic, and it is connected to the standard interface of the electronic bidding platform to intercept high-risk bidding entities before the bid opening.
9. The bidding entity liability risk identification and dynamic prevention and control system integrating multi-source data as described in claim 8, characterized in that, The dynamic scoring and execution module generates an interception command and simultaneously outputs a structured risk report. The structured risk report includes fragments of original risk evidence, data generation timestamps, and unique fingerprints of the source data, meeting the requirements for electronic data fixation, traceability, and judicial evidence collection.
10. The bidding entity liability risk identification and dynamic prevention and control system integrating multi-source data as described in claim 8, characterized in that, Also includes: The risk situation visualization module is configured to display the risk propagation path between bidding entities in the form of a network graph, automatically identify and mark high-risk entity nodes and risk transmission network edges, and realize panoramic visualization monitoring of the risk situation.
