Abnormal bid procurement quotation risk identification system

Abstract

The invention, which relates to the technical field of bid-invitation purchasing, discloses a bid-invitation purchasing abnormal quotation risk identification system comprising a database module, a data preprocessing module, an abnormal quotation identification module, a risk grading early warning module and a bid-invitation prediction module. And the database module is used for integrating total credible data and providing data input for each module. The bid invitation purchasing abnormal quotation risk identification system can perform abnormal quotation identification, identify overhigh quotation or abnormal deviation price, verify quotation data, perform quotation behavior tracing, perform cross-project review on finished historical projects, and identify potential risks; in addition, the bid invitation prediction module analyzes historical bid invitation data, industry trend data and policy oriented data, and is beneficial to bid invitation prediction based on a multi-model fusion technology, thereby helping an enterprise to turn from passive response bid invitation to active layout, capturing potential bid invitation opportunities in advance, and matching suppliers through classification.

Description

Technical Field

[0001] This invention relates to the field of bidding and procurement technology, specifically to a system for identifying risks associated with abnormal bidding and procurement prices. Background Technology

[0002] Tender procurement refers to the process by which the procuring party, as the tendering party, proposes the conditions and requirements for procurement in advance, invites numerous companies to participate in the bidding, and then selects the best trading partner from among them in a one-time process according to the prescribed procedures and standards, and signs an agreement with the bidder who offers the most favorable conditions. In the tender procurement process, it is necessary to identify risks associated with abnormal quotations in order to avoid risks.

[0003] For example, the invention with application number 202511211188.9 discloses an intelligent system for risk warning of bidding documents. Similar risk identification systems currently have the following shortcomings: although they can perform risk assessment on bidding documents, they only focus on some anomalies and are difficult to identify violations such as bid rigging, collusion, and deliberate circumvention of rules. Furthermore, they do not predict potential bidding opportunities in advance, and enterprises can only respond passively. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a system for identifying risks associated with abnormal pricing in bidding and procurement, thus resolving the problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a system for identifying abnormal bidding prices in bidding and procurement, comprising a database module, a data preprocessing module, an abnormal bidding price identification module, a risk classification and early warning module, and a bidding prediction module; The database module is used to integrate all reliable data and provide data input for each module. The data preprocessing module includes an announcement identification and extraction module, an announcement structuring processing module, and a target classification and matching module; The abnormal pricing identification module shown includes a multi-dimensional price comparison module, a pricing data verification module, a pricing behavior tracing module, and a historical project review module. The multi-dimensional price comparison module is used to access enterprise-specific price index data as a benchmark for judging the reasonableness of prices, and integrate multi-dimensional price data, including but not limited to the mode price, the highest price, the lowest price, the average price of government procurement, the average price of e-commerce, and the average price of enterprises; finally, it compares the bid price with the above-mentioned multi-dimensional prices to identify excessively high bids or abnormally deviating prices. The quotation data verification module is used to perform format verification, logic verification, and threshold verification. Format verification is used to check the completeness of the quotation and the legality of the data type; logic verification is used to verify whether the quotation responds to the bidding technical parameters, whether it is within the budget, whether the sum of the unit prices is consistent with the total price, and whether the tax rate calculation is compliant; threshold verification is used to compare the quotation with the project budget, historical winning bid prices, and market benchmark prices to determine whether it exceeds the preset threshold. The bidding behavior tracing module is used to capture the entire online bidding process in real time, generate a behavior trajectory map, and identify evasive operations. The historical project review module is used to conduct cross-project reviews of completed historical projects to identify potential risks such as regular pricing and collusive pricing. The risk classification and early warning module is used to identify and classify risks based on the identification results of the abnormal quotation identification module, and the risk classification and early warning module includes a risk analysis and assessment module, a risk visualization display module, and a risk intelligent early warning module. The bidding prediction module includes a data acquisition and processing module, a data analysis module, a feature extraction module, a multi-model fusion prediction module, and a classification and matching module.

[0006] Furthermore, the database module includes basic data, standard data, and related data. The basic data includes, but is not limited to, bidding announcements and price quotations. The standard data includes, but is not limited to, a product database and a price index database. The related data includes, but is not limited to, a company database.

[0007] Furthermore, the announcement recognition and extraction module is used to perform multimodal recognition and extraction of announcements, covering image OCR recognition, multi-attachment parsing, and converting the parsed announcements into a unified MD document, providing standardized input for subsequent structured processing.

[0008] Furthermore, the announcement structured processing module extracts key entities and relationships in the bidding and tendering process from the standardized MD document by calling the large model and RAG technology, and outputs JSON-formatted structured data (including 20+ core fields such as project name, purchaser, package, candidate, publicized amount, and various times).

[0009] Furthermore, the target classification and matching module is used to perform multi-level classification of targets in structured data, construct vectors and semantic similarity, and achieve accurate matching between targets and product categories by constructing vector and reordering models.

[0010] Furthermore, the risk analysis and assessment module is used to integrate risk information from all dimensions, including price comparison, data verification, behavior tracing, and historical review. It also constructs a behavior risk scoring model based on historical cases, assigns values ​​to abnormal behaviors such as high-frequency modifications and concentrated operations late at night, and overlays quotation data to conduct a comprehensive risk assessment.

[0011] Furthermore, the risk visualization display module is used to generate multi-dimensional risk analysis reports and visualize the risk analysis and assessment results. The report content includes, but is not limited to, abnormal price distribution, risk type proportion, high-risk enterprise ranking, and industry risk trends. The risk intelligent early warning module is used to classify and grade the risk analysis and assessment results and intelligently push the risk identification results to relevant review nodes to realize risk early warning.

[0012] Furthermore, the data acquisition and processing module is used to collect historical bidding data, industry trend data, and policy guidance data, and to clean, structure, transform, and store the raw data to generate high-quality standardized data. The data analysis module extracts the periodic and correlation characteristics of historical bidding projects through time series analysis, association rule mining, and clustering algorithms. At the same time, it predicts growth rates, identifies demand gaps, and analyzes the dynamic impact of competitors based on industry data. It also extracts and classifies policy elements, and then quantitatively models the policy impact, transforming policy texts into quantifiable bidding driving factors.

[0013] Furthermore, the feature extraction module is used to integrate the three-dimensional analysis results of the data analysis module, construct a unified and efficient predictive feature system, generate time features, industry features, policy features and correlation features, and then normalize the features.

[0014] Furthermore, the multi-model fusion prediction module outputs accurate prediction results for potential bidding projects through multi-model selection, fusion, and verification. Finally, it fuses the basic model output to predict future potential bidding projects and generates a model prediction report. The classification and matching module is used to recommend suitable suppliers for the predicted bidding projects based on the target object classification and matching results, and prioritizes matching suppliers with strong collaboration and no risk record.

[0015] This invention provides a system for identifying abnormal pricing risks in bidding and procurement, which has the following beneficial effects: 1. This bidding and procurement abnormal pricing risk identification system integrates enterprise-specific price indices and multi-dimensional price data through a multi-dimensional price comparison module. Combined with a triple verification mechanism in the pricing data verification module, it can quickly screen for explicit problems such as missing items and incorrect data formats, and accurately identify price anomalies that exceed budget thresholds or deviate from the market average. The pricing behavior tracing module captures the trajectory of the entire online pricing process and, together with the cross-project correlation analysis of the historical project review module, facilitates the identification of hidden risks, thereby enabling the identification of violations such as bid rigging, collusion, and deliberate rule circumvention.

[0016] 2. This bidding and procurement abnormal pricing risk identification system utilizes OCR recognition and large-scale model + RAG technology in its data preprocessing module to facilitate rapid parsing and structured extraction of announcements from multiple sources. The risk visualization module generates multi-dimensional reports, intuitively presenting key information such as the distribution of abnormal pricing and the ranking of high-risk companies. The risk grading and early warning module, through classification and intelligent push notifications, enables managers to quickly grasp the core risks, thereby reducing the rejection rate and subsequent dispute costs through early warnings. Furthermore, the bidding prediction module analyzes historical bidding data, industry trend data, and policy guidance data. Based on multi-model fusion technology, it facilitates bidding prediction, helping companies shift from passively responding to bidding to proactively planning and capturing potential bidding opportunities in advance. Combined with the classification and matching module, it recommends compliant and high-quality suppliers, helping companies optimize resource allocation, increase the success rate of bidding, and reduce the time and financial losses caused by blind bidding. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the system flow of a bidding and procurement abnormal quotation risk identification system according to the present invention. Detailed Implementation

[0018] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and should not be construed as limiting the scope of the invention.

[0019] like Figure 1 As shown, the present invention provides a technical solution: a system for identifying abnormal bidding and procurement risks, including a database module, a data preprocessing module, an abnormal bidding identification module, a risk classification and early warning module, and a bidding prediction module; The database module is used to integrate all reliable data and provide data input for each module. The database module includes basic data, standard data and related data. Basic data includes, but is not limited to, bidding announcements and price lists. Standard data includes, but is not limited to, product databases and price index databases. Related data includes, but is not limited to, enterprise databases. The data preprocessing module includes an announcement recognition and extraction module, an announcement structuring module, and a target item classification and matching module. The announcement recognition and extraction module is used for multimodal recognition and extraction of announcements, covering image OCR recognition and multi-attachment parsing, and converting the parsed announcements into a unified MD document to provide standardized input for subsequent structuring processing. The announcement structuring module extracts key entities and relationships in bidding and tendering from the standardized MD document by calling large models and RAG technology, and outputs JSON format structured data, including more than 20 core fields such as project name, purchaser, package, candidate, publicized amount, and various times. The target item classification and matching module is used to perform multi-level classification of targets in the structured data, construct vectors and semantic similarity, and achieve accurate matching of targets and product categories by constructing vector and re-ranking models. The abnormal pricing identification module includes a multi-dimensional price comparison module, a pricing data verification module, a pricing behavior tracing module, and a historical project review module. The multi-dimensional price comparison module accesses enterprise-specific price index data as a benchmark for price reasonableness judgment and integrates multi-dimensional price data, including but not limited to the mode price, highest price, lowest price, government procurement average price, e-commerce average price, and enterprise average price. Finally, it compares the bid price with the aforementioned multi-dimensional prices to identify excessively high prices or abnormal price deviations. The pricing data verification module performs format verification, logic verification, and threshold verification. Format verification checks the completeness of the bid and the legality of data types. Logic verification verifies whether the price responds to the bidding technical parameters, is within the budget, the consistency of the sum of unit prices with the total price, and the compliance of tax rate calculation. Threshold verification compares the price with the project budget, historical winning bid prices, and market benchmark prices to determine if it exceeds a preset threshold. The pricing behavior tracing module captures the entire online pricing process in real time and generates a behavior trajectory map, while also identifying evasive operations. The historical project review module conducts cross-project reviews of completed historical projects to identify potential risks of regular pricing and collusive pricing. The risk grading and early warning module is used to identify and grade risks based on the results of the abnormal pricing identification module. This module includes a risk analysis and assessment module, a risk visualization display module, and a risk intelligent early warning module. The risk analysis and assessment module integrates multi-dimensional risk information from price comparison, data verification, behavior tracing, and historical review. It constructs a behavioral risk scoring model based on historical cases, assigning values ​​to abnormal behaviors such as high-frequency modifications and concentrated late-night operations, and overlaying pricing data for comprehensive risk assessment. The risk visualization display module generates multi-dimensional risk analysis reports, visually displaying the risk analysis and assessment results. Report content includes, but is not limited to, abnormal pricing distribution, risk type percentages, high-risk enterprise rankings, and industry risk trends. The risk intelligent early warning module categorizes and grades the risk analysis and assessment results and intelligently pushes the risk identification results to relevant review nodes to achieve risk warnings. The bidding forecasting module includes a data acquisition and processing module, a data analysis module, a feature extraction module, a multi-model fusion forecasting module, and a classification and matching module. The data acquisition and processing module collects historical bidding data, industry trend data, and policy guidance data, and cleans, structures, transforms, and stores the raw data to generate high-quality standardized data. The data analysis module extracts the periodic and correlational characteristics of historical bidding projects through time series analysis, association rule mining, and clustering algorithms; it also predicts growth rates based on industry data, identifies demand gaps, and analyzes the dynamic impact of competitors; it extracts and classifies policy elements, and then quantitatively models the policy impact, transforming policy texts into quantifiable bidding driving factors. The feature extraction module integrates the three dimensions of the data analysis module. The analysis results are used to construct a unified and efficient predictive feature system, generating time features, industry features, policy features, and correlation features, and then normalizing these features. The multi-model fusion prediction module selects, fuses, and validates multiple models to output accurate prediction results for potential bidding projects. Periodic projects are trained using LSTM and Prophet models, policy-driven projects using XGBoost and LightGBM models, and complex projects using Transformer models. Finally, the outputs of the base models are fused to predict future potential bidding projects and generate a model prediction report. The classification and matching module recommends suitable suppliers for predicted bidding projects based on the target object classification and matching results, prioritizing suppliers with strong collaboration and no risk records.

[0020] Example: Taking a provincial state-owned enterprise's collaborative procurement platform as an application scenario; First, the database module integrates all the data. The basic data covers 150,000 bidding announcements and 80,000 quotation sheets from the past 5 years. The standard data includes 200,000 products in a 5-level classification database and a price index database for enterprises. The related data includes equity, qualification and other information of 50,000 enterprises. Next, the data preprocessing module efficiently processes announcements from multiple media sources. For PDF-format offshore wind power equipment procurement announcements, it uses OCR to identify technical parameters in the images, parses the itemized budgets in the Excel attachments, and converts them into a unified MD document. Then, it extracts more than 20 core fields using large model and RAG technology, outputting structured JSON data. Finally, it matches "5MW wind turbine components" to the 5-level category of "new energy equipment - wind power equipment - wind turbine components - permanent magnet synchronous wind turbine - 5MW level", and associates it with standard data of government procurement average price of 6.5 million yuan / set. Then, the abnormal pricing identification module, through the price multi-dimensional comparison module, discovered that a bidder's price of 8.2 million yuan per set was 26.15% higher than the average government procurement price, marking it as an abnormal high price; the pricing data verification module detected that another bid had a compliant format and consistent logic, but threshold verification showed that its price was slightly higher than the preset threshold value; the pricing behavior tracing module captured that a bidder modified its price three times near the deadline, ultimately matching the price of another company, and that the IP addresses belonged to the same city; the historical project review module, through cross-project analysis, discovered that the two companies jointly bid in three projects and that the prices changed regularly, determining the risk of bid rigging. The risk analysis and assessment module of the risk classification and early warning module integrates multi-dimensional risk information, calculates a comprehensive risk score of 85, and determines it to be high risk; the risk visualization display module generates a report, which intuitively shows that the proportion of high-risk bids in the new energy field is 65% and the proportion of bid-rigging risk is 40%; the risk intelligent early warning module pushes the early warning information to the review node, and simultaneously attaches screenshots of behavioral trajectories and historical related risks. The bidding prediction module can also collect and standardize historical bidding, industry trend, and policy data through its data processing module; the data analysis module can mine the 6-month bidding cycle of hydrogen refueling station projects and predict a 22% increase in new energy bidding volume in 2025; the feature extraction module can generate standardized features such as time, industry, and policy; multi-model fusion can predict that 3 hydrogen refueling station construction projects will be launched in Q2 of 2025, with a budget of 150 million yuan and a probability of occurrence of 89%; and the classification and matching module can recommend 5 high-quality suppliers with no risk record to help companies prepare in advance. After the implementation of this embodiment, the efficiency of abnormal quotation screening is improved, the accuracy of hidden risk identification reaches 88%, the procurement rejection rate drops from 12% to 3.5%, enterprises can capture bidding opportunities 3-6 months in advance, and the efficiency of winning bid response is improved.

[0021] Based on the above, this invention provides a bidding and procurement abnormal pricing risk identification system. The abnormal pricing identification module integrates enterprise-specific price indices and multi-dimensional price data through a multi-dimensional price comparison module, combined with a triple verification mechanism in the pricing data verification module. This allows for rapid screening of explicit issues such as missing items and incorrect data formats, while accurately identifying price anomalies exceeding budget thresholds and deviating from the market average. The pricing behavior tracing module captures the entire online pricing process and, combined with cross-project correlation analysis from the historical project review module, facilitates the identification of hidden risks. The data preprocessing module utilizes OCR recognition and large-scale model + RAG technology to facilitate rapid parsing and structured extraction of announcements from multiple sources, enabling risk visualization. The presentation module can generate multi-dimensional reports, intuitively displaying key information such as abnormal price distribution and high-risk enterprise rankings. The risk classification and early warning module, through grading and intelligent push, enables managers to quickly grasp the core risks, thereby facilitating the reduction of bid rejection rates and subsequent dispute costs through early warnings. In addition, the bidding prediction module analyzes historical bidding data, industry trend data, and policy guidance data. Based on multi-model fusion technology, it is conducive to bidding prediction, thereby helping enterprises shift from passively responding to bidding to proactively planning and capturing potential bidding opportunities in advance. Combined with the classification and matching module, it recommends compliant and high-quality suppliers, helping enterprises optimize resource allocation, increase the success rate of bidding, and reduce the time and money losses caused by blind bidding.

[0022] The embodiments of the present invention are given for illustrative and descriptive purposes only, and are not intended to be exhaustive or to limit the invention to the forms disclosed. Many modifications and variations will be apparent to those skilled in the art. The embodiments were chosen and described in order to better illustrate the principles and practical application of the invention, and to enable those skilled in the art to understand the invention and to design various embodiments with various modifications suitable for a particular purpose.

Claims

1. A system for identifying abnormal pricing risks in bidding and procurement, characterized in that: It includes a database module, a data preprocessing module, an abnormal quotation identification module, a risk classification and early warning module, and a bidding prediction module; The database module is used to integrate all reliable data and provide data input for each module. The data preprocessing module includes an announcement identification and extraction module, an announcement structuring processing module, and a target classification and matching module; The abnormal pricing identification module shown includes a multi-dimensional price comparison module, a pricing data verification module, a pricing behavior tracing module, and a historical project review module. The multi-dimensional price comparison module is used to access enterprise-specific price index data as a benchmark for judging the reasonableness of prices, and integrate multi-dimensional price data, including but not limited to the mode price, the highest price, the lowest price, the average price of government procurement, the average price of e-commerce, and the average price of enterprises; finally, it compares the bid price with the above-mentioned multi-dimensional prices to identify excessively high bids or abnormally deviating prices. The quotation data verification module is used to perform format verification, logic verification and threshold verification, wherein the format verification is used to check the integrity of the quotation and the legality of the data type; Logical verification is used to check whether the quotation responds to the technical parameters of the tender, whether it is within the budget, whether the sum of the unit prices is consistent with the total price, and whether the tax rate calculation is compliant. Threshold verification is used to compare the quoted price with the project budget, historical winning bid prices, and market benchmark prices to determine whether it exceeds a preset threshold. The bidding behavior tracing module is used to capture the entire online bidding process in real time, generate a behavior trajectory map, and identify evasive operations. The historical project review module is used to conduct cross-project reviews of completed historical projects to identify potential risks such as regular pricing and collusive pricing. The risk classification and early warning module is used to identify and classify risks based on the identification results of the abnormal quotation identification module, and the risk classification and early warning module includes a risk analysis and assessment module, a risk visualization display module, and a risk intelligent early warning module. The bidding prediction module includes a data acquisition and processing module, a data analysis module, a feature extraction module, a multi-model fusion prediction module, and a classification and matching module.

2. The bidding and procurement abnormal quotation risk identification system according to claim 1, characterized in that: The database module includes basic data, standard data, and related data. The basic data includes, but is not limited to, bidding announcements and price quotations. The standard data includes, but is not limited to, a product database and a price index database. The related data includes, but is not limited to, a company database.

3. The bidding and procurement abnormal quotation risk identification system according to claim 1, characterized in that: The announcement recognition and extraction module is used to perform multimodal recognition and extraction of announcements, covering image OCR recognition, multi-attachment parsing, and converting the parsed announcements into a unified MD document, providing standardized input for subsequent structured processing.

4. The bidding and procurement abnormal quotation risk identification system according to claim 1, characterized in that: The announcement structured processing module extracts key entities and relationships in the bidding and tendering process from the standardized MD document by calling the large model and RAG technology, and outputs JSON-formatted structured data (including 20+ core fields such as project name, purchaser, package, candidate, publicized amount, and various times).

5. The bidding and procurement abnormal quotation risk identification system according to claim 1, characterized in that: The target classification and matching module is used to perform multi-level classification of targets in structured data, construct vectors and semantic similarity, and achieve accurate matching between targets and product categories by constructing vector and reordering models.

6. The bidding and procurement abnormal quotation risk identification system according to claim 1, characterized in that: The risk analysis and assessment module integrates risk information from all dimensions, including price comparison, data verification, behavior tracing, and historical review. It also constructs a behavior risk scoring model based on historical cases, assigns values ​​to abnormal behaviors such as high-frequency modifications and concentrated operations late at night, and overlays quotation data to conduct a comprehensive risk assessment.

7. The bidding and procurement abnormal quotation risk identification system according to claim 1, characterized in that: The risk visualization display module is used to generate multi-dimensional risk analysis reports and visualize the risk analysis and assessment results. The report content includes, but is not limited to, abnormal price distribution, risk type proportion, high-risk enterprise ranking, and industry risk trends. The intelligent risk warning module is used to classify and grade the risk analysis and assessment results, and intelligently push the risk identification results to relevant review nodes to realize risk warning.

8. The bidding and procurement abnormal quotation risk identification system according to claim 1, characterized in that: The data acquisition and processing module is used to collect historical bidding data, industry trend data, and policy guidance data, and to clean, structure, transform, and store the raw data to generate high-quality standardized data. The data analysis module extracts the periodic and correlational characteristics of historical bidding projects through time series analysis, association rule mining, and clustering algorithms; it also predicts growth rates, identifies demand gaps, and analyzes the dynamic impact of competitors based on industry data; and it extracts and classifies policy elements, then quantifies and models the policy impact, transforming policy texts into quantifiable bidding driving factors.

9. The bidding and procurement abnormal quotation risk identification system according to claim 1, characterized in that: The feature extraction module is used to integrate the three-dimensional analysis results of the data analysis module, construct a unified and efficient predictive feature system, generate time features, industry features, policy features and correlation features, and then normalize the features.

10. The bidding and procurement abnormal quotation risk identification system according to claim 1, characterized in that: The multi-model fusion prediction module selects, fuses, and verifies multiple models to output accurate prediction results for potential bidding projects. Finally, it fuses the output of the basic model to predict future potential bidding projects and generates a model prediction report. The classification and matching module recommends suitable suppliers for the predicted bidding projects based on the classification and matching results of the target, and prioritizes matching suppliers with strong collaboration and no risk record.

Images