Bidding person association mode identification method oriented to collection process

Abstract

The invention discloses a bidder association mode identification method oriented to a collection flow, and belongs to the technical field of collection supervision, and the method comprises the following steps: S100, bidder data uniform characterization: employing an interface butt joint mode, static information data, dynamic information data, document information data and the like of different bidders participating in the same bid packet are collected from the collection system; s200, bidder data preprocessing is carried out; s300, analyzing static data of the bidder; s400, analyzing the dynamic data of the bidder; s500, bidder document data analysis is carried out; and S600, identifying the association mode between the bidders. According to the method, the abnormal behaviors can be quickly identified by analyzing the behavior modes of the bidder in different bid invitation stages, so that the monitoring efficiency of a supervision mechanism on the bid invitation process is improved, the compliance of bid invitation activities is ensured, unauthorized competitive behaviors such as bid enclosing, bid confusion and the like are disclosed and prevented, and the fairness of a bid invitation market is maintained.

Description

Technical Field

[0001] The present invention relates to the technical field of bidding and procurement supervision, and in particular to a method for identifying association patterns between bidders in a bidding and procurement process. Background Art

[0002] Bidding and procurement (abbreviated as bidding and procurement) selects the best suppliers or service providers through fair, just and transparent competition mechanisms and procurement processes to ensure the high quality and economic benefits of the project. However, faced with challenges such as complex competition environment, diversified bidder behaviors and the difficulty of comprehensive regulatory measures, the bidding and procurement process still has unfair competition, which undermines the fairness of the market, leads to the decline of project or service quality, damages the interests of the country and companies, and affects the rights of consumers. Although countries have passed strict laws and regulations to regulate bidding and procurement, such as China's Bidding Law and Anti-Unfair Competition Law, the bidding and procurement process involves a large amount of data and information, and traditional regulatory methods often find it difficult to identify and combat violations in a timely and comprehensive manner.

[0003] Analysis of bidder behavior patterns is the key to understanding how bidders compete in bidding and procurement. Normal behavior patterns include bidders preparing, submitting and responding to documents in accordance with specifications, and maintaining reasonable consistency in operation time, IP address, geographic location and content of bid documents. However, some bidders will take abnormal measures, such as bid rigging, allocation of market share, etc., to exclude competitors and manipulate market prices, thereby gaining unfair competitive advantages. Such abnormal behavior is usually manifested in the similarity of operation time between multiple bidders, the use of the same IP address, or highly consistent content of bid documents. Such behavior is highly concealed and poses great challenges to supervision. Summary of the invention

[0004] In order to solve the defects and deficiencies of the existing technology, the present invention proposes a method for identifying association patterns between bidders for the bidding process. By effectively detecting and identifying abnormal behavior patterns, the present invention introduces the technical means of big data analysis, machine learning algorithms and community network construction, deeply mines and analyzes the bidders' operation time, IP address, file characteristics and other multi-dimensional data, builds a behavior model and detects abnormal behavior patterns, and reveals potential abnormal behaviors and associations. This method not only improves the ability to discover abnormal behaviors, but also monitors the bidding process in real time, identifies and records suspicious patterns, ensures that regulatory authorities can take timely measures, significantly improves the fairness and transparency of the bidding and procurement process, and ensures the maintenance of market competition order.

[0005] The technical solution adopted by the present invention is: a method for identifying association patterns between bidders in a bidding process, comprising the following steps:

[0006] S100: Unified representation of bidder data: Use interface docking to collect static information data, dynamic information data, document information data, etc. of different bidders participating in the same bid package from the procurement system.

[0007] S200: Bidder data preprocessing.

[0008] S300: Bidder static data analysis.

[0009] S400: Dynamic data analysis of bidders.

[0010] S500: Bidder document data analysis.

[0011] S600: Identification of association patterns between bidders.

[0012] Preferably, the step S100 specifically includes the following steps:

[0013] S110: Unified representation of bidder static data. Static information data refers to the basic information of bidders stored in the bidding system. Static information data includes but is not limited to {company name, corporate legal person, equity, bidder contact information, contact address, email address, senior management};

[0014] S120: Unified representation of bidder dynamic data. Dynamic behavior data refers to data related to each bidding activity conducted by bidders in the bidding system, including but not limited to {browsing and retrieval record data, bidder registration data, bidding document download record data, bidder upload bid record data, bidder second quotation record data, bidder viewing winning notice record data}. Among them, browsing and retrieval record data include but are not limited to {the method by which the bidder obtains the bidding project, the IP address of the bidder's login system, and the MAC address of the bidder's login system}; the bidder registration data include but are not limited to {the bidder registration time, the bidder registration IP address, and the bidder registration MAC address}; the bidding document download record data include but are not limited to {the time of downloading the bidding document, the IP address of downloading the bidding document, and the MAC address of downloading the bidding document}; the bidder upload bid record data include but are not limited to {the bidder submit bid time, the bidder submit bid IP address, and the bidder submit bid MAC address}; the bidder second quotation record data include but are not limited to {the bidder second quotation time, the bidder second quotation IP address, and the bidder second quotation MAC address}; the bidder viewing the winning notice record data include but are not limited to {the bidder viewing time, the bidder IP address, and the bidder MAC address}.

[0015] S130: Unified representation of bidder document data. Document information data refers to the document file data prepared by the bidder for bidding in the tendering and procurement system, including but not limited to {bidding documents, bidding documents, quotation documents, secondary quotation documents, technical clarification documents, negotiation documents}.

[0016] Preferably, the step S200 specifically includes the following steps:

[0017] S210: Preprocessing of bidder static data. The fields in the static information data are divided into two categories, one is the unique field, and the other is the association field. The unique field includes but is not limited to {company name, corporate legal person, bidder contact information, contact address, email address}; the association field includes but is not limited to {equity, senior management}. In order to ensure the efficiency of equity-related companies, shareholders with less than 5% of the company's equity are eliminated.

[0018] S220: Preprocessing of bidder dynamic data: Perform basic data processing on the bidder dynamic data.

[0019] S230: Preprocessing of bidder document data. Perform document parsing on the bidder document data to obtain basic information of the document. Preferably, step S220 specifically includes the following steps:

[0020] S221: Processing the time interval of bidder dynamic data: Calculate the time interval between two adjacent operations for the time field in the bidder dynamic information data.

[0021] S222: Extracting the bidder's dynamic data hardware identification: Record the IP address and MAC address fields in the bidder's dynamic information data as the bidder's dynamic data hardware identification data.

[0022] S223: Extracting bidder dynamic data operation behavior: The behavior mode field in the bidder dynamic information data is recorded as bidder dynamic data operation behavior data.

[0023] Preferably, the step S230 specifically includes the following steps:

[0024] S231: Document information processing. The documents in the bidder's document data are processed in an information manner to obtain the text content of the documents. S232: Document structure information discovery. The text content information in the document is structurally analyzed through paragraph discovery and title structure extraction technology to obtain the document structure information and document title description information in the document file.

[0025] S233: Extraction of evaluation factor standard requirements. Use table extraction technology and information extraction technology to obtain the evaluation factor requirement information in the bidding documents. The evaluation factor requirement information includes but is not limited to {bidder qualification requirements, bidder form requirements, commercial part scoring standards, technical part scoring standards, performance requirements, financial statements, quotation part scoring standards, project manager qualification requirements, enterprise personnel requirements}.

[0026] Preferably, the step S300 specifically includes the following steps:

[0027] S310: Unique field consistency comparison. The unique field in step S210 is used for consistency comparison, and the full text comparison algorithm is used to compare the unique field in the static information of each bidder to obtain the consistency result of the unique field.

[0028] S320: Equity association discovery. Use the equity field in the association field in step S210 to discover the association, and use the equity-related enterprise comparison analysis algorithm to calculate the consistency result of the bidder's equity association for the equity field information in the association field in the static information. In order to ensure the efficiency of the equity-related enterprise comparison, only the shareholders with equity holdings higher than 5% of the enterprise are calculated for consistency.

[0029] S330: Similarity analysis of senior management of enterprises. The senior management field in the association field in step S210 is used to calculate the similarity analysis of senior management of enterprises. For the senior management field information in the association field in the static information, the personnel name similarity comparison and analysis algorithm is used to calculate the similarity analysis results of senior management of the bidder's enterprise. When the similarity value of senior management is higher than Sim1, Sim1 of this patent is set to 0.85, and the consistency results of the two enterprises are set to 1.

[0030] S340: Identification of bidder static data patterns. Calculate the consistency results of all bidder static information to obtain the field consistency matrix between bidders, and record the maximum value of the static information consistency results between bidders as the bidder's static information similarity result, and obtain the bidder's static data pattern probability matrix.

[0031] Preferably, the step S400 specifically includes the following steps:

[0032] S410: Time interval volatility analysis: Utilize the group network algorithm to calculate the correlation results of the operation time intervals between different stages.

[0033] S420: Hardware information consistency comparison. Use the full text comparison algorithm to calculate the hardware identification data in the dynamic information of each bidder to obtain a consistency result. When the hardware identification data of the bidders are consistent, the correlation result between the bidders is set to 1.

[0034] S430: Analysis of bidder edge nodes. Special processing is performed on some operation behaviors in the dynamic behavior data. This part of the operation behavior includes but is not limited to {browsing and retrieving record data}. Under normal circumstances, bidders browse the bidding projects on the tendering and procurement system normally. When there are bidders in a certain bidding project who exceed the ratio and obtain the bidding project by retrieving the name of the bidding project through information, then the bidders in the bidding project have abnormal patterns to a large extent. Therefore, it is necessary to perform weighted calculation of the correlation results of the bidders through information retrieval. The ratio of this patent is set to 80%. If the bidders of the bidding project obtain the bidding project through information retrieval and 80% of the bidders of the bidding project obtain the bidding project through information retrieval, then the correlation results between all bidders need to be multiplied by 1.07.

[0035] S440: Dynamic data pattern recognition of bidders. Through S410, S420 and S430, the correlation results of all dynamic information of bidders are calculated to obtain the correlation result matrix of different fields between bidders, and the maximum value of the dynamic information correlation result between bidders is recorded as the dynamic correlation result of the bidder, thus obtaining the dynamic pattern probability matrix of the bidder.

[0036] Preferably, the step S500 specifically includes the following steps:

[0037] S510: Positioning the description of evaluation factors in the bid document. Use text similarity technology to position the title description information in the bid document and the evaluation factor requirement information in the tender document, that is, obtain the module content areas such as qualification requirements, form requirements, business, technology, quotation, performance, financial statements, enterprise personnel and the text description content of the evaluation factor requirement information from the bid document text.

[0038] S520: Comparison of similarity of content of bid document evaluation factors. Use text similarity detection technology to analyze the text content in different modules of different bidders for similarity and obtain similarity results of their bidder document information.

[0039] S530: Detection of the regularity of bidder quotations. Use a regularity detection algorithm to perform regularity detection and analysis on the itemized quotations and total quotations in the quotations of different bidders. When a set of bidders with regular itemized quotations and total quotations appears, the document information similarity matrix of the bidder set is set to 1.

[0040] S540: Bidder document data pattern recognition, through S520, S530 to calculate the similarity results of all the bidder's document data content, the quotation regularity analysis to obtain the document data correlation result matrix between the bidders, and the maximum value of the document data correlation result between the bidders is recorded as the bidder's dynamic correlation result, and the bidder's document data pattern probability matrix is ​​obtained. Preferably, the step S600 specifically includes the following steps:

[0041] S610: Calculate the weights of different types of pattern recognition results by weight addition; use the bidders' static information pattern probability matrix, dynamic information pattern probability matrix, and document information pattern probability matrix to calculate the weights and obtain the basic probability matrix of the association pattern between bidders in the current bidding.

[0042] S620: Recall of abnormal pattern results between bidders. Under normal circumstances, when a bidding project releases the winning results, the companies participating in the bidding will enter the bidding system to check the winning bidders. However, after a bidding project releases the winning results, only one or two companies check the winning results and download the winning results. In this case, the bidding project has pattern recognition to a large extent. When pattern recognition occurs in the bidding section, other people with abnormal behavior patterns will not check the winning results and download the winning documents during the public announcement of the winning results. Only the main bidder will check the winning results and download the winning documents. Then, when this happens, it is only necessary to record the bidders who check and download the winning results during the public announcement of the winning results. When this happens, the basic probability matrix of the association pattern between bidders needs to be multiplied by a coefficient of 1.2 to obtain a new probability matrix of the association pattern between bidders.

[0043] S630: Identification of association patterns between bidders. Analyze the probability matrix results of association pattern identification between bidders. When the probability value of the probability matrix of association patterns between two bidders is higher than P, it indicates that there is a suspicion of abnormal pattern between the two bidders.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] 1. Improve regulatory efficiency: By analyzing the behavior patterns of bidders at different bidding stages, the present invention can quickly identify abnormal behaviors, thereby improving the monitoring efficiency of the regulatory agency over the bidding process and ensuring the compliance of bidding activities.

[0046] 2. Maintain market fairness: The present invention helps to expose and prevent unfair competition such as bid rigging and collusion, maintains the fairness of the bidding market, and enables bidders with real strength and integrity to compete in a fair environment.

[0047] 3. Reduce procurement risks: By identifying potential unfair behaviors, the present invention helps to reduce risks in the procurement process, avoid waste of resources and quality problems caused by unfair competition, and ensure the smooth progress of procurement projects.

[0048] 4. Promote technological progress and innovation: The application of this invention has promoted technological progress in the field of bidding and procurement, encouraged relevant enterprises and institutions to adopt advanced data analysis technology, and promoted the informatization construction and innovative development of the entire industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a flow chart of the method for identifying association patterns between bidders in the procurement process.

[0050] Figure 2 This is the detailed process of unified representation of bidder data in step S100 of the present invention.

[0051] Figure 3 This is the detailed process of the bidder data preprocessing in step S200 of the present invention.

[0052] Figure 4 This is the detailed process of the bidder static data preprocessing in step S210 of the present invention.

[0053] Figure 5 This is the detailed process of the bidder dynamic data preprocessing in step S220 of the present invention.

[0054] Figure 6 This is the detailed process of preprocessing bidder document data in step S230 of the present invention.

[0055] Figure 7 This is the detailed process of the bidder static data analysis in step S300 of the present invention.

[0056] Figure 8 This is the detailed process of the bidder dynamic data analysis in step S400 of the present invention.

[0057] Fig. 9 This is the detailed process of the bidder document data analysis in step S500 of the present invention.

[0058] Fig.10 This is the detailed process of identifying the association pattern between bidders in step S600 of the present invention. DETAILED DESCRIPTION

[0059] In order to better illustrate the technical solution of the present invention, the present invention is now further described in detail according to the accompanying drawings and embodiments.

[0060] Embodiment 1:

[0061] like Figure 1As shown, assuming that xxx procurement project is tendered, bidders A, B, and C participate in the preparation of the tender project and the overall bidding process.

[0062] The first step is the bidder's bidding data normalization module.

[0063] like Figure 2 As shown, by using the interface docking method, static information data, dynamic information data, document information data, etc. of different bidders participating in the same bidding package are collected from the bidding system.

[0064] S110: Unified representation of bidders’ static data. Static information data refers to the basic information of bidders saved in the bidding and procurement system. Static information data includes but is not limited to company name, corporate legal person, equity, bidder contact information, contact address, email address, and senior management; for example, the static information of bidder A is {Suzhou Pavis Information Technology Co., Ltd., Li Yiyi, ((Suzhou Material Information Technology Co., Ltd., 35%), (Kunshan Statistics Information Technology Co., Ltd., 65%)), Zhang*:19900003334, No. **, Zhuyuan Road, Suzhou High-tech Zone, huafang@163.com, (Li Xkang, Fang X)}, and the static information of bidder B is {Suzhou Fangtai Information Technology Co., Ltd., Li Yier, ((Suzhou Material Information Technology Co., Ltd., 5.2%), (Wuxi Physical Information Technology Co., Ltd. Information Technology Co., Ltd., 19.2%), (Jiangsu XX Information Technology Co., Ltd., 75.6%)), Sun*: 19900001234, No. **, Zhuyuan Road, Suzhou High-tech Zone, huating***@163.com, (Hu Xjun, Xie X)}, the static information of Bidder C is {Suzhou Hengsheng Information Technology Co., Ltd., Li**, ((Wuxi Tizhi ​​Information Technology Co., Ltd., 0.4%), (Wuxi XX Information Technology Co., Ltd., 90.6%)), Wang*: 199****4534, No. **, Zhuyuan Road, Suzhou High-tech Zone, huafang@163.com, (Li Xkang, Fang X, Wang XX)}.

[0065] S120: Unified representation of bidder dynamic data. Dynamic behavior data refers to data related to each bidding activity conducted by bidders in the bidding system, including but not limited to browsing and retrieval record data {the bidder's method of obtaining bidding projects, the bidder's system login IP address, the bidder's system login MAC address}; bidder registration data {the bidder's registration time, the bidder's registration IP address, the bidder's registration MAC address}; bidding document download record data {the time of downloading bidding documents, the IP address of downloading bidding documents, the MAC address of downloading bidding documents}; bidder upload bid record data {the bidder's bid submission time, the bidder's bid submission IP address, the bidder's bid submission MAC address}; bidder second quotation record data {the bidder's second quotation time, the bidder's second quotation IP address, the bidder's second quotation MAC address}; bidder viewing winning notice record data {the bidder's viewing time, the bidder's IP address, the bidder's MAC address}. For example, the dynamic information of bidder A is {{information retrieval, 2024-xx-1014:26:31,180.xxx.xxx.145,52:54:00:xx:xx:cf},{2024-xx-1014:27:21,180.xxx.xxx.145,52:54:00:xx:xx:cf},{2024-xx-15:14:27:31,180.xxx.xxx.145,52:54:00:xx:xx:cf},{2024-xx-15 11:27:31,180.xxx.xxx.145,52:54:00:xx:xx:cf},{2024-xx-1810:20:35,180.xxx.xxx.145,52:54:00:xx:xx:cf},{2024-xx-25

[0066] 13:23:01,180.xxx.xxx.145,52:54:00:xx:xx:cf}}, the dynamic information of bidder B is {{browse, 2024-xx-13 10:26:31,10.xxx.xxx.215,2e:08:xx:xx:a8:3a},{2024-xx-13 17:27:31,10.xxx.xxx.215,2e:08:xx:xx:a8:3a},{2024-xx-15

[0067] 16:27:41,10.xxx.xxx.215,2e:08:xx:xx:a8:3a},{2024-xx-15

[0068] 14:27:31,18.xxx.xxx.15,2e:08:xx:xx:a8:3a},{2024-xx-18

[0069] 11:03:05,18.xxx.xxx.15,2e:08:xx:xx:a8:3a},{}}, the dynamic information of bidder C is {{Information retrieval, 2024-xx-10 14:26:31,120.xxx.xxx.105,1e:15:xx:xx:5c:92},{2024-xx-1014:27:31,120.xxx.xxx.105,1e:15:xx:xx:5c:92},{2024-xx-15

[0070] 14:27:41,120.xxx.xxx.105,1e:15:xx:xx:5c:92},{2024-xx-15

[0071] 11:28:00,120.xxx.xxx.105,1e:15:xx:xx:5c:92},{2024-xx-18

[0072] 11:10:09,130.xxx.xxx.125,1e:15:xx:xx:5c:92},{}}.

[0073] S130: Unified representation of bidder document data. Document information data refers to the document file data prepared by bidders for bidding in the tendering and procurement system, including but not limited to {bidding documents, bidding documents, quotation documents, secondary quotation documents, technical clarification documents, and negotiation documents}. For example, the document information of bidder A is {xxx project bidding documents.docx, bidder A bidding documents.pdf, bidder A quotation documents.pdf, bidder A secondary quotation documents.pdf, bidder A technical clarification documents.pdf, bidder A negotiation documents.pdf}, the document information of bidder B is {xxx project bidding documents.docx, bidder B bidding documents.pdf, bidder B quotation documents.pdf, bidder B secondary quotation documents.pdf, bidder B technical clarification documents.pdf, bidder B negotiation documents.pdf}, and the document information of bidder C is {xxx project bidding documents.docx, bidder C bidding documents.pdf, bidder C quotation documents.pdf, bidder C secondary quotation documents.pdf, bidder C technical clarification documents.pdf, bidder C negotiation documents.pdf}, etc.

[0074] Step 2: Figure 3 As shown, the bidder data preprocessing module.

[0075] S210: Preprocessing of bidder static data. The fields in the static information data are divided into two categories, one is the unique field, and the other is the association field. Among them, the unique fields include but are not limited to {company name, corporate legal person, bidder contact information, contact address, email address}; the association field includes but is not limited to {equity, senior management}. In order to ensure the efficiency of equity-related companies, shareholders with less than 5% of the company's equity are eliminated. The equity field information of bidder C is ((Wuxi XX Information Technology Co., Ltd., 90.6%)), and the equity field information of bidder A is ((Suzhou Material Information Technology Co., Ltd., 35%), (Kunshan Statistical Information Technology Co., Ltd., 65%)); the equity field information of bidder B is ((Suzhou Material Information Technology Co., Ltd., 5.2%), (Wuxi Physical Information Technology Co., Ltd., 19.2%), (Jiangsu XX Information Technology Co., Ltd., 75.6%)).

[0076] S220: Preprocessing of bidder dynamic data. Perform basic data processing on bidder dynamic data; divide the fields in dynamic information data into three categories: one is processed from the time of bidder operation behavior, i.e. bidder dynamic data time interval; another is processed from the equipment hardware operated by bidder, i.e. bidder dynamic data hardware identification; and the other is processed from the operation behavior of bidder, i.e. bidder dynamic data operation behavior;

[0077] S221: Time interval processing of bidder dynamic data. Calculate the time interval between two adjacent operations for the time field in the bidder dynamic information data; for example, the time of each bidding activity performed by bidder A is {{browse search record, 2024-xx-10 14:26:31}, {bidder registration, 2024-xx-1014:27:21}, {bid document download, 2024-xx-15:14:27:31}, {upload bid, 2024-xx-1511:27:31}, {second quotation, 2024-xx-1810:20:35}, {check winning notice, 2024-xx-2513:23:01}}

[0078] S222: Extract the hardware identification of the bidder's dynamic data. The IP address and MAC address fields in the bidder's dynamic information data are recorded as the bidder's dynamic data hardware identification data; for example, the hardware identification data of bidder A is {(180.xxx.xxx.145), (52:54:00:xx:xx:cf)}; the hardware identification data of bidder B is {(18.xxx.xxx.15), (2e:08:xx:xx:a8:3a)}; the hardware identification data of bidder C is {(120.xxx.xxx.105, 130.xxx.xxx.125), (1e:15:xx:xx:5c:92)}.

[0079] S223: Extract the dynamic data operation behavior of the bidder, and record the behavior mode field in the bidder's dynamic information data as the bidder's dynamic data operation behavior data; for example, the dynamic data operation behavior data of bidder A is {(information retrieval, view the winning bidder)}; the hardware identification data of bidder B is {(browse and view)}; the hardware identification data of bidder C is {(information retrieval)}.

[0080] S230: Preprocessing of bidder document data: parsing the bidder document data to obtain basic information of the document; for example, performing OCR recognition on the bid document uploaded by bidder A, bidder A bid document.pdf, to obtain its text content. And obtain the title paragraph information and table text information in bidder A's bid document based on the OCR recognition content.

[0081] S231: Document information processing; perform information processing on the documents in the bidder's document data to obtain the text content of the document; for example, parse the bidding document, perform OCR recognition on the bidding document of bidder A to obtain its text content.

[0082] S232: Document structure information discovery: Use paragraph discovery and title structure extraction technology to perform structural analysis on the text content information in the document, and obtain the document structure information and document title description information in the document file. For example, the document title structure of the bidder A's bid document is """ Document - Quotation Letter

[0083] Document 2: Power of attorney, identity certificate of legal representative

[0084] File 3 Quotation Form

[0085] Document 4 Service Plan

[0086] Technical Solution

[0087] Service Team

[0088] About the Team ...

[0090] File Five Deviation Table

[0091] Document 6 Qualification Review Information

[0092] Business license

[0093] Qualification Certificate

[0094] Historical performance

[0095] Previous Years' Financial Reports

[0096] Document 7 Compliance Commitment Letter

[0097] Document 8 Other information

[0098] ...”””

[0100] S233: Extraction of evaluation factor standard requirements. Use table extraction technology and information extraction technology to obtain the evaluation factor requirement information in the bidding documents. The evaluation factor requirement information includes, but is not limited to, {bidder qualification requirements, bidder form requirements, commercial part scoring criteria, technical part scoring criteria, performance requirements, financial statements, quotation part scoring criteria, project manager qualification requirements, enterprise personnel requirements}. The extraction results of the evaluation factors for the xxx bidding project are as follows {[["Scoring Factor", "Scoring Standard"], ["Financial Status", "Evaluate the financial status of the bidder (based on the audited financial statements for 2022 and 2023). The scoring results are divided into five grades from A to E."], ["Previous Similar Project Performance and Experience", "Evaluate the historical performance provided by the bidder in combination with the requirements of this project: including, but not limited to, the matching degree, scale, implementation effect, quantity, etc. of the performance. The scoring results are divided into five grades from A to E."], ["Credit Evaluation", "Conduct unified scoring based on the latest released annual supplier credit evaluation results. The credit scores for grades A, B, and C are 100, 85, and 60 points respectively. If the supplier submits the initial quotation, the evaluation group shall reasonably determine the credit rating for this evaluation based on its previous performance and contract performance in other units."], ["Reasonableness of Quotation", "Conduct a reasonableness review of the quotation composition, comprehensive unit price, etc. of each bidder. The scoring results are divided into five grades from A to E."], ["Inspection of Confidential Content in Quotation Documents", "The bidder shall not submit or upload any quotation documents containing classified or misclassified content. If it is found during the inspection, this item will get 0 points; if not found during the inspection, this item will get 100 points."], ["Price Score", "When 0 < Di ≤ 3%, 1 point will be deducted for every 1% increase; when 3% < Di ≤ 6%, 1.5 points will be deducted for every 1% increase; when 6% < Di, 2 points will be deducted for every 1% increase; when -3% < Di ≤ 0, no points will be deducted; when -6% < Di ≤ -3%, 1 point will be deducted for every 1% decrease; when -9% < Di ≤ -6%, 1.5 points will be deducted for every 1% decrease; when Di ≤ -9%, 2 points will be deducted for every 1% decrease; the full score is 100 points, and the minimum score is 60 points."], ["Research Plan", "The responsiveness of the research plan to the procurement technical documents, and the reasonableness and feasibility of the plan design, research content, equipment investment, research mechanism, etc. The degree of understanding and grasp of the key points and difficulties of the project. The scoring results are divided into five grades from A to E."], ["Technical Service Team", "The title, professional ability and similar performance of the project leader, the composition of the project team, etc. The scoring results are divided into five grades from A to E."], ["Project Implementation Plan", "The reasonableness of the construction period arrangement, workload planning for each stage and progress control measures. The scoring results are divided into five grades from A to E."]], {"Qualification Requirements": "Having independent legal person status", "Performance Requirements": "Performance in the past 5 years", "Reputation Requirements": "Not within the professional scope and period of bidding restrictions", "Financial Requirements": "No consecutive losses in two years", "Project Manager": "PMP certificate"}}.

[0101] Step 3: Figure 4 As shown, the bidder static data analysis module.

[0102] S310: Unique field consistency comparison. Use the unique field in S210 for consistency comparison, and use the full-text comparison algorithm to compare the unique field in each bidder's static information to obtain the consistency result of the unique field. For example, the consistency result calculated for the bidder A's company name "Suzhou Pavis Information Technology Co., Ltd." and the bidder B's company name "Suzhou Fotile Information Technology Co., Ltd." is "0", and the consistency result calculated for the bidder A's email address "huafang@163.com" and the bidder C's email address "huafang@163.com" is "1".

[0103] S320: Equity association discovery. Use the equity field in the association field in S210 to discover the association, and use the equity-related enterprise comparison analysis algorithm to calculate the consistency result of the bidder's equity association for the equity field information in the association field in the static information. In order to ensure the efficiency of the comparison of equity-related enterprises, only shareholders with equity holdings higher than 5% of the company are calculated for consistency. For example, the consistency result calculated for bidder A's equity "(Suzhou Material Information Technology Co., Ltd., 35%)" and bidder B's company name "(Suzhou Material Information Technology Co., Ltd., 5.2%)" is "1"; the consistency result calculated for bidder B's equity "(Wuxi Physical Information Technology Co., Ltd., 5.2%)" and bidder C's company name "(Wuxi Physical Information Technology Co., Ltd., 0.4%)" is "0".

[0104] S330: Similarity analysis of senior management of enterprises; use the senior management field in the association field in S210 to calculate the similarity analysis of senior management of enterprises. For the senior management field information in the association field in the static information, use the personnel name similarity comparison analysis algorithm to calculate the similarity analysis results of senior management of bidder enterprises. When the similarity value of senior management is higher than Sim1, Sim1 is set to 0.85 in this patent, and the consistency results of the two enterprises are set to 1. For example, the senior management "(Li Xkang, Fang X)" of bidder A and the senior management "(Li Xkang, Fang X)" of bidder B can be calculated to have a similarity value of 1, so the consistency result of the senior management information of bidder A and bidder B is "1".

[0105] S340: Identification of bidder static data patterns. Calculate the consistency results of all static information of bidders to obtain the field consistency matrix between bidders, and record the maximum value of the static information consistency results between bidders as the static information similarity result of the bidders, and obtain the static information similarity matrix of the bidders. If all static information consistency results of bidder A and bidder B are (0,1,1), then the static information similarity result of bidder A and bidder B is 1; if all static information consistency results of bidder B and bidder C are (0,0,0), then the static information similarity result of bidder B and bidder C is 0. The bidder static data pattern probability matrix of bidders A, B, and C is shown below.

[0106]

[0107] Step 4: Figure 5 As shown, the bidder dynamic data analysis module.

[0108] S410: Time interval volatility analysis. The group network algorithm is used to calculate the correlation results of the operation time intervals (S221) between different stages. The time interval volatility analysis between bidders is performed to calculate the dynamic data pattern probability matrix. For example, through calculation, it can be concluded that the correlation result between bidder A and bidder C is 0.873.

[0109] S420: Hardware information consistency comparison. Use the full-text comparison algorithm to calculate the hardware identification data in the dynamic information of each bidder to obtain the consistency result. When the hardware identification data (S222) between two bidders are consistent, the dynamic data pattern probability matrix result between the two bidders is set to 1. For example, if the hardware identification data of bidder A {(180.xxx.xxx.145), (52:54:00:xx:xx:cf)} is inconsistent with the hardware identification data of bidder B {(18.xxx.xxx.15), (2e:08:xx:xx:a8:3a)}, then the dynamic data pattern probability between bidders A and B will not be modified.

[0110] S430: Analysis of bidder edge nodes; special processing of some operation behaviors in dynamic behavior data. This part of the operation behavior includes but is not limited to {browsing and retrieving record data}. Under normal circumstances, bidders browse the bidding projects on the tendering and procurement system normally. When there is a bidder in a bidding project that exceeds (includes) the ratio and obtains the bidding project by retrieving the name of the bidding project through information, then the bidders in the bidding project have abnormal patterns to a large extent, so it is necessary to perform weighted calculation of the correlation results of the bidders through information retrieval. The ratio of this patent is set to 80%. For example, if bidder B obtains the bidding project through information retrieval and 80% of the bidders of the bidding project obtain the bidding project through information retrieval, then the correlation results between all bidders need to be multiplied by 1.07. That is, the correlation results between bidder A and bidder C need to be weighted and adjusted to 0.938.

[0111] S440: Bidder dynamic data pattern recognition: The bidder dynamic data pattern probability matrix is ​​calculated by the above method.

[0112] The dynamic data pattern probability matrix of bidders A, B, and C is shown below.

[0113]

[0114] Step 5: Figure 6 As shown, the bidder document data analysis module.

[0115] S510: Positioning the description of evaluation factors in the bid document. Use text similarity technology to position the title description information in the bid document and the evaluation factor requirement information in the tender document, that is, obtain the module content areas such as qualification requirements, form requirements, business, technology, quotation, performance, financial statements, enterprise personnel and the text description content of the evaluation factor requirement information from the bid document text.

[0116] S520: Comparison of similarity of content of bid document evaluation factors. Use text similarity detection technology to analyze the similarity of text content in different modules of different bidders to obtain similarity results of their bidder document information. For example, the similarity result of document information between bidder A and bidder B is 0.338, and the similarity result of document information between bidder A and bidder C is 0.657.

[0117] S530: Detection of the regularity of bidder quotations. Use a regularity detection algorithm to perform regularity detection and analysis on the itemized quotations and total quotations in the quotation documents of different bidders. When a set of bidders with regular itemized quotations and total quotations appears, the document information similarity matrix of the bidder set is set to 1.

[0118] S540: Bidder document data pattern recognition: Calculate the bidder document data pattern probability matrix in different ways.

[0119] According to the above process, the probability matrix of the bidder document data pattern of bidders A, B, and C is:

[0120]

[0121] Step 6: Bidder association pattern recognition module

[0122] S610: Calculate the weights of different types of pattern recognition results. Figure 7-9 As shown, the bidders' static information pattern probability matrix, dynamic information pattern probability matrix, and document information pattern probability matrix are used for weighted calculation to obtain the probability matrix result of the association pattern between bidders in the current bidding.

[0123] S620: recall of abnormal pattern results between bidders; Under normal circumstances, when a bidding project releases the winning results, the enterprises participating in the bidding will enter the bidding system to check the winning bidders. However, after a bidding project releases the winning results, only one or two enterprises check the winning results and download the winning results. In this case, the bidding project has pattern recognition to a large extent. When pattern recognition occurs in the bidding section, other abnormal behavior pattern people will not check the winning results and download the winning documents during the public announcement of the winning results. Only the main bidder will check the winning results and download the winning documents. Then, when this happens, it is only necessary to record the bidders who check and download the winning results during the public announcement of the winning results. If only bidder A among bidders A, B, and C checks and downloads the public announcement of the winning results, then there is a high probability of pattern recognition among bidders A, B, and C, that is, the probability of bidders A, B, and C is weighted, and the coefficient of 1.2 is multiplied on the basic probability matrix of the association pattern between bidders. Obtain the latest probability matrix results of association pattern recognition between bidders.

[0124] S630: Identify the association pattern between bidders. Fig.10 As shown, the result of the probability matrix of the association pattern recognition between bidders is analyzed. When the probability value of the probability matrix of the association pattern between two bidders is higher than P, it indicates that there is a suspicion of an abnormal pattern between the two bidders. P is set to 0.85 in the present invention. For example, according to the result of the probability matrix of the association pattern recognition between bidders calculated in S620, it can be detected that the probability value of bidder A and bidder C is 0.865, which is greater than P. Therefore, there is an abnormal pattern between bidder A and bidder C.

[0125]

[0126] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying association patterns between bidders in the procurement process, characterized in that: The steps include: S100: Unified representation of bidder data: Use interface docking to collect static information data, dynamic information data, document information data, etc. of different bidders participating in the same bid package from the procurement system; S200: bidder data preprocessing; S300: Bidder static data analysis; S400: Dynamic data analysis of bidders; S500: bidder document data analysis; S600: Identification of association patterns between bidders.

2. The method for identifying association patterns between bidders in the procurement process according to claim 1 is characterized in that: The step S100 specifically includes the following steps: S110: unified representation of bidder static data, including enterprise name, corporate legal person, equity, bidder contact information, contact address, email address, and senior executives; S120: unified representation of bidders' dynamic data, including browsing and retrieval record data, bidder registration data, bidding document download record data, bidder upload record data, bidder second quotation record data, and bidder viewing winning notice record data; The browsing and retrieval record data includes the bidder's method of obtaining the bidding project, the bidder's login system IP address, and the bidder's login system MAC address; The bidder registration data includes the bidder registration time, the bidder registration IP address, and the bidder registration MAC address; The bidding document download record data includes the time of downloading the bidding document, the IP address of downloading the bidding document, and the MAC address of downloading the bidding document; The bidder's bid upload record data includes the bidder's bid submission time, the bidder's bid submission IP address, and the bidder's bid submission MAC address; The bidder's second quotation record data includes the bidder's second quotation time, the bidder's second quotation IP address, and the bidder's second quotation MAC address; The bidder's viewing of the winning notice record data includes the bidder's viewing time, the bidder's IP address, and the bidder's MAC address; S130: unified representation of the bidder's document data, the document data includes bidding documents, bidding documents, quotation documents, secondary quotation documents, technical clarification documents, and negotiation documents.

3. The method for identifying association patterns between bidders in the procurement process according to claim 1 is characterized in that: The step S200 specifically includes the following steps: S210: Preprocessing the bidder's static data, dividing the fields in the static data into two categories, one is a unique field, and the other is an associated field, wherein the unique field includes the company name, corporate legal person, bidder contact information, contact address, and email address; the associated field includes equity and senior management; S220: Preprocessing of bidder dynamic data, performing basic data processing on the bidder dynamic data; S230: Preprocessing the bidder's document data, parsing the bidder's document data to obtain basic information of the document.

4. The method for identifying association patterns between bidders in the procurement process according to claim 3 is characterized in that: The step S220 specifically includes the following steps: S221: Processing the time interval of the bidder's dynamic data, calculating the time interval between two adjacent operations for the time field in the bidder's dynamic data; S222: Extracting the bidder's dynamic data hardware identification, and recording the IP address and MAC address fields in the bidder's dynamic data as the bidder's dynamic data hardware identification data; S223: Extract the bidder's dynamic data operation behavior, and record the behavior mode field in the bidder's dynamic data as the bidder's dynamic data operation behavior data.

5. The method for identifying association patterns between bidders in the procurement process according to claim 3 is characterized in that: The step S230 specifically includes the following steps: S231: Document information processing, performing information processing on the documents in the bidder's document data to obtain the text content of the documents; S232: Document structure information discovery: structural analysis of the text content information in the document is performed through paragraph discovery and title structure extraction technology to obtain document structure information and document title description information in the document file; S233: Extraction of evaluation factor standard requirements, using table extraction technology and information extraction technology to obtain the evaluation factor requirement information in the bidding documents. The evaluation factor requirement information includes bidder qualification requirements, bidder form requirements, commercial part scoring standards, technical part scoring standards, performance requirements, financial statements, quotation part scoring standards, project manager qualification requirements, and enterprise personnel requirements.

6. The method for identifying association patterns between bidders in the procurement process according to claim 1 is characterized in that: The step S300 specifically includes the following steps: S310: Unique field consistency comparison, using the unique field in step S210 to perform consistency comparison, using the full text comparison algorithm to perform consistency comparison on the unique field in the static information of each bidder, and obtaining the consistency result of the unique field; S320: Equity association relationship discovery, using the equity field in the association field in step S210 to discover the association relationship, and using the equity field information in the association field in the static information to calculate the equity association consistency result of the bidder using the equity-related enterprise comparison analysis algorithm; S330: Enterprise senior management similarity analysis, using the senior management field in the association field in step S210 to calculate the enterprise senior management similarity analysis, for the senior management field information in the association field in the static information, using the personnel name similarity comparison analysis algorithm to calculate the similarity analysis results of the bidder's senior management personnel, when the senior management personnel similarity value is higher than Sim1, the consistency results of the two companies are set to 1; S340: Identify the bidder's static data pattern, calculate the consistency results of all the bidder's static information to obtain the field consistency matrix between the bidders, and record the maximum value of the static information consistency result between the bidders as the bidder's static information similarity result, and obtain the bidder's static information similarity matrix.

7. The method for identifying association patterns between bidders in the procurement process according to claim 1 is characterized in that: The step S400 specifically includes the following steps: S410: Time interval volatility analysis, using a swarm network algorithm to calculate the correlation results of the operation time intervals between different stages; S420: hardware information consistency comparison, using a full text comparison algorithm to calculate the hardware identification data in the dynamic information of each bidder to obtain a consistency result. When the hardware identification data of the bidders are consistent, the correlation result between the bidders is set to 1; S430: Bidder edge node analysis, special processing of some operation behaviors in dynamic behavior data, including browsing and retrieving record data, and weighted calculation of relevance results for bidders retrieved through information. Under normal circumstances, bidders browse tender projects in the tendering and procurement system normally. When bidders exceeding the ratio in a tender project obtain the tender project by retrieving the tender project name through information, the bidders in the tender project have an abnormal pattern to a large extent. S440: Dynamic data pattern recognition of bidders. Through S410, S420 and S430, the correlation results of all dynamic information of bidders are calculated to obtain the correlation result matrix of different fields between bidders, and the maximum value of the dynamic information correlation result between bidders is recorded as the dynamic correlation result of the bidder, thus obtaining the dynamic pattern probability matrix of the bidder.

8. The method for identifying association patterns between bidders in the procurement process according to claim 1 is characterized in that: The step S500 specifically includes the following steps: S510: Positioning the description of evaluation factors in the bid document, using text similarity technology to position the title description information in the bid document and the evaluation factor requirement information in the tender document, including obtaining the qualification requirements, form requirements, business, technology, quotation, performance, financial statements, enterprise personnel module content areas and evaluation factor requirement information text description content from the bid document text; S520: Comparison of similarity of content of bid document evaluation factors, using text similarity detection technology to analyze the similarity of text content in different modules of different bidders to obtain similarity results of their bidder document information; S530: Detection of the regularity of bidder quotations, using a regularity detection algorithm to perform regularity detection and analysis on the itemized quotations and the total quotations in the quotations of different bidders. When a set of bidders with regular itemized quotations and total quotations appears, the document information similarity matrix of the bidder set is set to 1. S540: Identify the bidder's document data pattern. Calculate the content similarity results of all the bidder's document data through S520 and S530, and analyze the quotation regularity to obtain the document data correlation result matrix between the bidders. The maximum value of the document data correlation result between the bidders is recorded as the bidder's dynamic correlation result, and the bidder's document data pattern probability matrix is ​​obtained.

9. The method for identifying association patterns between bidders in the procurement process according to claim 1, characterized in that: The step S600 specifically includes the following steps: S610: weighting and calculating the weights of different types of pattern recognition results; using the bidders' static data pattern probability matrix, dynamic data pattern probability matrix, and document data pattern probability matrix to calculate the weights, and obtain the basic probability matrix of the association pattern between bidders in the current bidding; S620: recall of abnormal pattern results between bidders. When pattern recognition occurs in a bidding section, the basic probability matrix of the correlation pattern between bidders needs to be multiplied by a coefficient of 1.2 to obtain a new probability matrix of the correlation pattern between bidders. S630: Identify the correlation pattern between bidders, analyze the probability matrix results of the correlation pattern identification between bidders, when the probability value of the correlation pattern probability matrix between two bidders is higher than P, it indicates that there is a suspicion of abnormal pattern between the two bidders.

10. The method for identifying association patterns between bidders in a procurement process according to claim 6 or 9, characterized in that: The Sim1 is set to 0.85, the ratio is set to 80%, and the P is set to 0.85.

Images