Method for judging procurement risk based on relationship tree model and storage medium

Abstract

The application discloses a procurement risk discrimination method based on a relationship tree model and a storage medium. The method comprises the following steps: establishing a relationship tree model, wherein the relationship tree model comprises a weight value relationship between a procurement party and a first procurement project and a weight value relationship between a supplier and the first procurement project; the weight value relationship between the supplier and the first procurement project is determined according to the procurement content similarity of a procurement project in which the first supplier has participated before the first procurement project; and risk procurement discrimination, relationship procurement discrimination, high-price procurement discrimination or subcontracting discrimination are performed according to the relationship tree model. The above scheme can construct a weight value tree model for the procurement party, the contractor and the procurement project, and simultaneously perform risk discrimination on the related procurement project, so that timely risk prompt can be given.

Description

Technical Field

[0001] This invention relates to the field of data analysis, and in particular to a method and storage medium for assessing procurement risks. Background Technology

[0002] The quality of goods or services procured by enterprises is closely related to their production and operation. The selection of suppliers and the signing of procurement contracts largely determine the survival and sustainable development of enterprises. Therefore, all enterprises have established strict internal control systems for procurement, strengthened supplier access conditions, and improved supplier quality. However, various irregularities still exist in enterprise bidding processes. How to better detect these irregularities has become a key focus and challenge for many enterprises in risk management. Currently, the main methods used are to formulate sound systems and processes, competitive bidding and negotiation, public announcement of winning bids, and contract auditing. However, the intelligent risk assessment aspect is relatively weak and urgently needs to be strengthened. Summary of the Invention

[0003] Therefore, there is a need to provide a risk assessment method that can meet the risk assessment requirements of existing technologies for relevant data in procurement projects.

[0004] To achieve the above objectives, the inventors provide a procurement risk assessment method based on a relationship tree model, comprising the following steps: establishing a relationship tree model, wherein the relationship tree model includes the weight relationships between the procuring entity and the first procurement project, and between the supplier and the procurement project; the weight relationship between the supplier and the first procurement project is determined based on the similarity of procurement content of procurement projects in which the first supplier participated prior to the time of the first procurement project.

[0005] The relationship tree model is used to determine risk-based procurement, relationship-based procurement, high-price procurement, or subcontracting.

[0006] Specifically, the risk procurement assessment includes the following steps: when the weighting relationship between the second procurement project and the first supplier is lower than the first threshold, and the first supplier's historical winning bids in all procurement projects with the same procuring party as the second procurement project are greater than the fifth threshold, it is assessed as a risk procurement.

[0007] Specifically, the process of determining related-party procurement includes the following steps: when the weighted relationship between the supplier and the second procurement project is less than a second threshold, it is determined to be related-party procurement.

[0008] Furthermore, the high-price procurement identification process specifically includes the following steps: calculating a weighted average reference price based on the winning bid price and weighting relationship between the supplier and each procurement project; generating a third threshold based on the reference price; and identifying a high-price procurement when the winning bid price of the supplier and the third procurement project is higher than the third threshold.

[0009] Furthermore, the specific steps involved in subcontracting identification are as follows:

[0010] The similarity of the content of new procurement projects initiated by the supplier within a preset time period after winning the fourth procurement project to the content of the fourth procurement project is determined.

[0011] If there are more than a preset number of new procurement projects with a procurement content similarity greater than the fourth threshold, they are identified as subcontracting risk procurement.

[0012] A procurement risk assessment storage medium based on a relationship tree model stores a computer program. When run, the computer program executes the following steps: establishing a relationship tree model, which includes the weight relationships between the procuring entity and the first procurement project, and between the supplier and the procurement project. The weight relationship between the supplier and the first procurement project is determined based on the similarity of procurement content of procurement projects in which the first supplier participated prior to the time of the first procurement project.

[0013] The relationship tree model is used to determine risk-based procurement, relationship-based procurement, high-price procurement, or subcontracting.

[0014] Specifically, the computer program performs risk procurement judgment steps, and when the weight relationship between the second procurement project and the first supplier is lower than the first threshold, and the number of historical winning bids of the first supplier in all procurement projects with the same purchaser as the second procurement project is greater than the fifth threshold, it is judged as risk procurement.

[0015] Specifically, the computer program performs a specific step of determining relationship-based procurement. When the weight relationship between the supplier and the second procurement project is less than a second threshold, it is determined to be relationship-based procurement.

[0016] Specifically, the computer program performs the following steps to determine high-price procurement: it calculates a weighted average reference price based on the winning bid price and weighting relationship between the supplier and each procurement project; it generates a third threshold based on the reference price; and it determines high-price procurement when the winning bid price of the supplier and the third procurement project is higher than the third threshold.

[0017] Furthermore, the computer program performs specific steps for packet segmentation and discrimination.

[0018] The similarity of the content of new procurement projects initiated by the supplier within a preset time period after winning the fourth procurement project to the content of the fourth procurement project is determined.

[0019] If there are more than a preset number of new procurement projects with a procurement content similarity greater than the fourth threshold, they are identified as subcontracting risk procurement.

[0020] Unlike existing technologies, the above solution can construct a weighted tree model for the purchaser, contractor, and procurement project, while also assessing the risks of the procurement project and providing timely risk warnings. Attached Figure Description

[0021] Figure 1 The flowchart of the procurement risk assessment method based on the relationship tree model is shown in the specific implementation method.

[0022] Figure 2 This is a schematic diagram of the procurement risk assessment storage medium described in the specific implementation method. Detailed Implementation

[0023] To explain in detail the technical content, structural features, objectives, and effects of the technical solution, the following description is provided in conjunction with specific embodiments and accompanying drawings.

[0024] In practice, the applicant noted that "transparent procurement" refers to enterprises and institutions acquiring products or services from the supply market as their own resources according to the principles of "openness, fairness, and impartiality" and "quality priority, price priority." Currently, national and local government agencies and enterprises have also established transparent procurement platforms, using information technology to further rationalize procurement and supervision systems, effectively promoting the openness and transparency of transparent procurement. Through reasonable competitive bidding and negotiation, procurement costs are effectively reduced, procurement efficiency is improved, and corruption such as under-the-table deals and kickbacks are avoided. At the same time, this also provides a research topic for better identifying irregularities in the procurement process. This invention, in this context, constructs a diversified weighted tree model based on procurement big data, and uses this model to identify several common irregularities in the procurement process, such as: relationship-based procurement, preference-based procurement, designated supplier procurement, high-price procurement, and subcontracting.

[0025] Therefore, in such Figure 1 The illustrated embodiment provides a procurement risk assessment method based on a relationship tree model, comprising the following steps: S1, establishing a relationship tree model, wherein the relationship tree model includes the weight relationships between the procuring entity and the first procurement project, and between the supplier and the procurement project. The weight relationship between the supplier and the first procurement project is determined based on the weight relationships of all procurement projects in which the first supplier participated prior to the time of the first procurement project.

[0026] S2 performs risk-based procurement judgment, relationship-based procurement judgment, high-price procurement judgment, or subcontracting judgment based on the aforementioned relationship tree model. Through this scheme, a weighted tree model can be constructed for the procuring entity, contractor, and procurement project, while simultaneously assessing the risks associated with the procurement project and providing timely risk alerts.

[0027] The relationship tree model includes a weight G(A, P) between procurement item i and supplier A. i Z i ), where A is the supplier matching degree, P i Z represents the bid price for this tender. i This indicates whether the bid was successful, with a value of 0 or 1, where 1 indicates success. Supplier matching degree can be calculated by summing, averaging, or weighted averaging the similarity between the supplier's past procurement activities and the current procurement activity. For example, the similarity between the supplier's past procurement activities and the current procurement activity can be weighted according to their proximity to the current time, with more recent activities receiving higher weights.

[0028] The relationship tree model also includes the weighted relationship between procurement item i and procurement item j, which is the similarity S(V i V j The similarity of projects can be calculated using techniques similar to project text similarity, and existing technologies can be utilized; this invention does not impose any limitations on this. In other embodiments, a procurement content similarity algorithm is used to calculate the procurement content similarity, which serves as the weighting relationship between procurement projects i and j.

[0029] The calculation method for the similarity of procurement content is as follows: For procurement projects i and j, retrieve the procurement bidding document data and remove the templated content of the bidding documents, such as: bidding announcement, bidding instructions, contract terms, bid document format requirements, evaluation methods, etc.

[0030] In the tender documents, the procurement content is processed by using regular expressions to identify the title of each sentence, forming a title list T. l The title sentence is then segmented and tagged with parts of speech to form a noun vector V. tn and verb vector V tv Other word vectors V to The dimension of the vector is the union of the non-repeating words, and the weight of each dimension is the word frequency.

[0031] Each sentence in the procurement section of the tender document is segmented and tagged with parts of speech to form a noun vector V. cn and verb vector V cv Other word vectors V co The dimension of the vector is the union of the non-repeating words, and the weight of each dimension is the word frequency.

[0032] Stop words and commonly used terms in the business domain, such as terms frequently used in software information technology construction projects: high cohesion, low coupling, microservices, etc. Common terms in the business domain can be manually annotated.

[0033] Improve the noun vector V formed by word segmentation of the title sentence. tnand verb vector V tv Other word vectors V to Each dimension has a weight, which is the word frequency multiplied by a weighting coefficient Δt (Δt>1, the title is more important than the body text).

[0034] The word vectors formed by merging the title and other content in the body are summed, and the weight values ​​of each dimension for the same word are added together to obtain:

[0035] Noun vector: Vn = V tn ∪V cn

[0036] Verb vector: Vt = V tv ∪V cv

[0037] Other word vectors: Vo = V to ∪V co

[0038] Improve noun word vector V n and verb word vector V t The weight of each dimension, i.e., the noun word vector V. n The weight of each dimension is the word frequency multiplied by Δn, and the verb word vector V t The weight of each dimension is the word frequency multiplied by Δv (Δn>1, Δv>1, the core content of procurement is mainly nouns and verbs, so the weight of nouns and verbs is increased).

[0039] Merge noun word vectors V n Verb vector V t Together with other word vectors Vo, they form the final procurement tender document word vector V.

[0040] The tender documents for the two procurements that need to be compared are processed according to the above steps to obtain the word vectors for the two procurements, denoted as: V i and V j Calculate the cosine similarity of word vectors or simple shared words as the similarity C(i,j). Finally, obtain the similarity of the procurement content.

[0041] For example, in some specific embodiments, the supplier matching degree can be calculated as follows:

[0042] 1) Initialize the supplier procurement content similarity SC to 0.

[0043] 2) Iterate through each bid in which the supplier participated and calculate the similarity of the content of each procurement, denoted as Ci. At the same time, calculate the time correction factor Ti for each procurement.

[0044] 3) Accumulate the similarity of supplier procurement content, SC=SC+Ci*Ti.

[0045] 4) Final cumulative value SC: Supplier procurement content similarity.

[0046] The time correction factor is positively correlated with the time of each bid; the closer a bid is to the present time, the larger the value of the time correction factor. Considering that bid content, supplier qualifications, etc., change over time, and their reference value in the bidding process weakens over time, a time correction factor T is introduced for entities with time-related factors. In some embodiments, the time correction factor T is defined as follows:

[0047] 1) Define the base time as 00:00:00 on January 1, 1970.

[0048] 2) Calculate the time difference ΔTa between the bidding time of historical bid A and the benchmark time, in seconds.

[0049] 3) Calculate the time difference ΔT between the current bidding time and the benchmark time, in seconds.

[0050] 4) Then T = ΔTa / ΔT.

[0051] The base time can be set as needed. In some other specific embodiments,

[0052] The time correction factor Ti at the i-th bid satisfies:

[0053] Ti=(ti-t0) / (tc-t0)

[0054] ti is the time of the i-th bid, tc is the time of this bid, and t0 is the base time constant.

[0055] By storing the supplier matching scores in a weighted tree, the weighted relationship between suppliers and specific procurement projects can be obtained, thus meeting the needs for risk assessment.

[0056] In some specific embodiments, risk procurement identification includes the following steps: when the weighting relationship between the second procurement project and the first supplier is lower than a first threshold, and the first supplier's historical winning bid count in all procurement projects with the same procuring entity as the second procurement project is greater than a fifth threshold, it is identified as risk procurement. The weighting relationship can use only the supplier matching degree index within the weighting relationship. When procuring entity A publishes the second procurement project and supplier B submits a bid, the data is entered into the system, and risk procurement identification is performed. If supplier B's supplier matching degree with the second procurement project is lower than the first threshold (e.g., the first threshold is 1), and supplier B's winning bid count in all procurement projects published by procuring entity A is greater than the second threshold (e.g., 8 times), a risk procurement warning is triggered, and it is identified as risk procurement. By setting up the above method, the risk procurement relationship types existing in the model can be effectively identified, preventing unqualified suppliers from winning bids multiple times. After being identified as risk procurement, a system alarm can be triggered, notifying manual intervention for review, thus improving the risk identification capability of the solution.

[0057] In other specific embodiments, the identification of related-party procurement includes the following steps: when the weighted relationship between the supplier and the second procurement project is less than a second threshold, it is identified as related-party procurement. When the weighted relationship between the supplier and the second procurement project, especially the supplier matching degree, is less than the second threshold (e.g., less than 0.5), it indicates that the supplier's qualifications or professional experience are likely insufficient, suggesting that there may be a situation where the bid is won through connections. After being identified as related-party procurement, a system alarm can be triggered, and manual intervention can be required for review. The above method improves the ability to identify related-party procurement.

[0058] In some further embodiments, the high-price procurement determination specifically includes the steps of calculating a weighted average reference price based on the winning bid prices and weighting relationships between the supplier and each procurement project; generating a third threshold based on the reference price; and determining that a procurement is high-price when the winning bid price of the supplier and the third procurement project is higher than the third threshold. In some embodiments,

[0059] Reference price guarantee = ∑ * (winning bid price * weighting relationship) / number of winning bids.

[0060] The weighting relationship can be based solely on the supplier matching degree dimension, with the reference price as the threshold. When the winning bid price of a supplier's new procurement project is higher than the reference price, a high-price procurement warning is triggered, allowing for manual review. This approach also improves the system's ability to identify high-price procurement.

[0061] In other specific embodiments, the packet segmentation detection process includes the following steps:

[0062] The system retrieves the content similarity between new procurement projects initiated by the supplier within a preset timeframe after winning the fourth procurement project and the fourth procurement project. If the content similarity of the procurement projects exceeds a preset number and is greater than a fourth threshold, it is identified as subcontracting risk procurement. In the weighted tree model, a supplier can be a supplier for one procurement project or a buyer for another. The same company can be identified by its name. If a company initiates multiple procurement projects for bidding within a preset timeframe (e.g., 3 months) after winning a procurement project, it is highly likely that it is subcontracting the winning project. A procurement content similarity index is introduced to compare the company's winning projects with the bidding projects. If the number of bidding projects exceeding the fourth threshold (e.g., 0.8) exceeds 3, it is identified as subcontracting risk procurement, triggering a subcontracting procurement warning. Manual intervention is allowed for review. This approach also improves the system's ability to identify subcontracting procurement.

[0063] In some specific implementations, the procurement data of a central state-owned enterprise, involving more than 1,000 bids and 200 suppliers, was used for model training and analysis.

[0064] Without model optimization, the similarity was less than 0.2261101603 and the number of successful bids (procurement out-degree / supply in-degree) was greater than 7. The system initially determined this to be a suspected relationship procurement behavior, pending manual verification.

[0065] If the similarity is less than 0.3128226399 and the number of successful bids (procurement out-degree / supply in-degree) is greater than or equal to 8, the system initially judges it as suspected biased procurement behavior, pending manual verification.

[0066] The number of successful bids (procurement output / supply input) is 12. The system initially judges it as a suspected designated procurement, pending manual verification.

[0067] The matching threshold for high-priced procurement is 0.7845130842. When the calculated result is greater than this threshold, the system initially judges it as a suspected high-priced procurement behavior, pending manual verification.

[0068] In such Figure 2 In the illustrated embodiment, this solution also provides a procurement risk assessment storage medium 2 based on a relationship tree model, which stores a computer program. When the computer program is run, it executes the following steps: establishing a relationship tree model, which includes the weight relationships between the procuring entity and the first procurement project, and between the supplier and the procurement project. The weight relationship between the supplier and the first procurement project is determined based on the weight relationships of all procurement projects in which the first supplier participated prior to the time of the first procurement project.

[0069] The relationship tree model is used to determine risk-based procurement, relationship-based procurement, high-price procurement, or subcontracting.

[0070] Specifically, the computer program performs risk procurement judgment steps, and when the weight relationship between the second procurement project and the first supplier is lower than the first threshold, and the number of historical winning bids of the first supplier in all procurement projects with the same purchaser as the second procurement project is greater than the fifth threshold, it is judged as risk procurement.

[0071] Specifically, the computer program performs a specific step of relationship-based procurement judgment. When the weight relationship between the supplier and the second procurement project is less than a second threshold, it is judged as a risky procurement.

[0072] Specifically, the computer program performs the following steps to determine high-price procurement: it calculates a weighted average reference price based on the winning bid price and weighting relationship between the supplier and each procurement project; it generates a third threshold based on the reference price; and when the winning bid price of the supplier and the third procurement project is higher than the third threshold, it is determined to be a risky procurement.

[0073] Furthermore, the computer program performs specific steps for packet segmentation and discrimination.

[0074] The similarity of the content of new procurement projects initiated by the supplier within a preset time period after winning the fourth procurement project to the content of the fourth procurement project is determined.

[0075] If there are more than a preset number of new procurement projects with a procurement content similarity greater than the fourth threshold, they are identified as subcontracting risk procurement.

[0076] It should be noted that although the above embodiments have been described herein, this does not limit the scope of patent protection of the present invention. Therefore, any changes and modifications made to the embodiments described herein based on the innovative concept of the present invention, or equivalent structural or procedural transformations made using the content of the present invention's specification and drawings, directly or indirectly applying the above technical solutions to other related technical fields, are all included within the scope of patent protection of the present invention.

Claims

1. A procurement risk assessment method based on a relational tree model, characterized in that, The process includes the following steps: establishing a relationship tree model, which includes the publication relationship between the procuring entity and the first procurement project, and the weighting relationship between the supplier and the procurement project. The weighting relationship between the supplier and the first procurement project is determined based on the similarity of the procurement content of procurement projects in which the first supplier participated before the time of the first procurement project. Based on the aforementioned relationship tree model, risk procurement judgment, relationship procurement judgment, high-price procurement judgment, or subcontracting judgment are performed; The relationship tree model includes a weight G(A, P) between procurement item i and supplier A. i Z i ), where A is the supplier matching degree, P i Z represents the bid price for this tender. i This indicates whether the bid was successful; the value can be 0 or 1, with 1 indicating success. Supplier matching score is calculated by weighted average of the similarity between the supplier's past procurement activities and the current procurement activity, including: The similarity between the supplier's past procurement activities and the current procurement activity is weighted according to the time elapsed since the current time, with more recent activities receiving higher weights: 1) Initialize the supplier procurement content similarity SC to 0; 2) Iterate through each bid in which the supplier has participated and calculate the similarity of the content of each procurement, denoted as Ci. At the same time, calculate the time correction factor Ti for each procurement. 3) Accumulate the similarity of supplier purchase content, SC = SC + Ci * Ti; 4) Final cumulative value SC: Supplier procurement content similarity; The time correction factor T is defined as follows: 1) Define the base time as 00:00:00 on January 1, 1970; 2) Calculate the time difference ΔTa between the bidding time of historical bid A and the benchmark time, in seconds; 3) Calculate the time difference ΔT between the current bidding time and the benchmark time, in seconds; 4) Then T = ΔTa / ΔT; The time correction factor Ti at the i-th bid satisfies: Ti = (ti - t0) / (tc - t0) ti is the time of the i-th bid, tc is the time of this bid, and t0 is the base time constant.

2. The procurement risk assessment method based on a relational tree model according to claim 1, characterized in that, The risk procurement assessment includes the following steps: when the weighting relationship between the second procurement project and the first supplier is lower than the first threshold, and the number of historical winning bids of the first supplier in all procurement projects with the same procuring party as the second procurement project is greater than the fifth threshold, it is assessed as a risk procurement.

3. The procurement risk assessment method based on a relational tree model according to claim 1, characterized in that, The specific steps for determining related-party procurement include: when the weighting relationship between the supplier and the second procurement project is less than a second threshold, it is determined to be related-party procurement.

4. The procurement risk assessment method based on a relational tree model according to claim 1, characterized in that, The process of identifying high-price procurement includes the following steps: calculating a weighted average reference price based on the winning bid price and weighting relationship between the supplier and each procurement project; generating a third threshold based on the reference price; and identifying high-price procurement when the winning bid price of the supplier and the third procurement project is higher than the third threshold.

5. The procurement risk assessment method based on a relational tree model according to claim 1, characterized in that, The specific steps involved in subcontracting identification are as follows: The similarity of the content of new procurement projects initiated by the supplier within a preset time period after winning the fourth procurement project to the content of the fourth procurement project is determined. If there are more than a preset number of new procurement projects with a procurement content similarity greater than the fourth threshold, they are identified as subcontracting risk procurement.

6. A storage medium for procurement risk assessment based on a relational tree model, characterized in that, The system stores a computer program that, when executed, includes the following steps: establishing a relationship tree model, which includes the weighted relationships between the procuring entity and the first procurement project, and between the supplier and the procurement project. The weighted relationship between the supplier and the first procurement project is determined based on the similarity of the procurement content of procurement projects in which the first supplier participated prior to the time of the first procurement project. Based on the aforementioned relationship tree model, risk procurement judgment, relationship procurement judgment, high-price procurement judgment, or subcontracting judgment are performed; The weight between the procurement project i and the supplier A in the relationship tree model is G(A, P i , Z i ), wherein A is a supplier matching degree, P i represents a bid price of this bid, Z i represents whether to win the bid, and takes a value of 0 or 1, 1 indicating winning the bid; Supplier matching score is calculated by weighted average of the similarity between the supplier's past procurement activities and the current procurement activity, including: The similarity between the supplier's past procurement activities and the current procurement activity is weighted according to the time elapsed since the current time, with more recent activities receiving higher weights: 1) Initialize the supplier procurement content similarity SC to 0; 2) Iterate through each bid in which the supplier has participated and calculate the similarity of the content of each procurement, denoted as Ci. At the same time, calculate the time correction factor Ti for each procurement. 3) Accumulate the similarity of supplier purchase content, SC = SC + Ci * Ti; 4) Final cumulative value SC: Supplier procurement content similarity; The time correction factor T is defined as follows: 1) Define the base time as 00:00:00 on January 1, 1970; 2) Calculate the time difference ΔTa between the bidding time of historical bid A and the benchmark time, in seconds; 3) Calculate the time difference ΔT between the current bidding time and the benchmark time, in seconds; 4) Then T = ΔTa / ΔT; The time correction factor Ti at the i-th bid satisfies: Ti = (ti - t0) / (tc - t0) ti is the time of the i-th bid, tc is the time of this bid, and t0 is the base time constant.

7. The procurement risk discrimination storage medium based on a relational tree model according to claim 6, characterized in that, The computer program performs risk procurement judgment in the following specific steps: when the weight relationship between the second procurement project and the first supplier is lower than the first threshold, and the number of historical winning bids of the first supplier in all procurement projects with the same procuring party as the second procurement project is greater than the fifth threshold, it is judged as risk procurement.

8. The procurement risk discrimination storage medium based on the relational tree model according to claim 6, characterized in that, The computer program performs specific steps for determining relationship-based procurement. When the weighted relationship between the supplier and the second procurement project is less than a second threshold, it is determined to be relationship-based procurement.

9. The procurement risk discrimination storage medium based on a relational tree model according to claim 6, characterized in that, The computer program performs the following specific steps to identify high-price procurement: it calculates a weighted average reference price based on the winning bid price and weighting relationship between the supplier and each procurement project; it generates a third threshold based on the reference price; and it identifies high-price procurement when the winning bid price of the supplier and the third procurement project is higher than the third threshold.

10. The procurement risk discrimination storage medium based on a relational tree model according to claim 6, characterized in that, The computer program performs specific steps for packet segmentation and identification. The similarity of the content of new procurement projects initiated by the supplier within a preset time period after winning the fourth procurement project to the content of the fourth procurement project is determined. If there are more than a preset number of new procurement projects with a procurement content similarity greater than the fourth threshold, they are identified as subcontracting risk procurement.

Images