Electronic bid invitation management system integrating bidding behavior intelligent identification

Abstract

The invention provides an electronic bid invitation management system integrated with bid behavior intelligent identification, which comprises a user display layer, a business logic layer and a data layer, and is characterized in that the business logic layer comprises a bid invitation management module, a bid invitation management module, a CA authentication module, a bid opening and evaluation module, a crawler module, a bid anomaly identification module and a notification module; in addition, through the bidding abnormity identification module, an intelligent identification method combining a supervised learning model and bidder relevance judgment is provided, whether bidders have price combination and bidding document defect setting on purpose to cause low bid evaluation score or waste bids is comprehensively considered, and the bidding success rate is improved. According to the method and the system, whether the bidders have combined bidding behaviors or not is identified in multiple aspects according to the multi-dimensional correlation between the bidders, so that the fairness of bidding and tendering is ensured, and a good market order is maintained.

Description

Technical Field

[0001] The present application relates to the field of electronic bidding technology, and in particular to an electronic bidding management system that integrates intelligent identification of bidding behavior. Background Art

[0002] Engineering bidding is a standardized transaction activity that applies the interaction of technical and economic methods and market competition mechanisms and selects the best through legal procedures. At present, the bidding process usually includes multiple links, such as the release of bidding announcements, submission of bidding documents, and bid evaluation. These links often rely on multiple independent systems or platforms. This decentralized system architecture leads to cumbersome processes and low efficiency, increasing the complexity of operations. Secondly, the problem of information islands is serious, and data sharing and circulation are not smooth, which increases management costs and time.

[0003] In addition, there are unfair competition behaviors such as joint bidding in the bidding process. Such behaviors deliberately restrict market competition through private collusion between bidders, exclude compliant bidders, and seek unfair benefits for specific stakeholders, which seriously disrupts the market order, damages the legitimate rights and interests of other participants, and also poses a potential threat to the quality and benefits of the project.

[0004] However, the existing bidding system cannot efficiently and accurately identify the joint bidding behavior in the bidding process. Therefore, there is an urgent need for an integrated electronic bidding and tendering electronic management system that integrates intelligent identification of bidding behavior. Summary of the invention

[0005] The present application provides an electronic bidding management system that integrates intelligent identification of bidding behavior, which can achieve integrated management of the entire bidding process and intelligent identification of joint bidding behavior.

[0006] The present application provides an electronic bidding management system integrating intelligent identification of bidding behavior, comprising: a user presentation layer, a business logic layer and a data layer, wherein the data layer is communicatively connected to the business logic layer, and the business logic layer is communicatively connected to the user presentation layer;

[0007] The user presentation layer is configured to display data to users and accept user input based on terminal devices, thereby realizing interaction between users and the system; the business logic layer is configured as the part that provides core services in the system architecture, and is used to provide users with a variety of business functions for electronic bidding and tendering management; the data layer is used for data storage, retrieval and management.

[0008] The business logic layer includes the bidding management module, tender management module, CA authentication module, bid opening and evaluation module, crawler module, bidding anomaly identification module and notification module;

[0009] The bidding management module is used for the bidder to upload or enter bidding data, and the bidding data at least includes the bidder information and bidding documents;

[0010] The bidding management module is used for bidders to download bidding documents, upload or enter qualification materials and bidding data, and the bidding data at least includes bidder information and bidding documents;

[0011] The CA authentication module is used to authenticate the identity and encrypt and decrypt the bidding documents and tender documents during the project bidding process.

[0012] The bid opening and evaluation module is used to decrypt the bid documents to obtain the bid structured data, and provide the evaluation committee members with online veto and online scoring functions to generate an evaluation report. The evaluation report at least includes the winning bid, the bidders who have been rejected, and the evaluation scores of each bidder.

[0013] The crawler module is used to obtain the basic enterprise information of the bidder from the Internet according to the bidder information in response to the instruction of the bid anomaly identification module, and perform data preprocessing.

[0014] The bidding anomaly identification module is used to obtain bidding information, evaluation report information, basic corporate information of bidders and historical joint bidding information, and determine whether bidders are suspected of joint bidding through intelligent identification methods, and generate automatic evaluation results for anomaly identification.

[0015] The notification module is used to push the abnormal identification automatic evaluation results determined by the bidding abnormality identification module, and the abnormal identification automatic evaluation results include the notification information of "no target joint bid collusion suspicion found", or the notification information of "target joint bid collusion suspicion exists", and the abnormal evaluation report sent together with the notification information of "target joint bid collusion suspicion exists", so as to facilitate the manual review of the bidding staff. The abnormal evaluation report includes the enterprise information of the bidder suspected of the first joint bid collusion, the bid quotation, the bid evaluation score, the bidder who cancelled the bid, and the correlation information between the bidders.

[0016] Further optionally, the notification module is also used to publish a winning bid announcement and a winning bid notice, or to publish a notice that the bidding is invalid due to illegal behavior.

[0017] As an optional embodiment of the present application, the bidding anomaly identification module includes a first identification module and a second identification module. The first identification module is used to construct a bidding behavior anomaly identification model to determine whether there is a suspicion of first joint collusion based on the bidder's bid quotation, rejected bid and evaluation score data and the bidding behavior anomaly identification model. The second identification module is used to determine the bidder suspected of first joint collusion according to the abnormal identification result sequence determined by the first identification module, make a correlation judgment on the suspected bidders, determine whether there is a suspicion of target joint collusion, and generate an abnormality assessment report.

[0018] The construction of the abnormal bidding behavior identification model determines whether there is a suspicion of first joint bidding collusion based on the bidder's bid quotation, bid rejection and bid evaluation score data and the abnormal bidding behavior identification model, including:

[0019] Step S110: Acquire historical bidding data sets and perform data preprocessing to obtain multiple sets of user behavior feature sequences and anomaly recognition result sequences.

[0020] Among them, the data set includes the user behavior feature data of the bidders in each bidding project, and the abnormal identification results of the bidders. The user behavior feature data includes the bidder number, whether the bid is successful, whether the bid is invalid, the bid price, the difference between the bid price and the bidding budget, and the evaluation score; the abnormal identification result is a mark of whether each bidder has a bidding joint.

[0021] Optionally, the step S110 further includes: obtaining synthetic sample data using a sample enhancement method to amplify the training data, and performing data preprocessing on the synthetic sample data to obtain multiple groups of user behavior feature sequences and anomaly recognition result sequences.

[0022] Step S120: construct a bidding behavior anomaly recognition model, and train the model based on multiple groups of user behavior feature sequences and anomaly recognition result sequences to obtain a converged bidding behavior anomaly recognition model.

[0023] In one embodiment, a bidding behavior anomaly identification model is constructed using a Transformer algorithm.

[0024] Optionally, in some embodiments, the bidding behavior anomaly identification model may also use a convolutional neural network (CNN), a multi-layer perceptron (MLP), a support vector machine (SVM) or other types of machine learning models.

[0025] Step S130: Obtain an abnormal identification result sequence according to the user behavior feature sequence of the bidder to be identified and the convergent bidding behavior abnormal identification model.

[0026] Step S140: Determine whether there is a first joint fraud suspicion based on the obtained abnormal identification result sequence.

[0027] At this time, if it is determined according to the first identification module that there is no suspicion of the first joint collusion in bidding, a notification message of "no suspicion of target joint collusion in bidding" is sent to the notification module; if it is determined according to the first identification module that there is suspicion of the first joint collusion in bidding, the second identification module further judges and identifies the bidders suspected of the first joint collusion in bidding based on the abnormal identification result sequence determined by the first identification module, makes a correlation judgment on the suspected bidders, determines whether there is suspicion of target joint collusion in bidding, and generates an abnormal evaluation report.

[0028] The second identification module determines the bidders suspected of first joint bid collusion according to the abnormal identification result sequence determined by the first identification module, makes a correlation judgment on the suspected bidders, determines whether there is a target joint bid collusion suspicion, and generates an abnormal evaluation report, including:

[0029] Step S210: determining a bidder suspected of first joint bidding collusion according to the abnormal recognition result sequence determined by the first recognition module;

[0030] Step S220: Obtain the basic enterprise information of each bidder through the crawler module, and determine whether each bidder belongs to the first affiliated enterprise based on the basic enterprise information of each bidder suspected of first joint bid collusion.

[0031] Step S230: if there is a first associated enterprise, determine at least one first associated enterprise set, and determine whether the proportion of the number of bidders in the first associated enterprise set among the bidders suspected of joint bidding collusion is greater than a second preset threshold; if it is greater than the second preset threshold, determine that there is a target joint bidding collusion suspicion, and generate an abnormal assessment report;

[0032] Step S240: If the number of bidders in the first associated enterprise set accounts for no more than the second preset threshold value among the bidders suspected of the first joint bidding collusion, or there is no first associated enterprise, the historical bidding data in the database is retrieved to determine whether the bidders suspected of the first joint bidding collusion have made multiple simultaneous bids in history. If so, it is determined that there is a target joint bidding collusion suspicion and an abnormal assessment report is generated. If not, it is determined that there is no target joint bidding collusion suspicion.

[0033] The second identification module conducts multi-dimensional comparison of the basic corporate information and historical joint bidding information of the bidders suspected of the first joint bidding collusion, discovers from multiple aspects whether there is any connection between the bidders, and further identifies whether the bidders suspected of the first joint bidding collusion have engaged in joint bidding collusion.

[0034] Compared with the prior art, this application has the following advantages:

[0035] 1. This application provides an electronic bidding management system that integrates intelligent identification of bidding behavior, realizes the integrated management of the entire process of bidding, tendering, bid opening, bid evaluation, and bid determination. Through the application of information technology, it greatly promotes the convenience of remote bidding, realizes data interoperability between different processes, and improves the transparency, confidentiality and work efficiency of the entire bidding process.

[0036] 2. This application provides an intelligent identification method that combines a supervised learning model and a bidder correlation judgment through a bidding anomaly identification module. It comprehensively considers whether bidders have price collusion, deliberately set defects in bidding documents to cause low scores or bid cancellation, and the correlation between bidders based on multi-dimensional comparisons. It identifies from multiple aspects whether bidders have engaged in joint bidding to ensure fairness and justice in bidding and maintain a good market order.

[0037] 3. This application combines the supervised learning model with the intelligent recognition method of the bidder correlation judgment, and preliminarily identifies the bidder's bid quotation, abandoned bid and evaluation score data based on the supervised learning model, and then further judges and identifies the preliminary recognition results through the bidder correlation judgment to determine whether there is suspicion of joint bidding. In the application scenario of abnormal bidding behavior analysis, this hierarchical and staged recognition method, on the one hand, overcomes the limitations of the traditional intelligent recognition method using behavioral feature analysis by introducing a supervised learning model, and improves the adaptability to complex and changeable bidding behavior situations in reality through the generalization ability of the model. On the other hand, through staged processing, it effectively reduces the complexity of the input feature data in the supervised learning model, making it easier for the model to train the correlation between the input features and abnormal bidding behavior, improve the recognition accuracy of the model, and significantly reduce the computing resource requirements and computing time, and improve the recognition efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0039] Figure 1 A schematic diagram of the architecture of an electronic bidding management system with integrated intelligent identification of bidding behavior provided in an embodiment of the present application;

[0040] Figure 2 A schematic diagram of the operation steps in the first identification module provided in an embodiment of the present application;

[0041] Figure 3 A schematic diagram of the operation steps in the second identification module provided in the embodiment of the present application;

[0042] Figure 4A schematic diagram of the Transformer model structure provided in an embodiment of the present application.

[0043] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0044] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0045] Figure 1 A schematic diagram of the architecture of an electronic bidding management system that integrates intelligent identification of bidding behavior provided in an embodiment of the present application.

[0046] like Figure 1 As shown, the present application provides an electronic bidding management system integrating intelligent identification of bidding behavior, including: a user presentation layer, a business logic layer and a data layer, wherein the data layer is communicatively connected to the business logic layer, and the business logic layer is communicatively connected to the user presentation layer;

[0047] The user presentation layer is configured to display data to users and accept user input based on terminal devices, thereby realizing interaction between users and the system; the business logic layer is configured as the part that provides core services in the system architecture, and is used to provide users with a variety of business functions for electronic bidding and tendering management; the data layer is used for data storage, retrieval and management.

[0048] The business logic layer includes a bidding management module 11, a bid management module 12, a CA authentication module 13, a bid opening and evaluation module 14, a crawler module 15, a bid anomaly identification module 16, and a notification module 17;

[0049] The bidding management module 11 is used for the bidder to upload or enter bidding data, and the bidding data at least includes the bidder information and bidding documents;

[0050] In one implementation of the present application, the bidding document is uploaded to the system in the form of an electronic attachment.

[0051] As another optional implementation, the bidding management module 11 provides multiple bidding document editing blocks, each of which generates corresponding bidding block information according to the input information of the bidder. The bidding block information includes the name of the corresponding bidding document editing block and the content of the filled information and / or the uploaded file. The bidding management module 11 generates a structured bidding document according to the multiple bidding block information.

[0052] For example, in the tender document editing block named "Tender Budget Range", the information that needs to be filled in is the upper and / or lower limit quotation range of the tender quotation; in the tender document editing block named "Technical Specifications and Requirements", the information that needs to be filled in is the detailed contents of the project's technical parameters, performance requirements, standards and specifications.

[0053] The bidding management module 12 is used for bidders to download bidding documents, and upload or enter qualification materials and bidding data, where the bidding data at least includes bidder information and bidding documents;

[0054] In one implementation of the present application, the bidding document is uploaded to the system in the form of an electronic attachment.

[0055] As another optional implementation, the bidding management module 12 provides multiple bidding editing blocks, each of which generates corresponding bidding block information according to the input information of the bidder. The bidding block information includes the name of the corresponding bidding editing block and the content of the filled information and / or the uploaded file. The bidding management module 12 generates a structured bidding document according to the multiple bidding block information.

[0056] The CA authentication module 13 is used to authenticate the bidding documents and tender documents in the project bidding process and encrypt and decrypt the data. The CA authentication module 13 uses a CA digital certificate to digitally sign the bidding documents and tender documents to verify the identities of the tenderer and the bidder. Preferably, the CA authentication module 13 uses an asymmetric encryption method to encrypt and decrypt the tender documents uploaded by the bidder to ensure the confidentiality of the bidding information and the transmission security.

[0057] The bid opening and evaluation module 14 is used to decrypt the bid documents to obtain the bid structured data, and provide the evaluation committee members with online rejection and online scoring functions to generate an evaluation report. The evaluation report at least includes the winning bid result, the bidders who have rejected the bid, and the evaluation scores of each bidder.

[0058] Among them, the bid is cancelled due to obvious defects in the bid, for example, the bid document fails to respond to the substantive requirements of the tender documents and has major deviations.

[0059] The crawler module 15 is used to obtain the basic enterprise information of the bidder from the Internet according to the bidder information in response to the instruction of the bid anomaly identification module 16, and perform data preprocessing.

[0060] The crawler module 15 is connected to third-party software through an interface, such as existing enterprise information query software such as Qizhidao and Tianyancha, and can also be obtained in other known ways, which is not limited in this application.

[0061] The bidding anomaly identification module 16 is used to obtain bidding information, evaluation report information, basic corporate information of bidders and historical joint bidding information, determine whether bidders are suspected of joint bidding through intelligent identification methods, and generate anomaly identification automatic evaluation results.

[0062] The notification module 17 is used to push the abnormal identification automatic evaluation results determined by the bidding abnormality identification module 16, and the abnormal identification automatic evaluation results include the notification information of "no target joint bid collusion suspicion found", or the notification information of "target joint bid collusion suspicion exists", and the abnormal evaluation report sent together with the notification information of "target joint bid collusion suspicion exists", so as to facilitate the manual review of the bidding staff. The abnormal evaluation report includes the enterprise information of the bidder suspected of the first joint bid collusion, the bid quotation, the bid evaluation score, the bidder who cancelled the bid, the correlation information between the bidders, etc.

[0063] Further optionally, the notification module 17 is also used to publish a winning bid announcement and a winning bid notice, or to publish a notice that the bidding is invalid due to illegal behavior.

[0064] Specifically, after the bidding staff manually intervenes and reviews the automatic evaluation results of abnormal identification pushed by the notification module 17, the final review conclusion is determined. The notification module 17 responds to the input of the bidding staff to publish the winning announcement and winning notice, or publishes a notice that the bidding is invalid due to violations.

[0065] As an optional embodiment of the present application, the bidding anomaly identification module 16 includes a first identification module and a second identification module. The first identification module is used to construct a bidding behavior anomaly identification model to determine whether there is a suspicion of first joint bidding collusion based on the bidder's bid quotation, rejected bid and evaluation score data and the bidding behavior anomaly identification model. The second identification module is used to determine the bidder suspected of first joint bidding collusion according to the abnormal identification result sequence determined by the first identification module, make a correlation judgment on the suspected bidders, determine whether there is a suspicion of target joint bidding collusion, and generate an abnormality evaluation report.

[0066] At present, in bidding, some bidders may collude on bidding matters, organize multiple related bidders to bid for the same bidding project, and make a certain stakeholder win the bid, thereby seeking profits and disrupting the market order.

[0067] When a project begins bidding, bidders who participate in joint bidding will bid for the project. On the one hand, the joint bidders usually exclude and suppress other bidders by means of price combination. On the other hand, the joint bidders will deliberately set loopholes and defects in the bidding documents by studying the bidding documents and bid evaluation methods, such as deviations in technical responses, which will lead to low evaluation scores or even bid cancellation. Therefore, it is necessary to identify potential joint bidding behaviors through multi-dimensional analysis of bidding behaviors to ensure fairness and justice in bidding and maintain a good market order.

[0068] At present, when identifying abnormal bidding behaviors, intelligent identification methods based on behavioral feature analysis are usually adopted, or multi-source feature data are input into supervised learning models for intelligent identification. However, intelligent identification methods based on behavioral feature analysis are usually highly subjective, difficult to adapt to complex and changing situations, and have poor generalization ability. On the other hand, in the application scenarios of abnormal bidding behavior identification, the feature data that need to be considered are diverse, such as the bidder's equity structure information, management information, credit status, historical joint bidding, bidding quotations, evaluation scores, winning bids, and cancelled bids. At this time, if the method of inputting multi-source feature data into a supervised learning model for intelligent identification is adopted, due to the various complex associations and rules between the input multi-source feature data and the abnormal bidding behavior identification results, the supervised learning model may find it difficult to capture the corresponding associations. In addition, multi-source data have different formats and structures. How to obtain the input feature vector to accurately characterize the multi-source data is also a major problem. In addition, due to the large amount of input data that needs to be processed, a large amount of computing resources and computing time are required.

[0069] To this end, the present application provides an intelligent identification method for a joint supervised learning model and a bidder correlation judgment. Specifically, the bidding anomaly identification module 16 includes a first identification module and a second identification module. The first identification module is used to construct a bidding behavior anomaly identification model, and determine whether there is a first joint bid collusion suspicion based on the bidder's bid quotation, bid rejection, and bid evaluation score data and the bidding behavior anomaly identification model. The second identification module is used to determine the bidder suspected of the first joint bid collusion according to the abnormal identification result sequence determined by the first identification module, perform correlation judgment on the suspected bidder, determine whether there is a target joint bid collusion suspicion, and generate an abnormal evaluation report.

[0070] like Figure 2As shown, the construction of the bidding behavior abnormality identification model determines whether there is a first joint bidding collusion suspicion based on the bidder's bid quotation, bid rejection and bid evaluation score data and the bidding behavior abnormality identification model:

[0071] Step S110: Acquire historical bidding data sets and perform data preprocessing to obtain multiple sets of user behavior feature sequences and anomaly recognition result sequences.

[0072] Among them, the data set includes the user behavior feature data of the bidders in each bidding project, and the abnormal identification results of the bidders. The user behavior feature data includes the bidder number, whether the bid is successful, whether the bid is invalid, the bid price, the difference between the bid price and the bidding budget, and the evaluation score; the abnormal identification result is a mark of whether each bidder has a bidding joint.

[0073] Among them, the anomaly recognition results in the historical bidding data set can be determined based on judgment rules, or can be determined using other known existing technologies, and this application does not limit this.

[0074] The data preprocessing includes feature encoding and normalization processing. Feature encoding is used to convert non-numerical features (such as bidder number, whether the bid is successful, whether the bid is invalid) into a numerical form that can be processed by the model. For example, label encoding, one-hot encoding, ordinal encoding, etc. can be used; for numerical features (such as bid price, the difference between the bid price and the bidding budget, and the bid evaluation score), minimum and maximum normalization (Min-Max) processing is performed.

[0075] After preprocessing the user behavior feature data and anomaly recognition results of all bidders in each bidding project, the user behavior feature sequence X = {x1, x2, ..., x m , …, x P} and the abnormal recognition result sequence Y = {y1, y2, ..., y m , …, y P}, where P is the number of bidders in the bidding project, x m represents the user behavior feature vector of each bidder, y m The abnormal identification result of each bidder, that is, the mark of whether each bidder has colluded in bidding, is represented by {0, 1}, where 0 indicates that there is no colluded bidding and 1 indicates that there is colluded bidding.

[0076] Optionally, the step S110 further includes: obtaining synthetic sample data using a sample enhancement method to amplify the training data, and performing data preprocessing on the synthetic sample data to obtain multiple groups of user behavior feature sequences and anomaly recognition result sequences.

[0077] Since the number of data samples of abnormal bidding behavior in bidding projects is usually much smaller than the number of data samples of normal bidding behavior, in order to make the results more accurate, a sample enhancement method is used to obtain sample data of abnormal bidding behavior to expand the training data. The sample enhancement method can be to synthesize sample data based on the characteristics of historical abnormal bidding behavior, or to generate synthetic sample data using a generative adversarial network (GAN), or other methods in the prior art can be used to obtain it.

[0078] Step S120: construct a bidding behavior anomaly recognition model, and train the model based on multiple groups of user behavior feature sequences and anomaly recognition result sequences to obtain a converged bidding behavior anomaly recognition model.

[0079] In one embodiment, a bidding behavior anomaly recognition model is constructed using the Transformer algorithm, such as Figure 4 As shown, the Transformer model includes an encoding layer and a decoding layer. The encoding layer may include M stacked encoders, and the decoding layer includes M stacked decoders. The encoder includes a multi-head self-attention module (Multi-head Self-Attention), a feedforward neural network module (FFN), and a residual normalization module (Add&Norm), wherein the number of residual normalization modules is two.

[0080] The decoder includes a feedforward neural network module, a residual normalization module, a multi-head self-attention module and a multi-head cross-attention module.

[0081] The multi-head self-attention mechanism is used to process a single input sequence. The input sequence is linearly transformed to obtain the query Q, key K and value V. The query Q of the multi-head cross-attention mechanism comes from the input of the decoding layer, and the key K and value V come from the output of the encoding layer. The calculation formula of the output attention vector Z after the multi-head self-attention module or the multi-head cross-attention module is as follows:

[0082]

[0083] In the encoding layer, the user behavior feature sequence is input, and the feature vector output after encoding will be used as the input of the decoding layer;

[0084] The decoding layer has two inputs, one is the output of the encoding layer, that is, the feature vector output by the encoding layer is used as the key K and value V of the multi-head cross attention mechanism module in the decoder, and the other is the object query sequence. The object query sequence (objectqueries) includes P object query vectors. The object query sequence is randomly initialized and encoded with the bidder number in sequence.

[0085] The object query outputs P vectors through the decoding layer, and finally obtains the abnormal recognition result sequence through the linear mapping layer, which is expressed as Y = {y1, y2, ..., y m , …, y P}.

[0086] Obtain multiple sets of user behavior feature sequences and anomaly recognition result sequences as training data for the Transformer model; use the training data as input, train the Transformer model using the loss function Loss, and use the back propagation algorithm to optimize and update the model parameters to obtain a converged Transformer model;

[0087] The calculation formula of the cross entropy loss function Loss is:

[0088]

[0089] In the formula, y i,m is the mth element in the anomaly recognition result sequence Y in the i-th sample, that is, y m The true output, p(y i,m ) is the output value of the Transformer model, and its value range is p(y i,m )∈[0,1]; N is the number of training data, and P is the number of elements in the anomaly recognition result sequence Y.

[0090] In this embodiment, the Transformer model can use the attention mechanism to capture long-distance dependencies and global context information in the sequence. For the bidding behavior data in the same bidding project, there are often complex global relationships and context dependencies. The Transformer model can better model these relationships, thereby improving the accuracy and robustness of anomaly detection.

[0091] Optionally, in some embodiments, the bidding behavior anomaly identification model may also use a convolutional neural network (CNN), a multi-layer perceptron (MLP), a support vector machine (SVM) or other types of machine learning models.

[0092] Step S130: Obtain an abnormal identification result sequence according to the user behavior feature sequence of the bidder to be identified and the convergent bidding behavior abnormal identification model.

[0093] Input the user behavior feature sequence X of the bidders to be identified in the current bidding project = {x1, x2, ..., x m , …, x P} to the convergent bidding behavior anomaly recognition model, and output the anomaly recognition result sequence Y = {y1, y2, ..., y m , …, y P}.

[0094] Step S140: Determine whether there is a first joint fraud suspicion based on the obtained abnormal identification result sequence.

[0095] According to the obtained abnormal recognition result sequence Y={y1, y2, ..., y m , …, y P}, determine the number of bidders who have engaged in joint bidding collusion. If the ratio of the number of bidders who have engaged in joint bidding collusion to the number of bidders P is greater than a first preset threshold T, it is determined that there is a first suspicion of joint bidding collusion. If not, it is determined that there is no first suspicion of joint bidding collusion. For example, T can be 30% based on experience.

[0096] Among them, for the element y in the obtained abnormal recognition result sequence Y m ,y m ∈[0, 1], the classification threshold can be set reasonably as needed, for example, the classification threshold is set to 0.5.

[0097] At this time, if it is determined according to the first identification module that there is no suspicion of the first joint collusion in bidding, a notification message of "no suspicion of target joint collusion in bidding" is sent to the notification module 17; if it is determined according to the first identification module that there is suspicion of the first joint collusion in bidding, the second identification module further judges and identifies the bidders suspected of the first joint collusion in bidding based on the abnormal identification result sequence determined by the first identification module, makes a correlation judgment on the suspected bidders, determines whether there is suspicion of target joint collusion in bidding, and generates an abnormal evaluation report.

[0098] Among them, Figure 3 As shown, the second identification module determines the bidders suspected of first joint bid collusion according to the abnormal identification result sequence determined by the first identification module, makes a correlation judgment on the suspected bidders, determines whether there is a target joint bid collusion suspicion, and generates an abnormal evaluation report, including:

[0099] Step S210: determining a bidder suspected of first joint bidding collusion according to the abnormal recognition result sequence determined by the first recognition module;

[0100] Step S220: Obtain the basic enterprise information of each bidder through the crawler module 15, and determine whether each bidder belongs to the first affiliated enterprise based on the basic enterprise information of each bidder suspected of first joint bid collusion.

[0101] The basic information of the bidder includes information of multiple dimensions, such as the business information of the bidder, equity relationship information, mutual investment relationship information, investment unit information, and information of the chairman, supervisor, senior executives and other management personnel of the bidder. If different bidders have the same keywords in the same dimension of information, it is judged that each bidder belongs to the first affiliated enterprise.

[0102] For example, by searching and comparing the shareholder information through the keyword of the name, it is found that the shareholders of the three bidders all include A, or by searching and comparing the investment units, it is found that the investment units of the three bidders all include B, then it is determined that the three bidders belong to the first affiliated enterprise.

[0103] Step S230: if there is a first associated enterprise, determine at least one first associated enterprise set, and determine whether the proportion of the number of bidders in the first associated enterprise set among the bidders suspected of joint bidding collusion is greater than a second preset threshold; if it is greater than the second preset threshold, determine that there is a target joint bidding collusion suspicion, and generate an abnormal assessment report;

[0104] Considering that the fact that some first related enterprises among the bidders jointly bid is usually considered as normal bidding behavior, and the probability of joint bidding behavior is relatively high when multiple related enterprises jointly bid, the second preset threshold is set as the judgment standard. Optionally, the second preset threshold can be set between 40% and 60% based on experience.

[0105] Among them, the abnormal evaluation report includes the corporate information of bidders suspected of the first joint bidding collusion, bid quotations, evaluation scores, and correlation information between bidders.

[0106] Step S240: If the number of bidders in the first associated enterprise set accounts for no more than the second preset threshold value among the bidders suspected of the first joint bidding collusion, or there is no first associated enterprise, the historical bidding data in the database is retrieved to determine whether the bidders suspected of the first joint bidding collusion have made multiple simultaneous bids in history. If so, it is determined that there is a target joint bidding collusion suspicion and an abnormal assessment report is generated. If not, it is determined that there is no target joint bidding collusion suspicion.

[0107] The step of judging whether the bidder suspected of first joint bidding collusion has made multiple simultaneous bids in the past, and if so, determining that there is a target joint bidding collusion suspicion; if not, determining that there is no target joint bidding collusion suspicion, includes:

[0108] If there are bidders with the first suspicion of joint bidding collusion whose number of historical simultaneous bids is greater than the third preset threshold and whose number of times is greater than the fourth preset threshold among the bidders with the first suspicion of joint bidding collusion, it is determined that there is suspicion of joint bidding collusion for the target; if not, it is determined that there is no suspicion of joint bidding collusion for the target. The third preset threshold can be set to 50%, for example, and the fourth preset threshold can be reasonably determined based on experience, for example, preferably 3 times.

[0109] The second identification module conducts multi-dimensional comparison of the basic corporate information and historical joint bidding information of the bidders suspected of the first joint bidding collusion, discovers from multiple aspects whether there is any connection between the bidders, and further identifies whether the bidders suspected of the first joint bidding collusion have engaged in joint bidding collusion.

[0110] The present application provides an intelligent identification method that combines a supervised learning model and a bidder correlation judgment through a bidding anomaly identification module, wherein a first identification module is used to construct a bidding behavior anomaly identification model, and determines whether there is a first joint bidding collusion suspicion based on the bidder's bid quotation, abandoned bid and bid evaluation score data and the bidding behavior anomaly identification model. If it is determined that there is no first joint bidding collusion suspicion, then it is determined that there is no target joint bidding collusion suspicion. If it is determined that there is a first joint bidding collusion suspicion, the second identification module further judges and identifies. The second identification module determines the bidder suspected of the first joint bidding collusion according to the abnormal identification result sequence determined by the first identification module, and makes a correlation judgment on the suspected bidders to determine whether there is a target joint bidding collusion suspicion.

[0111] It can be seen that through the first identification module and the second identification module, it is comprehensively considered whether the bidders have price collusion, deliberately set defects in the bidding documents to cause low scores or bid cancellation, and the correlation between bidders based on multi-dimensional comparison, thereby ensuring the accuracy and reliability of abnormal bidding behavior detection.

[0112] Furthermore, this application combines the supervised learning model with the intelligent recognition method of the bidder correlation judgment, and preliminarily identifies the bidder's bid quotation, abandoned bid and evaluation score data based on the supervised learning model, and then further judges and identifies the preliminary recognition results through the bidder correlation judgment to determine whether there is suspicion of joint bidding. In the application scenario of abnormal bidding behavior analysis, this hierarchical and staged recognition method, on the one hand, overcomes the limitations of the traditional intelligent recognition method using behavioral feature analysis by introducing a supervised learning model, and improves the adaptability to complex and changeable bidding behavior situations in reality through the generalization ability of the model. On the other hand, through staged processing, it effectively reduces the complexity of the input feature data in the supervised learning model, making it easier for the model to train the correlation between the input features and abnormal bidding behavior, improve the recognition accuracy of the model, and significantly reduce the computing resource requirements and computing time, and improve the recognition efficiency.

[0113] Preferably, if Figure 1 As shown, the user display layer displays system function data through PC clients, PC web pages, mobile apps, WeChat applets, etc., and develops the front-end interface through technologies such as HTML5, CSS3, VUE, and WebSocket.

[0114] The data layer is used to access the database and perform operations such as adding, deleting, modifying, updating, and searching the database. The database includes a relational database and / or a non-relational database, such as MySQL, SQL Server, and MongoDB.

[0115] In summary, the present application provides an electronic bidding management system that integrates intelligent identification of bidding behavior, realizes integrated management of the entire process of bidding, tendering, bid opening, bid evaluation, and bid determination. Through the application of information technology, it greatly promotes the convenience of remote bidding, realizes data interoperability between different processes, and improves the transparency, confidentiality and work efficiency of the entire bidding process.

[0116] Furthermore, the present application provides an intelligent identification method that combines a supervised learning model and a bidder correlation judgment through a bidding anomaly identification module. It comprehensively considers whether bidders have price collusion, deliberately set defects in bidding documents to result in low scores or bid cancellation, and the correlation between bidders based on multi-dimensional comparisons. It identifies from multiple aspects whether bidders have engaged in joint bidding collusion, so as to ensure the fairness of bidding and maintain a good market order.

[0117] In the above embodiments, the description of each embodiment has its own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant description of other embodiments. The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0118] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0119] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. An electronic bidding management system integrating intelligent identification of bidding behavior, characterized in that: include: A business logic layer, the business logic layer comprising a bidding management module (11), a bid management module (12), a bid opening and evaluation module (14), a crawler module (15) and a bid anomaly identification module (16); The bidding management module (11) is used for the bidder to upload or input bidding data, wherein the bidding data at least includes the bidder information and bidding documents; The bidding management module (12) is used for allowing bidders to download bidding documents, and upload or input qualification materials and bidding data, wherein the bidding data at least includes bidder information and bidding documents; The bid opening and evaluation module (14) is used to decrypt the bid documents to obtain bid structured data, and provide the evaluation committee members with online rejection and online scoring functions to generate an evaluation report; the evaluation report at least includes the winning bid result, the bidders whose bids were rejected, and the evaluation scores of each bidder; A crawler module (15) is used to obtain basic enterprise information of the bidder from the Internet according to the bidder information in response to the instruction of the bid anomaly identification module (16), and perform data preprocessing; The bidding anomaly identification module (16) comprises a first identification module and a second identification module, wherein the first identification module is used to construct a bidding behavior anomaly identification model, and determine whether there is a first joint bidding collusion suspicion based on the bidder's bid quotation, abandoned bid and bid evaluation score data and the bidding behavior anomaly identification model; the second identification module is used to determine the bidder suspected of the first joint bidding collusion according to the abnormal identification result sequence determined by the first identification module, perform correlation judgment on the suspected bidders, determine whether there is a target joint bidding collusion suspicion, and generate an abnormal evaluation report.

2. The electronic bidding management system with integrated intelligent identification of bidding behavior according to claim 1 is characterized in that: If it is determined according to the first identification module that there is no suspicion of the first joint bid-rigging, then it is determined that there is no suspicion of the target joint bid-rigging. If it is determined according to the first identification module that there is suspicion of the first joint bid-rigging, the second identification module will further judge and identify. The second identification module determines the bidder suspected of the first joint bid-rigging according to the abnormal identification result sequence determined by the first identification module, makes a correlation judgment on the suspected bidders, determines whether there is suspicion of the target joint bid-rigging, and generates an abnormal evaluation report.

3. The electronic bidding management system with integrated intelligent identification of bidding behavior according to claim 2 is characterized in that: The construction of the abnormal bidding behavior identification model, and determining whether there is a first joint bidding collusion suspicion based on the bidder's bid quotation, abandoned bid, and bid evaluation score data and the abnormal bidding behavior identification model, includes: Step S110: Obtain historical bidding data sets and perform data preprocessing to obtain multiple sets of user behavior feature sequences X and anomaly recognition result sequences Y; Step S120: construct a bidding behavior anomaly recognition model, and train the model based on multiple groups of user behavior feature sequences and anomaly recognition result sequences to obtain a converged bidding behavior anomaly recognition model; Step S130, obtaining an abnormal identification result sequence according to the user behavior feature sequence of the bidder to be identified and the convergent bidding behavior abnormal identification model; Step S140: Determine whether there is a first joint fraud suspicion based on the obtained abnormal identification result sequence.

4. The electronic bidding management system with integrated intelligent identification of bidding behavior according to claim 3 is characterized in that: The step S110 also includes: obtaining synthetic sample data using a sample enhancement method to amplify the training data, and performing data preprocessing on the synthetic sample data to obtain multiple groups of user behavior feature sequences and abnormality recognition result sequences.

5. The electronic bidding management system with integrated intelligent identification of bidding behavior according to claim 3 is characterized in that: The user behavior feature sequence is obtained by preprocessing the user behavior feature data, wherein the user behavior feature data includes the bidder number, whether the bid is successful, whether the bid is cancelled, the bid price, the difference between the bid price and the bidding budget, and the bid evaluation score; The abnormal identification result sequence is a mark of whether each bidder has a bidding combination.

6. The method according to claim 3, characterized in that The bidding behavior anomaly identification model adopts the Transformer model. During model training, the calculation formula of the cross entropy loss function Loss is: In the formula, y i,m is the true output of the mth element in the anomaly recognition result sequence Y in the i-th sample, p(y i,m ) is the output value of the Transformer model, and its value range is p(y i,m )∈[0,1]; N is the number of training data, and P is the number of elements in the anomaly recognition result sequence Y.

7. The electronic bidding management system with integrated intelligent identification of bidding behavior according to claim 2 is characterized in that: The second identification module determines the bidders suspected of first joint bid collusion according to the abnormal identification result sequence determined by the first identification module, makes a correlation judgment on the suspected bidders, determines whether there is a target joint bid collusion suspicion, and generates an abnormal evaluation report, including: Step S210: determining a bidder suspected of first joint bidding collusion according to the abnormal recognition result sequence determined by the first recognition module; Step S220: Obtaining basic enterprise information of each bidder through the crawler module (15), and judging whether each bidder belongs to the first affiliated enterprise based on the basic enterprise information of each first suspected joint bid collusion bidder; Step S230: if there is a first associated enterprise, determine at least one first associated enterprise set, and determine whether the proportion of the number of bidders in the first associated enterprise set among the bidders suspected of joint bidding collusion is greater than a second preset threshold; if it is greater than the second preset threshold, determine that there is a target joint bidding collusion suspicion, and generate an abnormal assessment report; Step S240: If the number of bidders in the first associated enterprise set accounts for no more than the second preset threshold value among the bidders suspected of the first joint bidding collusion, or there is no first associated enterprise, the historical bidding data in the database is retrieved to determine whether the bidders suspected of the first joint bidding collusion have made multiple simultaneous bids in history. If so, it is determined that there is a target joint bidding collusion suspicion and an abnormal assessment report is generated. If not, it is determined that there is no target joint bidding collusion suspicion.

8. The electronic bidding management system with integrated intelligent identification of bidding behavior according to claim 1 is characterized in that: The business logic layer also includes a notification module (17), which is used to push the identification results determined by the bidding anomaly identification module (16), including notification information of "no suspicion of joint bidding collusion among the target is found", or notification information of "there is suspicion of joint bidding collusion among the target", and an abnormality assessment report sent together with the notification information of "there is suspicion of joint bidding collusion among the target", so as to facilitate manual review by the bidding staff.

9. The electronic bidding management system with integrated intelligent identification of bidding behavior according to claim 1 is characterized in that: The business logic layer also includes a CA authentication module (13), which is used to perform identity authentication and data encryption and decryption on bidding documents and tender documents in the project bidding process.

10. The electronic bidding management system with integrated intelligent identification of bidding behavior according to claim 1 is characterized in that: The system further comprises a user display layer and a data layer, wherein the data layer is communicatively connected to the business logic layer, and the business logic layer is communicatively connected to the user display layer; the user display layer is configured to display data to the user and accept the user's input based on the terminal device, so as to realize the interaction between the user and the system; The business logic layer is configured as the part that provides core services in the system architecture, and is used to provide users with a variety of business functions for electronic bidding and tendering management; the data layer is used for data storage, retrieval and management.

Images