An intelligent supervision system for the whole process of project engineering bidding and tendering

Abstract

An intelligent supervision system for the whole process of engineering project bidding and tendering, which relates to the technical field of engineering project management, includes a project bidding data center. The project bidding data center is communicatively linked with a client, a pre-bid supervision module, an intelligent feature analysis module, a mid-bid supervision module, and a whole-process supervision module. The pre-bid supervision module is used to conduct multi-dimensional comparisons before the bid and generate pre-bid scoring coefficients for project bidders. The intelligent feature analysis module is used to obtain the weight coefficients of various data indicators of project bidders for associated feature factors. The mid-bid supervision module constructs an association relationship map, conducts horizontal feature comparisons on the bid document data of each project bidder, and conducts bid-rigging behavior analysis on the historical project bid data of each project bidder, and makes a secondary judgment on the suspected bid-rigging objects. The whole-process supervision module monitors the pre-bid scoring coefficients of each project bidder in real time, significantly improving the efficiency and effectiveness of the supervision of the bidding process.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering project management, and in particular to an intelligent supervision system for the entire process of bidding and tendering for an engineering project. Background Art

[0002] A Chinese patent with publication number CN114511162A discloses a method and system for assessing the risk of bidding for an engineering project, comprising the following steps: conducting a bidding risk analysis on the engineering project to be bid, establishing a game tree model based on the game logic relationship between the project integrator and the project contractor, obtaining the theoretical compliance probability of a qualified project contractor whose expected project revenue can meet the expected revenue demand of the project integrator based on the project stakeholder data determined by the game tree model, obtaining the actual existence probability of an actual qualified project contractor based on the Delphi method, and determining whether the qualified project contractor has a bidding risk based on the comparison result between the theoretical compliance probability and the actual existence probability.

[0003] There are many drawbacks in the bidding activities of the above-mentioned engineering projects. On the one hand, the supervision process relies on a large number of manual operations, which is inefficient and prone to omissions. For example, the review of the bidding documents needs to be manually checked word by word, which makes it difficult to quickly find illegal content. On the other hand, faced with massive transaction data and complex transaction relationships, bid rigging and collusion often occur, seriously disrupting the market order. In addition, there is a serious phenomenon of information islands between different departments, and there is a lack of effective data sharing mechanisms. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of the present invention is to provide an intelligent supervision system for the whole process of bidding and tendering for engineering projects, including a project bidding data center, wherein the project bidding data center is connected in communication with a client, a pre-bid supervision module, an intelligent feature analysis module, a bidding supervision module and a whole process supervision module;

[0005] The client has two roles: a project tenderer and a project bidder. The project tenderer is used to upload project tender information to the project tender data center, and the project bidder is used to send an access request to the project tenderer through the project tender data center.

[0006] The pre-bid supervision module is used to compare the project bidders and the project tenderers in multiple dimensions before bidding, and generate the pre-bid scoring coefficient of the project bidders based on the comparison results;

[0007] The intelligent feature analysis module is used to perform intelligent feature analysis on several historical bid winning records during the multi-dimensional comparison process before bidding, and obtain the weight coefficients of various data indicators of the project bidders for the associated feature factors;

[0008] The bid pre - supervision module is used to construct an association relationship graph, compare the horizontal features of the bid document data of bidders for each project, analyze the bid - rigging behavior of the bid data of several historical projects of bidders for each project, and conduct a secondary judgment on the suspected bid - rigging objects;

[0009] The whole - process supervision module is used to monitor the pre - bid scoring coefficients of bidders for each project in real - time and generate bid warning signals according to the monitoring results.

[0010] Furthermore, the process of the bid pre - supervision module comparing the bidder for a project with the project tenderer in multiple dimensions before the bid and generating the pre - bid scoring coefficient of the bidder for the project includes:

[0011] The bidder for a project logs in to the project tendering data center by inputting the identity ID and password, and the project tendering data center obtains the multi - dimensional data of the bidder for the project according to the identity ID. The multi - dimensional data includes basic data, performance data, and credit data, and inputs the identity ID of the project tenderer that the bidder for the project needs to access into the project tendering data center according to the access request. The project tendering data center obtains the project tendering information of the project tenderer, extracts the project features from the project tendering information, and obtains the project tendering features;

[0012] Extract each data index in the basic data, performance data, and credit data of the bidder for the project as evaluation indexes, set the index weights of the evaluation indexes according to the project tendering features. Specifically, extract each feature factor in the project tendering features, obtain the data indexes of the basic data, performance data, and credit data associated with each feature factor, obtain the weight coefficients of each data index for the associated feature factor, and obtain the membership degree matrix of the bidder for the project for the preset pre - bid scoring coefficient through fuzzy comprehensive evaluation;

[0013] Obtain the pre - bid scoring coefficient of the bidder for the project according to the membership degree matrix and the index weights, compare the pre - bid scoring coefficient of the bidder for the project with the preset pre - bid scoring coefficient threshold. If the pre - bid scoring coefficient is less than the pre - bid scoring coefficient threshold, reject the access request of the bidder for the project. If the pre - bid scoring coefficient is greater than or equal to the pre - bid scoring coefficient threshold, send the project tendering information of the project tenderer to the bidder for the project.

[0014] Furthermore, the process of obtaining the pre - bid scoring coefficient of the bidder for the project according to the membership degree matrix and the index weights includes:

[0015] Fuse the index weight and membership degree matrix of the evaluation index through a formula to obtain the fuzzy comprehensive evaluation matrix of the evaluation index, obtain the membership degree of the project tenderer for different pre-bid scoring coefficients according to the fuzzy comprehensive evaluation matrix, screen out the pre-bid scoring coefficient with the highest membership degree corresponding to the project tenderer, and use the pre-bid scoring coefficient with the highest membership degree corresponding to the project tenderer as the pre-bid scoring coefficient of the project tenderer;

[0016] Among them, the formula is:

[0017] ;

[0018] Among them, is the fuzzy comprehensive evaluation matrix of the evaluation index, is the index weight of the evaluation index, is the membership degree matrix, represents the multiplication of the elements at the corresponding positions of the weight matrix of the evaluation index and the membership degree matrix, is a weighted parameter used to control the balance between the weight matrix and the membership degree matrix in the fuzzy comprehensive evaluation matrix of the evaluation index.

[0019] Furthermore, the process of the intelligent feature analysis module performing intelligent feature analysis on a number of historical winning bids includes:

[0020] Extract a number of historical winning bids from the project bidding data center. The historical winning bids include project bidding characteristics, project winning bidders, and basic data, performance data, and credit data of each project tenderer. Extract each characteristic factor in the project bidding characteristics, obtain the data indicators of the basic data, performance data, and credit data associated with each characteristic factor, perform cluster analysis on the project bidding characteristics in a number of historical winning bids according to the project bidding characteristics, obtain the similarity between the project bidding characteristics in each historical winning bid and the current project bidding characteristics, perform clustering according to the similarity of the project bidding characteristics in each historical winning bid, and obtain a project bidding characteristic clustering circle. The similarity between the project bidding characteristics corresponding to each historical winning bid in the project bidding characteristic clustering circle is greater than the similarity threshold;

[0021] Perform statistical analysis on a number of historical winning bids in the project bidding characteristic clustering circle, obtain the Pearson correlation coefficient between the data indicators of the project winning bidder and each project tenderer and the associated characteristic factors, and at the same time perform a difference analysis on the data indicators of the project winning bidder and each project tenderer to obtain the difference coefficient of the data indicators for the associated characteristic factors;

[0022] Perform data distribution analysis on each data index in the project bidding feature clustering circle, obtain the frequency distribution histograms of each data index in the project winning bidder and the project tenderer respectively. The frequency distribution histogram uses the literal value of the data index as the abscissa and the frequency as the ordinate. Analyze the characteristics of the frequency distribution histogram to obtain the trend distribution deviation coefficients of each data index for the associated characteristic factors.

[0023] Further, the process of obtaining the trend distribution deviation coefficients of each data index for the associated characteristic factors includes:

[0024] Obtain the frequencies of different literal values in the frequency distribution histogram of each data index in the project winning bidder, mark the frequencies as the first frequencies, and the frequencies of different literal values in the frequency distribution histogram of the project tenderer, mark the frequencies as the second frequencies. According to the first frequencies and the second frequencies corresponding to different literal values of each data index, obtain the trend distribution coefficients of each data index in the project winning bidder and the project tenderer respectively. Compare the trend distribution coefficients of each data index in the project winning bidder and the project tenderer to obtain the trend distribution deviation coefficients of each data index for the associated characteristic factors.

[0025] Further, the specific process of obtaining the trend distribution deviation coefficients for the associated characteristic factors includes:

[0026] ;

[0027] , where The trend distribution deviation coefficient for the associated characteristic factor represents the frequency of the th literal value, is the number of literal values, is the bias coefficient.

[0028] Further, the process of obtaining the weight coefficients of each data index for the associated characteristic factors includes:

[0029] Standardize the Pearson correlation coefficients between the data indicators of the winning bidder of the project and each project bidder and the associated characteristic factors, the difference coefficients between the data indicators for the associated characteristic factors, and the trend distribution deviation coefficients between the data indicators for the associated characteristic factors. Based on the Pearson correlation coefficients between the data indicators of the winning bidder of the project and each project bidder and the associated characteristic factors, the difference coefficients between the data indicators for the associated characteristic factors, and the trend distribution deviation coefficients between the data indicators for the associated characteristic factors after standardization, obtain the weight coefficients of each data indicator for the associated characteristic factors.

[0030] Furthermore, the process of obtaining the weight coefficients of each data indicator includes:

[0031] Suppose there are data indicators and characteristic factors. Let represent the Pearson correlation coefficient between the -th data indicator of the winning bidder of the project and the associated characteristic factor , represent the Pearson correlation coefficient between the -th data indicator of the project bidder and the associated characteristic factor , represent the difference coefficient of the winning bidder of the project and each project bidder on the -th data indicator for the associated characteristic factor , represent the skewness coefficient difference of the -th data indicator for the associated characteristic factor ;

[0032] Perform standardization on , , , respectively. For the Pearson correlation coefficient , the standardized coefficient , where , represent the minimum and maximum values of all the Pearson correlation coefficients corresponding to the characteristic factor respectively. Similarly, obtain the standardized coefficients , , ;

[0033] ;

[0034] Among them, data indicator for the associated characteristic factor The weight coefficients, where , , are adjustment parameters used to adjust the relative importance of the Pearson correlation coefficient, the coefficient of difference, and the skewness coefficient difference when determining the weights, and , the above formulas are all calculated by removing the dimension and taking their numerical values. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters and preset thresholds in the formulas are set by those skilled in the art according to the actual situation or obtained by simulating a large amount of data.

[0035] Further, the process of the in-bid supervision module constructing the association relationship graph includes:

[0036] Obtain the main bodies in the current project bidding and tendering process and the association relationship information between the main bodies. The main bodies include project bidders, project tenders, tendering agencies, and bid evaluation experts. Take each main body as a node of the association relationship graph. Each node contains the attribute information of the corresponding main body. According to the association relationship information between the main bodies, construct different types of edges between the nodes. Each edge contains the corresponding association relationship. Preset the weight labels corresponding to different association relationships, and set the weight labels on each edge according to the association relationship corresponding to each edge.

[0037] Further, the process of the in-bid supervision module comparing the horizontal features of the bidding document data of each project bidder includes:

[0038] Obtain the bidding document data of each project bidder in the association relationship graph, perform word vector representation on the text content in the bidding document data to generate a word vector sequence, compare the cosine similarity of the word vector sequences corresponding to each bidding document data, obtain the text similarity between each bidding document data, and compare the text similarity between each bidding document data with the preset text similarity threshold;

[0039] If there is a text similarity between the bidding document data greater than the text similarity threshold, extract the numerical key indicators in the bidding document data (including bid price, bill of quantities, etc.), compare and analyze the numerical key indicators between the bidding document data, obtain the deviation values between various types of numerical key indicators, and perform weighted average processing on the deviation values between various types of numerical key indicators to obtain the comprehensive deviation rate between the bidding document data;

[0040] A standard threshold interval corresponding to a preset comprehensive deviation rate is set. Threshold points are selected within the standard threshold interval to divide sub-intervals of different reliability levels, and the reliability level corresponding to the comprehensive deviation rate between the bid document data is obtained. According to the reliability level and the weight labels of all the edges between the nodes to which the bid document data belongs, the risk loss value of the node to which the bid document data belongs is obtained. The pre-bid scoring coefficient of the node is updated according to the risk loss value, and the updated pre-bid scoring coefficient = the pre-bid scoring coefficient before update - the risk loss value.

[0041] Further, the process of obtaining the risk loss value of the node to which the bid document data belongs includes:

[0042] Let the number of edges between nodes be , the weight label of the th edge be , be the reliability level of the bid document data, be the total number of reliability levels;

[0043] ;

[0044] Among them, represents the risk loss value, is a regulation coefficient used to adjust the overall size of the risk loss value, .

[0045] Further, the process of the in-bid supervision module analyzing the bid-rigging behavior of the historical project bid data of each project bidder includes:

[0046] Extract the bid data of several historical projects of each project bidder in the associated relationship graph, preprocess the bid data of several historical projects, and use the normalization method to convert the bid price into a relative value to eliminate the influence of factors such as project scale and type on the price. According to the preprocessed bid data of several historical projects, obtain the bid price difference ratio between each project bidder and other project bidders for each project. The calculation formula is: price difference ratio = (bid price of bidder A - bid price of bidder B) / bid price of bidder B. Conduct a correlation analysis on the bid price difference ratio between each project bidder and other project bidders in each project to obtain the Pearson correlation coefficient between each project bidder and other project bidders. For example, if there are m projects, for bidders A and B, their price difference ratios in each project are (x1, x2,..., xm) and (y1, y2,..., ym) respectively. Calculate the correlation coefficient r according to the calculation method of the Pearson correlation coefficient. Judge the strength of the correlation according to the value of r. The closer r is to 1 or (-1), the stronger the correlation; the closer it is to 0, the weaker the correlation. Preset the first correlation coefficient threshold. If the Pearson correlation coefficient between a project bidder and other project bidders is greater than the first correlation coefficient threshold, mark the project bidder and other project bidders as suspected bid rigging objects.

[0047] Further, the process of secondary judgment on suspected bid rigging objects includes:

[0048] Build a behavior trend prediction model based on deep learning, extract project bidding characteristics and time characteristics from the bid data of several historical projects of suspected bid rigging objects, and obtain the bid price time series of suspected bid rigging objects for different project bidding characteristics. Use the bid price time series of suspected bid rigging objects for different project bidding characteristics as the training set and the test set. Input the training set into the behavior trend prediction model for training until the loss function is trained stably, and save the model parameters. Test the behavior trend prediction model through the test set until it meets the preset requirements, and output the behavior trend prediction model.

[0049] According to the behavior trend prediction model, output the predicted bid price of the suspected bid rigging object for the current project bidding characteristics, compare the bid price of the suspected bid rigging object for the current project bidding characteristics with the predicted bid price, obtain the bid price deviation value, and obtain the risk loss value of the node to which the suspected bid rigging object belongs according to the bid price deviation value and the weight labels of all edges between the nodes to which the suspected bid rigging object belongs. Update the pre-bid scoring coefficient of the node according to the risk loss value.

[0050] Further, the process of the whole-process supervision module for real-time monitoring of the pre-bid scoring coefficients of each project bidder and generating a bid warning signal according to the monitoring results includes:

[0051] Preset a bid warning score, and compare the pre-bid scoring coefficient of each project bidder with the bid warning score in real time. When the pre-bid scoring coefficient of the project bidder is less than the bid warning score, a bid warning signal is generated and fed back to the project tenderer.

[0052] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0053] 1. Multi-dimensional comprehensive evaluation: The pre-bid supervision module obtains multi-dimensional data of project bidders, makes multi-dimensional comparisons with the project tendering information of project tenderers, and uses fuzzy comprehensive evaluation to obtain the pre-bid scoring coefficient. This method changes the limitations of traditional single-index evaluation, comprehensively considers more factors related to the project, makes the preliminary screening of bidders more scientific and reasonable, can more accurately judge whether the bidder meets the project requirements, and improves the probability of screening out high-quality bidders.

[0054] 2. Deeply explore the value of historical data: The intelligent feature analysis module deeply analyzes a number of historical winning bid records, extracts the characteristic factors in the project tendering characteristics, and obtains the associated data indicators. Through methods such as cluster analysis and statistical analysis, Pearson correlation coefficients, difference coefficients, and trend distribution deviation coefficients are obtained, and then the weight coefficients of each data indicator for the associated characteristic factors are determined. Compared with the fuzzy comprehensive evaluation process of multi-dimensional matching in the prior art, where the weights of evaluation indicators are all determined according to the experience of experts, the present invention can more accurately grasp the key factors when evaluating bidders, provides a more targeted and scientific basis for pre-bid decision-making, and helps to select the most suitable bidder for the project.

[0055] 3. Detect bid rigging behavior in multiple dimensions: When comparing the horizontal features of bid document data, not only the text similarity is compared, but also the numerical key indicators are compared and analyzed, the comprehensive deviation rate is obtained and the risk loss value is calculated, and the pre-bid scoring coefficient of the node is updated. At the same time, analyze the bid rigging behavior of historical project bid data, mark the suspected bid rigging objects by calculating the correlation of the bid price difference ratio, and conduct a secondary judgment. This multi-dimensional detection method can more comprehensively and accurately identify illegal behaviors such as bid rigging, reduces the possibility of illegal behaviors occurring, and guarantees the fairness of the bidding process.

[0056] 4. Promote the continuity and comprehensiveness of supervision: The system covers the entire process of bidding, from pre-bid, in-bid to full-process supervision, forming a complete supervision system. Each module cooperates with each other and complements each other, realizing continuous and comprehensive supervision of the bidding process, helping to timely discover and solve problems arising in each link, and improving the efficiency and effect of supervision. Description of the Drawings

[0057] Figure 1 This is the schematic diagram of an intelligent supervision system for the whole process of project bidding and tendering in an embodiment of the present application. Specific implementation manners

[0058] Next, in combination with the accompanying drawings in the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts fall within the scope of protection of the present application.

[0059] As Figure 1 shown, an intelligent supervision system for the whole process of project bidding and tendering includes a project bidding data center, and the project bidding data center is communicatively linked with a client, a pre-bid supervision module, an intelligent feature analysis module, a mid-bid supervision module, and a whole-process supervision module;

[0060] The client has two roles: the project tenderer and the project bidder. The project tenderer is used to upload project bidding information to the project bidding data center, and the project bidder is used to send an access request to the project tenderer through the project bidding data center.

[0061] The pre-bid supervision module is used to conduct multi-dimensional comparison between the project bidder and the project tenderer before the bid, and generate a pre-bid scoring coefficient for the project bidder according to the comparison result.

[0062] The intelligent feature analysis module is used to conduct intelligent feature analysis on a number of historical winning bid records during the multi-dimensional comparison before the bid, and obtain the weight coefficients of various data indicators of the project bidder for the associated feature factors.

[0063] The mid-bid supervision module is used to construct an association relationship graph, conduct horizontal feature comparison on the bid document data of each project bidder, conduct bid-rigging behavior analysis on the historical project bid data of each project bidder, and conduct a secondary judgment on the suspected bid-rigging objects.

[0064] The whole-process supervision module is used to monitor the pre-bid scoring coefficients of each project bidder in real time, and generate a bid warning signal according to the monitoring result.

[0065] It should be further noted that, in the specific implementation process, the process of the pre-bid supervision module conducting multi-dimensional comparison between the project bidder and the project tenderer and generating a pre-bid scoring coefficient for the project bidder includes:

[0066] The project bidders log in to the project bidding data center by entering their identity ID and password. The project bidding data center obtains the basic data, performance data, and credit data of the project bidders based on the identity ID, and enters the identity ID of the project tenderers that the project bidders need to access into the project bidding data center according to the access request. The project bidding data center obtains the project bidding information of the project tenderers, extracts the project characteristics from the project bidding information, and obtains the project bidding characteristics. The project bidding characteristics include the complexity of technical requirements, precision requirements, the level of difference from existing technologies, contract type, payment method, margin requirements, and evaluation methods.

[0067] Extract each data index in the basic data, performance data, and credit data of the project bidders as evaluation indexes, including setting scoring criteria according to the registered capital, establishment years, etc. For example, enterprises with a high registered capital and a long establishment time may perform better in terms of stability and can obtain higher scores. The qualification level is also an important index, and a high-level qualification can correspond to a higher score. Set scoring criteria according to the winning bid amount, number of projects, project difficulty, etc. For example, those with a cumulative winning bid amount reaching a certain scale in the past three years can get high scores, and those who have undertaken similar large and complex projects can get bonus points. The quality evaluation results of the projects can also be used as the basis for scoring, and projects that have won high-quality project awards can get additional bonus points. Set scoring criteria according to credit data, etc. For example, those with a credit rating of AAA can be given a higher score, and those with bad credit records will be deducted points according to the severity. For example, enterprises without any disputes can get scores, and those involved in disputes should be deducted significantly. Set the index weights of the evaluation indexes according to the project bidding characteristics. Specifically, extract each characteristic factor in the project bidding characteristics, obtain the data indexes of the basic data, performance data, and credit data associated with each characteristic factor, and obtain the weight coefficients of each data index for the associated characteristic factor. For example, for projects with complex technologies, the weight of relevant technical performance in the performance data can be appropriately increased, and for projects with high requirements for enterprise reputation, the weight of credit data can be increased. Obtain the membership degree matrix of the project bidders for the preset pre-bid scoring coefficient through fuzzy comprehensive evaluation.

[0068] Obtain the pre-bid scoring coefficient of the project bidders according to the membership degree matrix and the index weights, and compare the pre-bid scoring coefficient of the project bidders with the preset pre-bid scoring coefficient threshold. If the pre-bid scoring coefficient is less than the pre-bid scoring coefficient threshold, reject the access request of the project bidders. If the pre-bid scoring coefficient is greater than or equal to the pre-bid scoring coefficient threshold, send the project bidding information of the project tenderers to the project bidders.

[0069] It should be further noted that in the specific implementation process, the process of obtaining the pre-bid scoring coefficient of the project bidders according to the membership degree matrix and the index weights includes:

[0070] Fuse the index weight and membership degree matrix of the evaluation index through a formula to obtain the fuzzy comprehensive evaluation matrix of the evaluation index, obtain the membership degree of the project tenderer for different pre-bid scoring coefficients according to the fuzzy comprehensive evaluation matrix, screen out the pre-bid scoring coefficient with the highest membership degree corresponding to the project tenderer, and use the pre-bid scoring coefficient with the highest membership degree corresponding to the project tenderer as the pre-bid scoring coefficient of the project tenderer;

[0071] Among them, the formula is:

[0072] ;

[0073] Among them, is the fuzzy comprehensive evaluation matrix of the evaluation index, is the index weight of the evaluation index, is the membership degree matrix, represents the multiplication of the elements at the corresponding positions of the weight matrix of the evaluation index and the membership degree matrix, is a weighted parameter used to control the balance between the weight matrix and the membership degree matrix in the fuzzy comprehensive evaluation matrix of the evaluation index.

[0074] It should be further noted that in the specific implementation process, the process of the intelligent feature analysis module performing intelligent feature analysis on a number of historical winning records includes:

[0075] Extract a number of historical winning records from the project bidding data center. The historical winning records include project bidding characteristics, project winning bidders, and basic data, performance data, and credit data of each project tenderer. Extract each characteristic factor in the project bidding characteristics. The characteristic factors include the complexity of technical requirements, precision requirements, the level of difference from existing technologies, contract type, payment method, guarantee requirements, and evaluation methods. Obtain the data indicators of the basic data, performance data, and credit data associated with each characteristic factor. For example, for the bidding characteristic of technical difficulty, the relevant indicators may include basic data such as the proportion of technical personnel of the winning bidder and tenderers, and R & D investment in technology, as well as performance data such as the completion of projects with similar technical difficulties. Perform clustering analysis on the project bidding characteristics in a number of historical winning records according to the project bidding characteristics, obtain the similarity between the project bidding characteristics in each historical winning record and the current project bidding characteristics, and perform clustering according to the similarity of the project bidding characteristics in each historical winning record to obtain a project bidding characteristic clustering circle. The similarity between the project bidding characteristics corresponding to each historical winning record in the project bidding characteristic clustering circle is greater than the similarity threshold;

[0076] Statistically analyze a number of historical winning bid records in the project bidding feature clustering circle, obtain the Pearson correlation coefficients between the winning bidders of the projects and various data indicators of each project bidder and the associated characteristic factors, and at the same time conduct a difference analysis on the various data indicators of the winning bidders of the projects and each project bidder to obtain the difference coefficients of the various data indicators for the associated characteristic factors;

[0077] Conduct a data distribution analysis on the various data indicators in the project bidding feature clustering circle, obtain the frequency distribution histograms of the various data indicators in the winning bidders of the projects and the project bidders respectively. The frequency distribution histogram takes the literal value of the data indicator as the abscissa and the frequency as the ordinate, and conduct a feature analysis on the frequency distribution histogram to obtain the trend distribution deviation coefficients of the various data indicators for the associated characteristic factors.

[0078] It should be further noted that in the specific implementation process, the process of obtaining the trend distribution deviation coefficients of the various data indicators for the associated characteristic factors includes:

[0079] Obtain the frequencies of different literal values in the frequency distribution histogram of the various data indicators in the winning bidders of the projects, mark the frequencies as the first frequencies, and the frequencies of different literal values in the frequency distribution histogram of the project bidders, mark the frequencies as the second frequencies. According to the first frequencies and the second frequencies corresponding to the different literal values of the various data indicators, obtain the trend distribution coefficients of the various data indicators in the winning bidders of the projects and the project bidders respectively, and compare the trend distribution coefficients of the various data indicators in the winning bidders of the projects and the project bidders respectively to obtain the trend distribution deviation coefficients of the various data indicators for the associated characteristic factors.

[0080] It should be further noted that in the specific implementation process, the specific process of obtaining the trend distribution deviation coefficients for the associated characteristic factors includes:

[0081] ;

[0082] , where The trend distribution deviation coefficient for the associated characteristic factor represents the frequency of the th literal value, is the number of literal values, is the bias coefficient. is the bias coefficient.

[0083] It should be further noted that in the specific implementation process, the process of obtaining the weight coefficients of the various data indicators for the associated characteristic factors includes:

[0084] Standardize the Pearson correlation coefficients between the data indicators of the winning bidder of the project and each project bidder and the associated characteristic factors, the difference coefficients between the data indicators for the associated characteristic factors, and the trend distribution deviation coefficients between the data indicators for the associated characteristic factors. Based on the standardized Pearson correlation coefficients between the data indicators of the winning bidder of the project and each project bidder and the associated characteristic factors, the difference coefficients between the data indicators for the associated characteristic factors, and the trend distribution deviation coefficients between the data indicators for the associated characteristic factors, obtain the weight coefficients of the data indicators for the associated characteristic factors.

[0085] It should be further noted that in the specific implementation process, the process of obtaining the weight coefficients of the data indicators includes:

[0086] Suppose there are data indicators, and characteristic factors. Let represent the Pearson correlation coefficient between the th data indicator of the winning bidder of the project and the associated characteristic factor represent the Pearson correlation coefficient between the th data indicator of the project bidder and the associated characteristic factor ; represent the difference coefficient of the winning bidder of the project and each project bidder on the th data indicator for the associated characteristic factor ; represent the skewness coefficient difference of the th data indicator for the associated characteristic factor ;

[0087] Standardize , , , respectively. For the Pearson correlation coefficient , the standardized coefficient , where , represent the minimum and maximum values of all the Pearson correlation coefficients corresponding to the characteristic factor respectively. Similarly, obtain the standardized coefficients , , ;

[0088] ;

[0089] Among them, data indicator For associated feature factors of the weight coefficient, where , , are adjustment parameters used to adjust the relative importance of the Pearson correlation coefficient, the difference coefficient, and the skewness coefficient difference when determining the weight, and . The above formulas are all calculated by removing the dimension and taking their numerical values. The formula is obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained by simulating a large amount of data.

[0090] It should be further noted that in the specific implementation process, the process of the in-bid supervision module constructing the association relationship graph includes:

[0091] Obtain the entities in the current project bidding and tendering process and the association relationship information between each entity. The entities include project bidders, project tenders, bidding agencies, and evaluation experts. The association relationship information includes business transactions (business cooperation records between each entity, such as a bidder participating in a project tender of a tenderer, a bidding agency providing bidding agency services for a tenderer, an evaluation expert participating in the evaluation of a bidding project, etc., information such as the time of business transactions and the project name), equity relationships, personnel appointment relationships, etc. Take each entity as a node of the association relationship graph. Each node contains the attribute information of the corresponding entity. According to the association relationship information between each entity, construct different types of edges between each node. Each edge contains the corresponding association relationship. Preset the weight labels corresponding to different association relationships, and set the weight labels on each edge according to the association relationship corresponding to each edge.

[0092] It should be further noted that in the specific implementation process, the process of the in-bid supervision module comparing the horizontal features of the bidding document data of each project bidder includes:

[0093] Obtain the bidding document data of each project bidder in the association relationship graph, perform word vector representation on the text content in the bidding document data to generate a word vector sequence, compare the cosine similarity of the word vector sequences corresponding to each bidding document data, and obtain the text similarity between each bidding document data. For example, if the text similarity of the key parts such as the construction steps and resource allocation plan in the construction organization designs of two project bidders exceeds 80%, there may be a suspicion of collusion. Compare the text similarity between each bidding document data with the preset text similarity threshold;

[0094] If the text similarity between the bid document data is greater than the text similarity threshold, extract the numerical key indicators in the bid document data (including bid price, bill of quantities, etc.), compare and analyze the numerical key indicators between the bid document data, obtain the deviation values between various types of numerical key indicators, and perform weighted average processing on the deviation values between various types of numerical key indicators to obtain the comprehensive deviation rate between the bid document data. For example, in the bills of quantities of different bidders, if the quantities of multiple items are exactly the same, these may all be signs of bid rigging;

[0095] Preset the standard threshold interval corresponding to the comprehensive deviation rate, select threshold points within the standard threshold interval to divide sub-intervals of different reliability levels, obtain the reliability level corresponding to the comprehensive deviation rate between the bid document data, and based on the reliability level and the weight labels of all the edges between the nodes to which the bid document data belongs, obtain the risk loss value of the node to which the bid document data belongs. Update the pre-bid scoring coefficient of the node according to the risk loss value. The updated pre-bid scoring coefficient = the pre-bid scoring coefficient before update - the risk loss value.

[0096] It should be further noted that in the specific implementation process, the process of obtaining the risk loss value of the node to which the bid document data belongs includes:

[0097] Let the number of edges between nodes be , the weight label of the th edge be , is the reliability level of the bid document data, is the total number of reliability levels;

[0098] ;

[0099] Among them, represents the risk loss value, is an adjustment coefficient used to adjust the overall size of the risk loss value, .

[0100] It should be further noted that in the specific implementation process, the process of the in-bid supervision module analyzing the bid rigging behavior of the historical project bid data of each project bidder includes:

[0101] Extract the bid data of several historical projects of each project bidder in the associated relationship graph, preprocess the bid data of several historical projects, and use the normalization method to convert the bid price into a relative value to eliminate the influence of factors such as project scale and type on the price. According to the preprocessed bid data of several historical projects, obtain the bid price spread ratio between each project bidder and other project bidders for each project. The calculation formula is: spread ratio = (bid price of bidder A - bid price of bidder B) / bid price of bidder B. Conduct a correlation analysis on the bid price spread ratio between each project bidder and other project bidders in each project to obtain the Pearson correlation coefficient between each project bidder and other project bidders. For example, if there are m projects, for bidders A and B, their spread ratios in each project are (x1, x2,..., xm) and (y1, y2,..., ym) respectively. Calculate the correlation coefficient r according to the calculation method of the Pearson correlation coefficient. Judge the strength of the correlation according to the value of r. The closer r is to 1 or (-1), the stronger the correlation; the closer it is to 0, the weaker the correlation. Preset the first correlation coefficient threshold. If the Pearson correlation coefficient between a project bidder and other project bidders is greater than the first correlation coefficient threshold, mark the project bidder and other project bidders as suspected bid rigging targets.

[0102] It should be further noted that in the specific implementation process, the process of secondary judgment on the suspected bid rigging targets includes:

[0103] Build a behavior trend prediction model based on deep learning, extract project bidding characteristics and time characteristics from the bid data of several historical projects of the suspected bid rigging targets, and obtain the bid price time series of the suspected bid rigging targets for different project bidding characteristics. Use the bid price time series of the suspected bid rigging targets for different project bidding characteristics as the training set and the test set. Input the training set into the behavior trend prediction model for training until the loss function is trained stably, and save the model parameters. Test the behavior trend prediction model through the test set until it meets the preset requirements, and output the behavior trend prediction model.

[0104] According to the behavior trend prediction model, output the predicted bid price of the suspected bid rigging targets for the current project bidding characteristics. Compare the bid price of the suspected bid rigging targets for the current project bidding characteristics with the predicted bid price to obtain the bid price deviation value. According to the bid price deviation value and the weight labels of all edges between the nodes to which the suspected bid rigging targets belong, obtain the risk loss value of the nodes to which the suspected bid rigging targets belong. Update the pre-bid scoring coefficient of the nodes according to the risk loss value.

[0105] Building a behavior trend prediction model based on deep learning is a complex process that involves multiple steps such as model selection, training, validation, and testing. The following is a detailed supplementary description of this process:

[0106] In this invention, a convolutional neural network (CNN) suitable for time series analysis is selected as the deep learning architecture, and the mean squared error loss function is chosen as the optimization objective. Subsequently, the prepared training set is input into the selected deep learning model for training. During the training process, the weights are continuously updated through the backpropagation algorithm, causing the loss function to gradually decrease until a stable state is reached. During this period, techniques such as Early Stopping are used to avoid overfitting. In addition to the basic training process, various parameters of the model are tuned through Grid Search. The parameters include the learning rate, batch size, regularization coefficient, etc.

[0107] When the model training is completed and the parameters are adjusted, the final evaluation is carried out through the test set to obtain the evaluation results of the model. The evaluation results include classification metrics such as accuracy, recall rate, F1 score, etc. According to the evaluation results on the test set, it is judged whether the model meets the expected standards. If the requirements are met, the model parameters are saved and ready for deployment; if not, it is necessary to return to a previous stage to re-examine issues such as data quality, model structure, or training strategy.

[0108] It should be further noted that in the specific implementation process, the whole-process supervision module monitors the pre-bid scoring coefficients of the bidders for each project in real time. The process of generating a bid warning signal based on the monitoring results includes:

[0109] Preset the bid warning score, and compare the pre-bid scoring coefficients of the bidders for each project with the bid warning score in real time. When the pre-bid scoring coefficient of a project bidder is less than the bid warning score, a bid warning signal is generated and fed back to the project tenderer.

[0110] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An intelligent supervision system for the entire process of bidding for engineering projects, characterized in that: It includes a project bidding data center, wherein the communication links of the project bidding data center include a client, a pre-bid supervision module, an intelligent feature analysis module, a bidding supervision module and a full-process supervision module; The client includes the project tenderer and the project bidder. The project tenderer is used to upload the project tender information to the project tender data center, and the project bidder is used to send an access request to the project tenderer through the project tender data center. The pre-bid supervision module is used to compare the project bidders and the project tenderers in multiple dimensions before bidding, and generate the pre-bid scoring coefficients of the project bidders, including: Obtain multi-dimensional data of project bidders and project bidding information of project tenderers, extract project features from the project bidding information, and obtain project bidding features; Extract various data indicators from multi-dimensional data as evaluation indicators, set the indicator weights of the evaluation indicators according to the project bidding characteristics, and obtain the membership matrix of the project bidders for the preset pre-bid scoring coefficients through fuzzy comprehensive evaluation; The indicator weights and the membership matrix of the evaluation indicators are integrated to obtain a fuzzy comprehensive evaluation matrix of the evaluation indicators, and the membership of the project bidder to different pre-bid scoring coefficients is obtained according to the fuzzy comprehensive evaluation matrix, and the pre-bid scoring coefficient with the highest membership corresponding to the project bidder is screened out, and the pre-bid scoring coefficient with the highest membership corresponding to the project bidder is used as the pre-bid scoring coefficient of the project bidder; A pre-bid scoring coefficient threshold is preset. If the pre-bid scoring coefficient is less than the pre-bid scoring coefficient threshold, the access request of the project bidder is rejected. If the pre-bid scoring coefficient is greater than or equal to the pre-bid scoring coefficient threshold, the project bidding information is sent to the project bidder. The intelligent feature analysis module is used to perform intelligent feature analysis on several historical bid winning records during the multi-dimensional comparison process before bidding, and obtain the weight coefficients of various data indicators of the project bidders for the associated feature factors; The bidding supervision module is used to build a correlation map, compare the horizontal characteristics of the bidding document data of each project bidder, analyze the bidding collusion behavior of several historical project bidding data of each project bidder, and make a secondary judgment on the suspected collusion objects; The full-process supervision module is used to monitor the pre-bid scoring coefficients of each project bidder in real time.

2. According to claim 1, a whole-process intelligent supervision system for bidding and tendering of engineering projects is characterized in that: The process of the intelligent feature analysis module performing intelligent feature analysis on several historical bid winning records includes: Extracting a number of historical bid winning records, wherein the historical bid winning records include project bidding characteristics, multi-dimensional data of the project winning party and each project bidder, extracting each characteristic factor in the project bidding characteristics, obtaining data indicators of the multi-dimensional data associated with each characteristic factor, performing cluster analysis on the number of historical bid winning records according to the project bidding characteristics, and obtaining a project bidding characteristic clustering circle; Conduct statistical analysis on each historical winning bid record in the project bidding feature clustering circle to obtain the Pearson correlation coefficient between each data indicator of the project winning bidder and each project bidder and the associated characteristic factors. At the same time, conduct difference analysis on each data indicator of the project winning bidder and each project bidder to obtain the difference coefficient between each data indicator for the associated characteristic factors. The distribution characteristics of various data indicators in the project bidding feature clustering circle are analyzed to obtain the trend distribution deviation coefficient of the associated characteristic factors between various data indicators.

3. According to claim 2, a whole-process intelligent supervision system for bidding and tendering of engineering projects is characterized in that: The process of obtaining the trend distribution deviation coefficients of the associated characteristic factors between various data indicators includes: Obtain frequency distribution histograms corresponding to the project winning party and the project bidder for each data indicator, the frequency distribution histogram uses the literal value of the data indicator as the horizontal coordinate and the frequency as the vertical coordinate, obtain the frequency of different literal values ​​of each data indicator in the frequency distribution histogram of the project winning party, mark the frequency as the first frequency, and the frequency of different literal values ​​in the frequency distribution histogram of the project bidder, mark the frequency as the second frequency, and obtain the trend distribution deviation coefficient of the associated characteristic factors between the data indicators according to the first frequency and the second frequency corresponding to the different literal values ​​of each data indicator.

4. According to claim 3, the intelligent supervision system for the whole process of bidding and tendering for engineering projects is characterized in that: The process of obtaining the weight coefficients of various data indicators for associated characteristic factors includes: The Pearson correlation coefficients between the data indicators of the project bidder and each project bidder and the associated characteristic factors, the difference coefficients between the data indicators for the associated characteristic factors, and the trend distribution deviation coefficients between the data indicators for the associated characteristic factors are standardized, and the weight coefficients of the data indicators for the associated characteristic factors are obtained based on the Pearson correlation coefficients between the data indicators of the project bidder and each project bidder and the associated characteristic factors, the difference coefficients between the data indicators for the associated characteristic factors, and the trend distribution deviation coefficients between the data indicators for the associated characteristic factors after the standardized processing.

5. According to claim 4, a whole-process intelligent supervision system for bidding and tendering of engineering projects is characterized in that: The process of constructing the association relationship map by the bidding supervision module includes: Obtain the subjects of the current project bidding process and the relationship information between each subject, wherein the subjects include the project bidders, and use each subject as a node of the relationship graph. According to the relationship information between each subject, construct different types of edges between each node, and set a weight label on each edge.

6. According to claim 5, the intelligent supervision system for the whole process of bidding and tendering for engineering projects is characterized in that: The process of the bidding supervision module comparing the horizontal characteristics of the bidding document data of each project bidder includes: Obtain the bidding document data of each project bidder in the association relationship map, perform cosine similarity comparison on the text content in the bidding document data, obtain the text similarity between each bidding document data, and preset a text similarity threshold; If the text similarity between the bidding document data is greater than the text similarity threshold, the numerical key indicators in the bidding document data are extracted, and the numerical key indicators between the bidding document data are compared and analyzed to obtain the comprehensive deviation rate between the bidding document data; A standard threshold interval corresponding to the preset comprehensive deviation rate is set, and threshold points are selected within the standard threshold interval to divide sub-intervals into different reliability levels. The reliability level corresponding to the comprehensive deviation rate between the bidding document data is obtained, and the risk loss value of the node to which the bidding document data belongs is obtained based on the reliability level and the weight labels of all edges between the nodes to which the bidding document data belongs. The pre-bid scoring coefficient of the node is updated based on the risk loss value.

7. According to claim 6, a whole-process intelligent supervision system for bidding and tendering of engineering projects is characterized in that: The process of the bidding supervision module analyzing the bidding behavior of several historical project bidding data of each project bidder includes: Extract several historical project bidding data of each project bidder in the association relationship map and preprocess them, obtain the bid price difference ratio between each project bidder and other project bidders for each project based on the preprocessed historical project bidding data, perform correlation analysis on the bid price difference ratio, obtain the Pearson correlation coefficient between each project bidder and other project bidders, preset a first correlation coefficient threshold, and if the Pearson correlation coefficient between the project bidder and other project bidders is greater than the first correlation coefficient threshold, mark the project bidder and other project bidders as suspected collusion objects.

8. According to claim 7, a whole-process intelligent supervision system for bidding and tendering of engineering projects is characterized in that: The process of conducting a secondary judgment on a suspected bid-rigging object includes: Based on deep learning, a behavioral trend prediction model is constructed to extract project bidding features and time features from several historical project bidding data of suspected bid-collusion objects, obtain the bidding price time series of the suspected bid-collusion objects for different project bidding features, and use the bidding price time series of the suspected bid-collusion objects for different project bidding features as training sets and test sets to obtain a trained behavioral trend prediction model; Output the estimated bid price of the suspected bidder for the current project bidding characteristics according to the behavioral trend prediction model, compare the bid price of the suspected bidder for the current project bidding characteristics with the estimated bid price, and obtain the bid price deviation value, and obtain the risk loss value of the node to which the suspected bidder belongs based on the bid price deviation value and the weight labels of all edges between the nodes to which the suspected bidder belongs, and update the pre-bid scoring coefficient of the node according to the risk loss value.

9. According to claim 8, a whole-process intelligent supervision system for bidding and tendering of engineering projects is characterized in that: The whole process supervision module conducts real-time monitoring of the pre-bid scoring coefficients of each project bidder, including: A preset bid warning score is used to compare the pre-bid score coefficient of each project bidder with the bid warning score in real time. When the pre-bid score coefficient of the project bidder is less than the bid warning score, a bid warning signal is generated and fed back to the project tenderer.

Images