Self-adaptive optimization method and platform for optimizing bid inviting and purchasing process

Abstract

The invention provides a self-adaptive optimization method and platform for bid invitation purchasing process optimization, and relates to the technical field of process optimization, and the method comprises the steps: employing an electronic file platform, and obtaining a plurality of electronic files of a to-be-optimized target project; extracting a plurality of main body features and a plurality of bidding information in the plurality of electronic files, and performing rationality analysis to obtain a rationality coefficient; according to the rationality coefficient, combining with a plurality of main body characteristics, performing auditor sequence self-adaptive optimization selection for automatic file auditing, and obtaining an optimized auditor sequence; and inputting a plurality of electronic files into the optimized auditor sequence, and outputting to obtain a plurality of auditing results. The technical problem that in the prior art, the electronic file auditing efficiency is low and the auditing is inaccurate in the bid inviting and purchasing process is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of process optimization, in particular to an adaptive optimization method and platform for tender procurement process optimization. BACKGROUND

[0002] In the prior art, the tender procurement process usually relies on manual review of each tender document, which is time-consuming and labor-intensive when facing a large number of documents, and is easily affected by subjective factors (such as fatigue leading to missed review). Moreover, with the advancement of government procurement and enterprise tender platform, a large number of tender processes have been transferred online, and the demand for automatic review of electronic documents has surged. Therefore, the traditional tender procurement process has the problems of low efficiency and inaccuracy in reviewing electronic documents. SUMMARY

[0003] The present application provides an adaptive optimization method and platform for tender procurement process optimization to solve the technical problems of low efficiency and inaccuracy in reviewing electronic documents in the prior art.

[0004] The technical solutions of the present application to solve the above technical problems are as follows: In a first aspect, the present application provides an adaptive optimization method for tender procurement process optimization, comprising: using an electronic document platform to obtain a plurality of electronic documents of a target project to be optimized; extracting a plurality of subject features and a plurality of bid information from the plurality of electronic documents, performing rationality analysis, and obtaining a rationality coefficient; performing adaptive optimization selection of an auditor sequence for automatic document review based on the rationality coefficient and the plurality of subject features, and obtaining an optimized auditor sequence; inputting the plurality of electronic documents into the optimized auditor sequence, and outputting a plurality of audit results.

[0005] In a second aspect, the present application provides an adaptive optimization platform for tender procurement process optimization, comprising: a file acquisition module for using an electronic document platform to obtain a plurality of electronic documents of a target project to be optimized; a rationality analysis module for extracting a plurality of subject features and a plurality of bid information from the plurality of electronic documents, performing rationality analysis, and obtaining a rationality coefficient; an auditor sequence optimization module for performing adaptive optimization selection of an auditor sequence for automatic document review based on the rationality coefficient and the plurality of subject features, and obtaining an optimized auditor sequence; an audit execution module for inputting the plurality of electronic documents into the optimized auditor sequence, and outputting a plurality of audit results.

[0006] In a third aspect, the present application provides an electronic device, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor implements the adaptive optimization method for tendering procurement process optimization as described in the first aspect by running the executable instructions.

[0007] In a fourth aspect, the present application provides a computer-readable storage medium having stored thereon computer instructions, which, when executed by a processor, implement the steps of the adaptive optimization method for tendering procurement process optimization as described in the first aspect.

[0008] The present application has the following beneficial effects: Compared with the prior art, the present application first uses an electronic file platform to obtain a plurality of electronic files of a target project to be optimized, thereby providing reliable data support for subsequent information extraction, process optimization and other operations. Secondly, a plurality of subject features and a plurality of bid information in the plurality of electronic files are extracted, a rationality analysis is performed, a rationality coefficient is obtained, the similarity between the subject features and the bid information is quantified as a distribution coefficient, and a rationality analyzer constructed in combination with historical data is used to output the rationality coefficient, thereby realizing quantitative evaluation of the risk of bid fraud and providing a decision basis for optimization of the tendering procurement process. Thirdly, according to the rationality coefficient, a sequence of auditors for automatic file auditing is adaptively optimized and selected in combination with the plurality of subject features, an optimized sequence of auditors is obtained, the number of auditors is dynamically configured according to the rationality coefficient, excessive auditing of low-risk projects is avoided, targeted auditing steps are added for high-risk projects, the auditing efficiency is effectively improved, and the accuracy of the auditing is improved through matching of the subject features with an optimal auditor combination. Finally, the plurality of electronic files are input into the optimized sequence of auditors, and a plurality of auditing results are output, thereby quickly and accurately completing the auditing of the electronic files and providing a reliable basis for tendering procurement decisions.

[0009] Through the above technical solutions, the present application receives a plurality of electronic files through an electronic file platform, automatically extracts a plurality of subject features and a plurality of bid information, calculates a rationality coefficient based on the same, quantifies the risk of abnormal events such as bid fraud, dynamically configures a sequence of auditors according to the rationality coefficient, obtains an optimized sequence of auditors, and finally outputs a plurality of auditing results through the optimized sequence of auditors. In this way, the auditing of the plurality of electronic files is completed through digital means, and the auditing efficiency and accuracy of the electronic files in the tendering procurement process are effectively improved. BRIEF DESCRIPTION OF DRAWINGS

[0010] Figure 1 A flowchart of the adaptive optimization method for tendering procurement process optimization provided by the present application; Figure 2 This is a structural diagram of the adaptive optimization platform for optimizing bidding and procurement processes provided by the present invention.

[0011] In the accompanying drawings, the components represented by the various symbols are as follows: File acquisition module 11, rationality analysis module 12, auditor sequence optimization module 13, audit execution module 14. DETAILED DESCRIPTION

[0012] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0013] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the specified features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0014] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or illustration". Any embodiment of the present invention described as "for example" is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed herein.

[0015] Example 1, as Figure 1 As shown, the embodiment of the present invention provides an adaptive optimization method for optimizing the bidding and procurement process, including: S10: using an electronic file platform to obtain multiple electronic files of the target project to be optimized; With the acceleration of the process of government procurement electronicization and the wide application of enterprise bidding platform, more than 80% of the bidding projects have realized the whole process online. Electronic bidding documents replace traditional paper documents with the advantages of convenient transmission and efficient storage, but also bring new challenges. According to statistics, the total amount of electronic files received by a single large bidding project can reach several GB, including PDF, Word, Excel and other formats, and the proportion of unstructured data is more than 60%.

[0016] To solve the above problems, the electronic file platform is used to obtain a plurality of electronic files of a target project to be optimized.

[0017] Specifically, step S10 in the method comprises: obtaining a target project to be optimized; using an electronic file platform to receive a plurality of electronic files of the target project.

[0018] In an embodiment of the application, first, a target project to be optimized is obtained. For example, the target project to be optimized is "xxxx engineering construction project".

[0019] Secondly, an electronic file platform is used to receive a plurality of electronic files of the target project. Specifically, the electronic file platform can be connected with enterprise ERP, OA system and government bidding platform through API (application program interface), receive a plurality of electronic files of the target project in the same format (such as unified PDF / Word / Excel format), such as project approval, fund proof, bidding document, etc. For example, in the electronic file platform, a plurality of electronic files of the target project (such as "xxxx engineering construction project") are received, such as bidding documents (PDF format) and bid forms (PDF format).

[0020] In summary, compared with the prior art, the electronic file platform is used to obtain a plurality of electronic files of the target project to be optimized, so that a plurality of electronic files of the target project are obtained, which provides reliable data support for subsequent information extraction, process optimization and other operations.

[0021] S20: Extracting a plurality of subject features and a plurality of bidding information in a plurality of electronic files, performing rationality analysis, and obtaining a rationality coefficient; By extracting key information of multiple electronic files, multiple subject characteristics and multiple bid information can be obtained. When the multiple subject characteristics and the multiple bid information are very similar, there is a risk of collusive bidding, i.e., multiple bidders collude privately to deliberately control the core elements of the bid document (such as price, qualification, technical solution, etc.) to achieve the purpose of illegal bidding, such as surrounding bidding and accompanying bidding. Further, when the multiple subject characteristics are highly convergent or the multiple bid information is abnormally similar, for example, the registered addresses of the bidders are concentrated in the same region (such as within a certain prefecture-level city), the qualification levels are completely consistent (such as five bidders are all "construction engineering general contractor level two"), the registered capital is similar (such as the registered capital of five bidders is 2-3 million), the enterprise equity structure is related (such as multiple bidders have common shareholders or cross-employment of senior managers), the price dispersion is extremely low, the construction period is the same, and the key parameters in the technical solution (such as equipment model, service standard) are consistent, there is a risk of collusive bidding. Therefore, by calculating the distribution coefficients of the subject characteristics and the bid information, the risk of collusive bidding can be quantified.

[0022] To solve the above problems, the application extracts multiple subject characteristics and multiple bid information in multiple electronic files, performs rationality analysis, and obtains a rationality coefficient.

[0023] Specifically, step S20 in the method includes: automatically extracting subject characteristics and bid information in the multiple electronic files to obtain multiple subject characteristics and multiple bid information; calculating subject distribution coefficients and bid distribution coefficients of the multiple subject characteristics and the multiple bid information; performing rationality analysis according to the subject distribution coefficients and the bid distribution coefficients to obtain a rationality coefficient.

[0024] In the embodiment of the application, first, the subject characteristics and the bid information in the multiple electronic files are automatically extracted to obtain multiple subject characteristics and multiple bid information. The subject characteristics include basic information of the bidders such as the enterprise name, the qualification level, and the registered capital, and the bid information includes bid information such as price data (such as total price, sub-item bid price, etc.), time elements (such as construction period, delivery period, etc.), and technical parameters (such as equipment model, service standard, etc.). For example, the subject characteristics (such as bidder registered address, qualification level, registered capital, etc.) and the bid information (such as total price, sub-item bid price, construction period, etc.) can be extracted from the electronic files by NLP, table analysis, etc.

[0025] Secondly, the subject distribution coefficient and the bidding distribution coefficient of the plurality of subject features and the plurality of bidding information are calculated, the subject distribution coefficient reflects the similarity of the plurality of bidders' basic information, the bidding distribution coefficient reflects the similarity of the plurality of bidders' bidding information, the higher the similarity, the greater the risk of collusive bidding. Further, the plurality of subject features and the plurality of bidding information mainly include two types of numerical data (such as registered capital, bid price, construction period, qualification level) and text data (such as technical scheme), for numerical data, the dispersion degree of a plurality of data can be calculated by standard deviation, and for text data, the text overlap degree can be calculated by TF-IDF combined with cosine similarity, and the plurality of distribution coefficients are added to obtain the subject distribution coefficient and the bidding distribution coefficient. Exemplarily, the registered capital of the five bidders of the xxxx engineering construction project is 2 million, 2.2 million, 2.2 million, 2 million and 2.3 million, the qualification level is all construction engineering general contractor level two, the total bid price is 500,000, 510,000, 520,000, 490,000 and 530,000, the standard deviation of the registered capital of the five bidders is 120,000, the standard deviation of the qualification level is 0, the standard deviation of the total bid price is 14,142, and the standard deviation of the registered capital and the qualification level is 120,000, the subject distribution coefficient (such as 120,000) and the bidding distribution coefficient (such as 14,142) of the project are obtained.

[0026] Finally, according to the subject distribution coefficient and the bidding distribution coefficient, a rationality analysis is performed to obtain a rationality coefficient. Specifically, based on the rationality analyzer, the subject distribution coefficient and the bidding distribution coefficient are input into the rationality analyzer to output the rationality coefficient, the rationality coefficient is based on the subject distribution information and the bidding information of the plurality of bidders, and predicts the probability of the occurrence of abnormal events such as collusive bidding, the greater the rationality coefficient, the smaller the probability of abnormality.

[0027] Further, the "according to the subject distribution coefficient and the bidding distribution coefficient, performing rationality analysis to obtain a rationality coefficient" includes: obtaining a rationality analyzer, wherein the rationality analyzer is constructed by a sample subject distribution coefficient set, a sample bidding distribution coefficient set and a sample rationality coefficient set collected in a historical period, each sample rationality coefficient is 1 minus the proportion of projects with abnormal events under different sample subject distribution coefficients and sample bidding distribution coefficients; inputting the subject distribution coefficient and the bidding distribution coefficient into the rationality analyzer to output the rationality coefficient.

[0028] In the embodiments of the present application, a rationality analyzer is first acquired, wherein the rationality analyzer is constructed by using a sample subject distribution coefficient set, a sample bidding distribution coefficient set and a sample rationality coefficient set collected in historical time, each sample rationality coefficient being 1 minus the proportion of projects with abnormal events under different sample subject distribution coefficients and sample bidding distribution coefficients, i.e., rationality coefficient = 1 - proportion of projects with abnormal events under different subject distribution coefficients and bidding distribution coefficients. Specifically, by collecting historical bidding project data, a sample set is constructed, subject distribution coefficients and bidding distribution coefficients are calculated for each project, the proportion of projects with abnormal events (such as events of bid cancellation, default, bid rigging) under multiple distribution coefficient combinations is counted, and the sample rationality coefficient is calculated, i.e., sample rationality coefficient = 1 - proportion of projects with abnormal events under different subject distribution coefficients and bidding distribution coefficients. For example, under a certain distribution coefficient combination, 20 out of 100 projects have abnormal events, and the rationality coefficient = 1 - 20 / 100 = 0.8.

[0029] Exemplarily, the rationality analyzer is a risk quantification model constructed based on historical bidding data, mainly composed of four parts of an input layer, a feature processing layer, a risk probability calculation layer and an output layer. Among them, the input layer receives the subject distribution coefficient (reflecting the similarity of multiple bidders) and the bidding distribution coefficient (reflecting the similarity of multiple bidding information) of the current project; the feature processing layer maps the double coefficients to the historical distribution space through kernel density estimation, and locates the similar project group; the risk probability calculation layer uses grouping statistics method and Bayesian smoothing technology to calculate the abnormal probability (such as the occurrence probability of bid rigging or fraud events) of the current coefficient combination; and finally the output layer outputs the rationality coefficient.

[0030] Exemplarily, the training process of the rationality analyzer can be achieved through the following technical path: 1. Data preparation: extract completed projects from the bidding history database, calculate the subject distribution coefficient and the bidding distribution coefficient of each project as the sample subject distribution coefficient set, the sample bidding distribution coefficient set, and statistically analyze the probability of abnormal events (such as bid rigging, false qualification, serious breach of contract, etc.) under different sample subject distribution coefficient sets and sample bidding distribution coefficient sets. The rationality coefficient is calculated as 1 minus the proportion of projects with abnormal events under different subject distribution coefficients and bidding distribution coefficients. A plurality of rationality coefficients are obtained as the sample rationality coefficient set. For example, under a certain distribution coefficient combination, 2 out of 10 projects are abnormal, and the rationality coefficient is 1-2 / 10=0.8. The sample subject distribution coefficient set, the sample bidding distribution coefficient set, and the sample rationality coefficient set are divided into a training set, a validation set, and a test set according to a ratio of 7:1.5:1.5; 2. Rationality analyzer construction: a lightweight hybrid architecture is adopted, combining statistical learning and interpolation prediction technology. The input layer receives the subject distribution coefficient and the bidding distribution coefficient and linearly compresses them to the [0, 1] interval to eliminate dimensional differences. The feature processing layer maps the normalized double coefficients to a 10*10 historical data grid coordinate system to detect the sample density in the grid in real time (threshold set to 30 projects). The risk probability calculation layer predicts through a double path. When the grid sample is ≥30, the rationality coefficient is directly output based on the Bayesian smoothing formula. When the sample is insufficient, the radial basis function (RBF) interpolation engine is started to integrate the rationality coefficients of the adjacent 8 grids with the current grid as the center. The output layer ensures that the result strictly falls within the [0, 1] interval through the Sigmoid function, and outputs the predicted rationality coefficient. 3. Rationality analyzer training: the model is trained through supervised training, with the subject distribution coefficient and the bidding distribution coefficient as input and the rationality coefficient as label output. Weighted mean square error (MSE) is used as the loss function, and L2 regularization and Dropout (0.3 probability) are introduced to suppress overfitting. Finally, when the MSE on the validation set does not decrease for 5 consecutive rounds and the accuracy on the test set is >90%, it is considered to be converged, the training is stopped, and the rationality analyzer is obtained.

[0031] Secondly, the subject distribution coefficient and the bidding distribution coefficient are input into the rationality analyzer to output the obtained rationality coefficient. Exemplarily, the subject distribution coefficient (such as 120000) and the bidding distribution coefficient (such as 14142) are input into the rationality analyzer to output the obtained rationality coefficient (such as 0.3). The rationality coefficient is the probability of abnormal events such as bid rigging predicted and output based on the subject distribution information and the bidding information of a plurality of bidders. The larger the rationality coefficient, the smaller the probability of abnormal events.

[0032] In summary, compared to existing technologies, this application extracts multiple subject features and multiple bid information from multiple electronic files, performs rationality analysis, and obtains rationality coefficients. By quantifying the similarity between subject features and bid information as a distribution coefficient and combining it with a rationality analyzer constructed from historical data, the rationality coefficient is output, thereby achieving a quantitative assessment of bid-rigging risk and providing a decision-making basis for optimizing the bidding and procurement process.

[0033] S30: performing adaptive optimization selection of a reviewer sequence for automatic file review based on the rationality coefficient and in combination with the plurality of subject features to obtain an optimized reviewer sequence; The traditional bidding review process relies on manual review of bid documents one by one, but this is time-consuming and labor-intensive when faced with a large number of documents, and is easily affected by subjective factors. Therefore, an auditor can be built to input multiple electronic files and quickly output multiple review results.

[0034] In response to the above problems, this application performs adaptive optimization selection of the reviewer sequence for automatic file review based on the rationality coefficient and combined with multiple subject features to obtain an optimized reviewer sequence.

[0035] Specifically, step S30 in the method includes: Obtain a set of pre-trained document reviewers, wherein each document reviewer is trained using a sample electronic document set and a sample review score set; Calculate the selection quantity based on the rationality coefficient and the quantity of all file reviewers in the file reviewer set; Randomly selecting the selected number of document reviewers from the document reviewer set to obtain a first reviewer sequence, and calculating a first optimization fitness based on a sample subject feature set of a subject corresponding to the sample electronic document set of each reviewer in the first reviewer sequence and the plurality of subject features; Continue to randomly select the selected number of file auditors to optimize the auditor sequence, and obtain the optimized auditor sequence with the maximum optimization fitness after the optimization converges.

[0036] In the embodiment of the present application, a set of pre-trained file reviewers is first obtained, where each file reviewer is trained using a set of sample electronic files and a set of sample review scores. Specifically, the set of file reviewers is composed of a combination of multiple file reviewers, each of which can independently obtain review results based on the electronic file output.

[0037] Secondly, according to the rationality coefficient, the number of all file auditors in the file auditor set is combined to calculate the selected number, wherein the selected number = [the number of all file auditors * (1-rationality coefficient)], [] is the rounding up, the greater the rationality coefficient, the lower the risk of abnormality, the fewer the number of selected file auditors, to improve efficiency; the smaller the rationality coefficient, the higher the risk of abnormality, the more the number of selected file auditors, to strengthen risk control. Exemplarily, when the number of all file auditors in the file auditor set is 5 and the rationality coefficient is 0.3, the selected number = [5 * (1-0.3)] = 4, thus, the number of called file auditors can be dynamically regulated according to the rationality coefficient, to realize high-risk strict auditing and low-risk simple auditing.

[0038] Thirdly, a first auditor sequence is obtained by randomly selecting the selected number of file auditors in the file auditor set, and a first optimization fitness is calculated according to the sample subject feature set of the subject corresponding to each sample electronic file set of the first auditor sequence and the plurality of subject features. Specifically, the first auditor sequence is obtained by randomly selecting the selected number of file auditors in the file auditor set, the plurality of sample electronic file sets in the file audit training data of all auditors in the first auditor sequence are obtained, the plurality of sample subject feature sets of the subjects corresponding to the plurality of sample electronic file sets are obtained, the similarity between the plurality of sample subject feature sets and the plurality of subject features is calculated respectively to obtain a plurality of first subject similarities, and finally the mean of the plurality of first subject similarities is calculated to obtain the first optimization fitness. The first optimization fitness reflects the matching degree of the first auditor sequence in auditing the current plurality of subject features, the greater the first optimization fitness, the higher the similarity between the training data of the first auditor sequence and the plurality of subject features, and the higher the accuracy in auditing.

[0039] Finally, continue to randomly select the selected number of file auditors, optimize the auditor sequence, and obtain the optimized auditor sequence with the maximum optimized fitness after optimization convergence. Illustratively, continue to randomly select the selected number of file auditors, and then iteratively optimize the sequence through genetic algorithm or simulated annealing algorithm, constantly improve the fitness through operations such as replacement, crossover, and the like, until the optimized auditor sequence with the maximum optimized fitness is obtained. For example, the number of all file auditors in the file auditor set is 5, the calculated selected number is 4, 4 file auditors are randomly selected in the file auditor set, and after multiple rounds of iterative optimization, the optimized auditor sequence with the maximum optimized fitness is obtained. Further, in the iterative optimization process, the overall fitness can be improved by constantly replacing auditors with low matching degree, for example, the initial sequence fitness is 0.7, by replacing 2 auditors with low matching degree, after 3 iterations, the fitness is improved to 0.92, and the optimized auditor sequence with the maximum optimized fitness is obtained.

[0040] Specifically, the "obtaining a pre-trained file auditor set" includes: According to the record data of a plurality of sample projects in a historical time, a plurality of sample electronic file sets are collected, each sample electronic file is scored and labeled according to the bidding result, and a plurality of sample scoring sets are obtained; The plurality of sample electronic file sets and the plurality of sample scoring sets are randomly divided with replacement to obtain a plurality of file audit training data; Using the plurality of file audit training data, a plurality of file auditors are trained based on machine learning to obtain a file auditor set.

[0041] In the embodiments of the present application, first, according to the record data of a plurality of sample projects in a historical time, a plurality of sample electronic file sets are collected, and each sample electronic file is scored and labeled according to the bidding result to obtain a plurality of sample scoring sets. Specifically, a plurality of sample project data in a historical time are collected, including bidding documents, bidding documents, bidder qualifications and other electronic files, and each sample electronic file is scored and labeled (such as 1-10 points) according to the bidding result (such as winning bid, bid cancellation, etc.), to obtain a plurality of sample scoring sets. The higher the scoring, the stronger the compliance. Illustratively, in a certain project bidding, the bidder is labeled as 4 points due to expired qualification, as a low compliance sample.

[0042] Secondly, the multiple sample electronic file sets and the multiple sample audit score sets are randomly divided with replacement to obtain multiple file audit training data. The random division with replacement is a statistical sampling method, and the process is as follows: from a data set D containing N samples, N times of random extraction (after each extraction, the sample is put back) are repeated to form a new training subset. Different training sets contain different repeated samples, which can enhance the model generalization ability, and when the historical data is limited, the random sampling with replacement can solve the small sample problem. Exemplarily, the 100 sample electronic file sets of the historical bidding project are randomly divided with replacement into 100 file audit training data.

[0043] Finally, multiple file auditors are trained based on machine learning using the multiple file audit training data to obtain a file auditor set. Specifically, the multiple file auditors can respectively undertake different audit tasks (such as enterprise qualification, bid price, technical scheme, etc.), so the file audit training data can be divided according to different audit tasks, and algorithms such as random forest and neural network are selected to train multiple auditors. For example, file auditor 1 can automatically identify certificate validity period abnormalities by learning 1000 sample files containing qualification information; file auditor 2 can locate unbalanced bid price risks by analyzing sub-item bid price data. In this way, by training multiple file auditors, the file auditors independently undertake different audit tasks, and finally a file auditor set covering all audit scenarios can be formed.

[0044] Further, the "obtaining a first optimization fitness according to the sample subject feature set of the subject corresponding to each auditor in the first auditor sequence and the multiple subject features" includes: Obtaining multiple sample electronic file sets in the file audit training data of all auditors in the first auditor sequence, and obtaining multiple sample subject feature sets of the subjects corresponding to the multiple sample electronic file sets; Respectively calculating the similarity of the multiple sample subject feature sets and the multiple subject features to obtain multiple first subject similarities; Calculating the mean of the multiple first subject similarities to obtain the first optimization fitness.

[0045] In the embodiments of the present application, in order to configure the optimal file auditor for different subject features, the similarity of the subject features and the subject features of the training data of all auditors in the first auditor sequence is calculated. The higher the similarity, the more familiar the file auditor is with the subject feature, and the more accurate the audit is. Specifically: Firstly, a plurality of sample electronic file sets in the file review training data of all auditors in the first auditor sequence are obtained, and a plurality of sample subject feature sets corresponding to the plurality of sample electronic file sets are obtained. Specifically, the training data of each auditor in the first auditor sequence is obtained, and a sample feature set (such as the registered address of the bidder, the qualification level, the registered capital, etc.) of the subject corresponding to the training data is obtained. For example, the electronic file set of a certain auditor includes 200 Shanghai construction engineering projects, and the sample subject features are: Shanghai + construction engineering + first-class qualification.

[0046] Secondly, the similarity of the plurality of sample subject feature sets and the plurality of subject features is calculated respectively to obtain a plurality of first subject similarities. Exemplarily, Jaccard similarity (categorical features) or cosine similarity (numerical features) can be used to calculate the similarity of the sample subject features and the current subject features respectively, for example, the current subject features are Shanghai + construction engineering + second-class qualification, and the Jaccard similarities of the plurality of sample subject feature sets are 0.6, 0.5, 0.7, and 0.6 respectively. A plurality of first subject similarities are obtained. The first subject similarity reflects the similarity degree of the sample subject features and the current subject features. The higher the first subject similarity is, the higher the similarity of the sample subject feature set training data and the current subject features is, and the higher the accuracy of the audit of the current electronic file is.

[0047] Finally, the mean of the plurality of first subject similarities is calculated to obtain a first optimization fitness. Specifically, the mean of the plurality of first subject similarities is taken as the first optimization fitness, for example, the plurality of first subject similarities are 0.6, 0.5, 0.7, and 0.6, and the first optimization fitness = (0.6 + 0.5 + 0.7 + 0.6) / 4 = 0.6. The higher the first optimization fitness is, the higher the matching degree of the first auditor sequence and the current subject features is, the higher the accuracy of the audit of the electronic file is, and the better the effect is.

[0048] In summary, compared with the prior art, the reasonable coefficient is used in combination with the plurality of subject features to perform self-adaptive optimization selection of the auditor sequence for automatic file review, and an optimized auditor sequence is obtained. In this way, the number of auditors is dynamically configured according to the reasonable coefficient, excessive review of low-risk projects is avoided, targeted review links of high-risk projects are increased, the review efficiency is effectively improved, and the accuracy of the review is improved through the optimal auditor combination of the subject features.

[0049] S40: inputting a plurality of electronic files into the optimized auditor sequence to output a plurality of audit results.

[0050] In the embodiment of the present application, a plurality of electronic files are input into the optimized auditor sequence to output a plurality of audit results. Specifically, Firstly, the input electronic files (such as bidding documents, etc.) are preprocessed according to the same method in the aforementioned step S20, the file structure is analyzed by NLP technology, the subject features and bidding information are extracted from the electronic files, the subject distribution coefficients and bidding distribution coefficients of the plurality of subject features and the plurality of bidding information are calculated and obtained, the rationality analysis is performed according to the subject distribution coefficients and the bidding distribution coefficients, and the rationality coefficient is obtained.

[0051] Secondly, the number of auditors is configured according to the rationality coefficient according to the same method in the aforementioned step S30, the first optimization fitness is calculated, and the optimized auditor sequence is obtained by iterative optimization.

[0052] Thirdly, the plurality of auditors in the optimized auditor sequence analyzes the plurality of electronic files respectively, for example, the auditor 1 audits the validity of the enterprise qualification, the auditor 2 audits the rationality of the sub-item bid, and finally outputs the plurality of audit results.

[0053] Finally, the results of the single auditors are integrated into the file-level audit report, all file results are summarized to generate the final audit result, and the final audit result is output in the form of PDF / Excel visualization.

[0054] Exemplarily, the plurality of electronic files (such as bidding documents, bid forms) are input into the optimized auditor sequence and the plurality of audit results (such as audit scores of 4, 4.5) are obtained, the audit results are scores of the plurality of electronic files, and the bidding procurement decision can be made accordingly.

[0055] In summary, compared with the prior art, the plurality of electronic files are input into the optimized auditor sequence, and the plurality of audit results are obtained, so that the electronic file audit can be quickly and accurately completed, and reliable basis is provided for the bidding procurement decision.

[0056] In summary, the embodiments of the present application have at least the following technical effects: Compared with the prior art, the present application first adopts the electronic file platform to obtain the plurality of electronic files of the target project to be optimized, so that the plurality of electronic files of the target project are obtained, and reliable data support is provided for subsequent information extraction, process optimization and other operations.

[0057] Secondly, the plurality of subject features and the plurality of bidding information in the plurality of electronic files are extracted, the rationality analysis is performed, and the rationality coefficient is obtained. In this way, the similarity between the subject features and the bidding information is quantified as a distribution coefficient, and the rationality analyzer constructed by combining historical data is used to output the rationality coefficient, thereby realizing the quantitative evaluation of the bid risk and providing a decision basis for the optimization of the bidding procurement process.

[0058] Again, the application selects the sequence of auditors for automatic file auditing according to the rationality coefficient and in combination with the plurality of subject features, to obtain an optimized sequence of auditors. In this way, the number of auditors is dynamically configured according to the rationality coefficient, excessive auditing of low-risk projects is avoided, and targeted auditing links are added for high-risk projects, effectively improving the auditing efficiency, and the accuracy of the auditing is improved by matching the optimal auditor combination with the subject features.

[0059] Finally, the application inputs a plurality of electronic files into the optimized sequence of auditors, and outputs a plurality of auditing results, so that the electronic file auditing can be quickly and accurately completed, providing a reliable basis for bidding and procurement decisions.

[0060] Through the above technical solutions, the application receives a plurality of electronic files through an electronic file platform, automatically extracts a plurality of subject features and a plurality of bidding information, and calculates a rationality coefficient accordingly, quantifying the risk of abnormal events such as bid rigging. The sequence of auditors is dynamically configured according to the rationality coefficient, an optimized sequence of auditors is obtained, and finally a plurality of auditing results are output through the optimized sequence of auditors. In this way, the auditing of a plurality of electronic files is completed through digital means, effectively improving the auditing efficiency and accuracy of electronic files in the bidding and procurement process.

[0061] Embodiment two, as shown in Figure 2 based on the same inventive concept of the adaptive optimization method for optimizing the bidding and procurement process provided in embodiment one, the application embodiment also provides an adaptive optimization platform for optimizing the bidding and procurement process, comprising: The file acquisition module 11 is configured to use an electronic file platform to acquire a plurality of electronic files of a target project to be optimized; The rationality analysis module 12 is configured to extract a plurality of subject features and a plurality of bidding information in the plurality of electronic files, perform rationality analysis, and obtain a rationality coefficient; The auditor sequence optimization module 13 is configured to select the sequence of auditors for automatic file auditing according to the rationality coefficient and in combination with the plurality of subject features, to obtain an optimized sequence of auditors; The auditing execution module 14 is configured to input a plurality of electronic files into the optimized sequence of auditors, and output a plurality of auditing results.

[0062] The file acquisition module 11 is configured to: acquire a target project to be optimized; use an electronic file platform to receive a plurality of electronic files of the target project.

[0063] The rationality analysis module 12 is configured to: automatically extracting subject features and bidding information in the plurality of electronic files to obtain a plurality of subject features and a plurality of bidding information; calculating subject distribution coefficients and bidding distribution coefficients of the plurality of subject features and the plurality of bidding information; performing rationality analysis according to the subject distribution coefficients and the bidding distribution coefficients to obtain a rationality coefficient.

[0064] Further, the "performing rationality analysis according to the subject distribution coefficients and the bidding distribution coefficients to obtain a rationality coefficient" comprises: obtaining a rationality analyzer, wherein the rationality analyzer is constructed by a sample subject distribution coefficient set, a sample bidding distribution coefficient set and a sample rationality coefficient set collected in historical time, and each sample rationality coefficient is 1 minus the proportion of projects with abnormal events under different sample subject distribution coefficients and sample bidding distribution coefficients; inputting the subject distribution coefficients and the bidding distribution coefficients into the rationality analyzer to output the rationality coefficient.

[0065] The sequence optimization module 13 is specifically configured to: obtain a pre-trained file auditor set, wherein each file auditor is trained by a sample electronic file set and a sample audit score set; obtain a selection number according to the rationality coefficient and the number of all file auditors in the file auditor set; select the selection number of file auditors in the file auditor set at random to obtain a first auditor sequence, and calculate a first optimization fitness according to the sample subject feature set of the subject corresponding to the sample electronic file set of each auditor in the first auditor sequence and the plurality of subject features; continue to select the selection number of file auditors at random to optimize the auditor sequence, and obtain an optimized auditor sequence with the maximum optimization fitness after optimization convergence.

[0066] Further, the "obtaining a pre-trained file auditor set" comprises: collecting a plurality of sample electronic file sets according to the record data of a plurality of sample projects in historical time, and obtaining a plurality of sample audit score sets by marking the audit scores of each sample electronic file according to the bidding results; randomly dividing the plurality of sample electronic file sets and the plurality of sample audit score sets with replacement to obtain a plurality of file audit training data; training a plurality of file auditors based on machine learning by using the plurality of file audit training data to obtain a file auditor set.

[0067] Further, the "calculating a first optimization fitness according to a sample subject feature set corresponding to each sample electronic file set of the first auditor sequence and the plurality of subject features", comprises: obtaining a plurality of sample electronic file sets in the file audit training data of all auditors in the first auditor sequence, and obtaining a plurality of sample subject feature sets corresponding to the plurality of sample electronic file sets; respectively calculating the similarity between the plurality of sample subject feature sets and the plurality of subject features to obtain a plurality of first subject similarities; calculating the mean of the plurality of first subject similarities to obtain the first optimization fitness.

[0068] The audit execution module 14 is specifically configured to: inputting the plurality of electronic files into the optimization auditor sequence to output a plurality of audit results.

[0069] To sum up, the embodiments of the present application have at least the following technical effects: Compared with the prior art, the present application firstly obtains a plurality of electronic files of the target project to be optimized through the file acquisition module, using an electronic file platform, which provides reliable data support for subsequent information extraction, process optimization and other operations. Secondly, through the rationality analysis module, the plurality of subject features and the plurality of bidding information in the plurality of electronic files are extracted for rationality analysis to obtain a rationality coefficient. The similarity between the subject features and the bidding information is quantified as a distribution coefficient, and a rationality analyzer is constructed by combining historical data to output the rationality coefficient, thereby realizing quantitative evaluation of the bid risk and providing a decision basis for optimization of the bidding and procurement process. Thirdly, through the auditor sequence optimization module, the optimization auditor sequence is obtained by selecting the auditor sequence adaptive optimization according to the rationality coefficient and the plurality of subject features. The number of auditors is dynamically configured according to the rationality coefficient to avoid excessive auditing of low-risk projects and increase targeted auditing links for high-risk projects, effectively improving the auditing efficiency and the accuracy of the audit through the subject feature matching optimal auditor combination. Finally, the audit execution module inputs the plurality of electronic files into the optimization auditor sequence to output a plurality of audit results, which can quickly and accurately complete the electronic file audit and provide a reliable basis for bidding and procurement decisions. In this way, the electronic file audit is completed through digital means, effectively improving the efficiency and accuracy of the electronic file audit in the bidding and procurement process.

[0070] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0071] In some embodiments, the electronic device further comprises an input interface and an output interface. The processor, the memory, the input interface and the output interface can be connected through a bus or a signal line. Each peripheral device can be connected to the input interface and the output interface through the bus, the signal line or a circuit board. The processor has a signal processing capability, and can be a microprocessor or any conventional processor; can be a general-purpose processor, a central processing unit (CPU), a network processor (NP), a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), etc. The memory can be a volatile memory or a non-volatile memory, or can include both a volatile memory and a non-volatile memory. a processor; a memory for storing processor-executable instructions; The processor implements the adaptive optimization method for tender procurement process optimization according to the first aspect by running the executable instructions.

[0072] In some embodiments, the electronic device further comprises an input interface and an output interface. The processor, the memory, the input interface and the output interface can be connected through a bus or a signal line. Each peripheral device can be connected to the input interface and the output interface through the bus, the signal line or a circuit board. The processor has a signal processing capability, and can be a microprocessor or any conventional processor; can be a general-purpose processor, a central processing unit (CPU), a network processor (NP), a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), etc. The memory can be a volatile memory or a non-volatile memory, or can include both a volatile memory and a non-volatile memory.

[0073] In some embodiments, the electronic device further comprises an input interface and an output interface. The processor, the memory, the input interface and the output interface can be connected through a bus or a signal line. Each peripheral device can be connected to the input interface and the output interface through the bus, the signal line or a circuit board. The processor has a signal processing capability, and can be a microprocessor or any conventional processor; can be a general-purpose processor, a central processing unit (CPU), a network processor (NP), a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), etc. The memory can be a volatile memory or a non-volatile memory, or can include both a volatile memory and a non-volatile memory.

[0074] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) containing computer-usable program code.

[0075] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0076] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0077] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0078] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the embodiments described and shown, and it is therefore intended that such additional variations and modifications be included within the scope of the application. Accordingly, although the present application has been described in detail with reference to particular embodiments, it should be apparent that various modifications can be made without departing from the spirit and / or scope of the application.

[0079] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the application, modifications and variations of this application can be practiced otherwise than as specifically enumerated herein.

Claims

1. An adaptive optimization method for optimizing bidding and procurement processes, characterized in that: The method comprises: Using an electronic file platform to obtain multiple electronic files of the target project to be optimized; Extract multiple subject features and multiple bidding information from multiple electronic files, conduct rationality analysis, and obtain rationality coefficients; Based on the rationality coefficient and in combination with the multiple subject features, adaptively optimizing and selecting a reviewer sequence for automatic file review to obtain an optimized reviewer sequence; Multiple electronic files are input into the optimized reviewer sequence, and multiple review results are obtained as output.

2. The adaptive optimization method for optimizing the bidding and procurement process according to claim 1, characterized in that: Use the electronic file platform to obtain multiple electronic files of the target project to be optimized, including: Obtain the target project to be optimized; An electronic file platform is used to receive multiple electronic files of the target project.

3. The adaptive optimization method for optimizing the bidding and procurement process according to claim 1, characterized in that: Extract multiple subject features and multiple bidding information from multiple electronic files, conduct rationality analysis, and obtain rationality coefficients, including: Automatically extracting subject features and bidding information from the plurality of electronic files to obtain a plurality of subject features and a plurality of bidding information; Calculating and obtaining the subject distribution coefficient and the bidding distribution coefficient of the plurality of subject characteristics and the plurality of bidding information; Based on the subject distribution coefficient and the bidding distribution coefficient, a rationality analysis is performed to obtain a rationality coefficient.

4. The adaptive optimization method for optimizing the bidding and procurement process according to claim 3 is characterized in that: Based on the subject distribution coefficient and the bidding distribution coefficient, a rationality analysis is performed to obtain a rationality coefficient, including: Obtaining a rationality analyzer, wherein the rationality analyzer is constructed using a sample subject distribution coefficient set, a sample bidding distribution coefficient set, and a sample rationality coefficient set collected from bidding data within a historical period, where each sample rationality coefficient is 1 minus the proportion of projects experiencing abnormal events under different sample subject distribution coefficients and sample bidding distribution coefficients; The subject distribution coefficient and the bidding distribution coefficient are input into the rationality analyzer, and the rationality coefficient is obtained as an output.

5. The adaptive optimization method for optimizing the bidding and procurement process according to claim 1, characterized in that: According to the rationality coefficient and in combination with the plurality of subject features, an adaptive optimization selection of an auditor sequence for automatic file audit is performed to obtain an optimized auditor sequence, including: Obtain a set of pre-trained document reviewers, wherein each document reviewer is trained using a sample electronic document set and a sample review score set; Calculate the selection quantity based on the rationality coefficient and the quantity of all file reviewers in the file reviewer set; Randomly selecting the selected number of document reviewers from the document reviewer set to obtain a first reviewer sequence, and calculating a first optimization fitness based on a sample subject feature set of a subject corresponding to the sample electronic document set of each reviewer in the first reviewer sequence and the plurality of subject features; Continue to randomly select the selected number of file auditors to optimize the auditor sequence, and obtain the optimized auditor sequence with the maximum optimization fitness after the optimization converges.

6. The adaptive optimization method for optimizing the bidding and procurement process according to claim 5, characterized in that: Get a collection of pre-trained file auditors, including: Based on the record data of multiple sample projects in the historical period, multiple sample electronic file sets are collected. Based on the bidding results, each sample electronic file is reviewed and scored to obtain multiple sample review and score sets; Performing random division with replacement on the multiple sample electronic file sets and the multiple sample review score sets to obtain multiple sets of file review training data; Using multiple document review training data, multiple document reviewers are trained based on machine learning to obtain a set of document reviewers.

7. The adaptive optimization method for optimizing the bidding and procurement process according to claim 5, characterized in that: Calculating a first optimization fitness based on a sample subject feature set of a subject corresponding to the sample electronic document set of each auditor in the first auditor sequence and the plurality of subject features includes: Acquire multiple sample electronic file sets from the file audit training data of all auditors in the first auditor sequence, and acquire multiple sample subject feature sets of subjects corresponding to the multiple sample electronic file sets; Calculating similarities between a plurality of sample subject feature sets and the plurality of subject features respectively to obtain a plurality of first subject similarities; Calculate the average of the similarities of the multiple first subjects to obtain the first optimization fitness.

8. The adaptive optimization platform for bidding and procurement process optimization is characterized by: Used to perform the method according to any one of claims 1 to 7, comprising: A file acquisition module is used to acquire multiple electronic files of the target project to be optimized using an electronic file platform; The rationality analysis module is used to extract multiple subject features and multiple bidding information in multiple electronic files, conduct rationality analysis, and obtain rationality coefficients; An auditor sequence optimization module, configured to perform adaptive optimization selection of an auditor sequence for automatic document review based on the rationality coefficient and in combination with the plurality of subject features, to obtain an optimized auditor sequence; The audit execution module is used to input multiple electronic files into the optimized auditor sequence and output multiple audit results.

9. An electronic device, characterized in that: The electronic device comprises: processor; a memory for storing processor-executable instructions; The processor implements the method according to any one of claims 1 to 7 by running the executable instructions.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images