Bidding document risk early warning method based on machine learning

Abstract

The invention provides a bidding file risk early warning method based on machine learning, and relates to the technical field of risk management, and the method comprises the steps: obtaining real-time bidding files, carrying out the file classification, obtaining a first classification file set, processing each bidding file in the first classification file set, and obtaining a second classification file set; determining a risk keyword and a corresponding risk rule based on the bidding type of the bidding file to obtain a first risk type; screening a matched machine learning algorithm based on the first risk type to obtain a first learning algorithm, and performing risk assessment on the bidding document; and analyzing the first risk type in combination with the risk assessment result to obtain a risk early warning scheme of the corresponding bidding file, thereby performing risk early warning. By classifying the bidding files, the risk type of each bidding type is determined, and the corresponding machine learning algorithm is matched to carry out risk assessment, so that the risk assessment efficiency can be improved, and risk early warning can be carried out more quickly and accurately.

Description

Technical Field

[0001] The present invention relates to the technical field of risk management, and in particular to a bidding document risk early warning method based on machine learning. Background Art

[0002] At present, in the bidding process, the quality of the bidding documents is directly related to the company's winning rate and the risk of project execution.

[0003] However, the existing bidding document review method mainly relies on manual review, which is not only time-consuming and laborious, but also easily affected by subjective factors, making it difficult to fully and accurately identify potential risks, so the risk warning performance of bidding documents is relatively weak. With the rapid development of big data and artificial intelligence technology, especially the widespread application of machine learning algorithms, new ideas and technical support have been provided for bidding document risk warning.

[0004] Therefore, the present invention provides a bidding document risk early warning method based on machine learning. Summary of the invention

[0005] The present invention provides a bidding document risk early warning method based on machine learning, which is used to solve the problems in the prior art that the review efficiency and accuracy of bidding documents are not high enough, and potential risks are also easy to be detected.

[0006] The present invention provides a bidding document risk early warning method based on machine learning, comprising:

[0007] Step 1: obtaining real-time bidding documents, and classifying documents based on the real-time bidding documents to obtain a first classified document set, and processing each bidding document in the first classified document set to obtain a second classified document set;

[0008] Step 2: Determine the risk keywords and corresponding risk rules corresponding to the second classification document subset based on the bidding type of the target bidding document to obtain the first risk type;

[0009] Step 3: Screening a matching machine learning algorithm based on the first risk type to obtain a first learning algorithm, and performing risk assessment on each bidding document in the second classification document set based on the first learning algorithm;

[0010] Step 4: Analyze the first risk type in combination with the corresponding risk assessment results to obtain a risk warning plan for the corresponding bidding document, thereby conducting corresponding risk warnings.

[0011] According to the present invention, the method of obtaining real-time bidding documents, classifying documents based on the real-time bidding documents to obtain a first classified document set, and processing each bidding document in the first classified document set to obtain a second classified document set includes:

[0012] Step 11: obtaining real-time bidding documents, and classifying the real-time bidding documents based on bidding types to obtain a first classified document set;

[0013] Step 12: Perform a file integrity check on each first classification file in the first classification file set, and remove first classification files whose integrity is lower than a preset integrity;

[0014] Step 13: Perform data cleaning and data standardization on the remaining first classification files in the first classification file set to obtain the first bidding file, thereby constructing the second classification file set.

[0015] According to the present invention, the corresponding risk keywords and corresponding risk rules are determined based on the bidding type of the target bidding document to obtain the first risk type, including:

[0016] Step 21: determining the standard risk keyword and the standard risk rule corresponding to the current bidding type based on the bidding type corresponding to each second classification document subset in the second classification document set, and obtaining the initial risk type corresponding to the second classification document subset, wherein each second classification document subset contains at least one first bidding document;

[0017] Step 22: Obtain historical bidding documents of the bidding system corresponding to the real-time bidding document, and select the first historical bidding document corresponding to the current bidding type from the historical bidding documents;

[0018] Step 23: Obtain historical risk keywords and historical risk rules of the first historical bidding document to obtain historical risk types, thereby optimizing the initial risk types to obtain the first risk type corresponding to the current second classification document subset.

[0019] According to the present invention, the historical risk keywords and historical risk rules of the first historical bidding document are obtained to obtain the historical risk type, thereby optimizing the initial risk type to obtain the first risk type corresponding to the current second classification document subset, including:

[0020] Step 231: Obtain historical risk keywords and historical risk rules of the first historical bidding document, and construct corresponding historical risk types based on the historical risk keywords and historical risk rules at the same historical moment, thereby obtaining a historical risk type set;

[0021] Step 232: performing a corresponding comparison based on each historical risk type in the historical risk type set, thereby extracting the median of the historical risk type in the historical risk type set, thereby obtaining a reference risk type;

[0022] Step 233: Compare the reference risk type with the initial risk type, and extract the sub-risk type with errors based on the comparison result;

[0023] Step 234: Optimize the sub-risk types corresponding to the initial risk types based on the difference values ​​of the sub-risk types, so as to obtain the first risk type corresponding to the current second classified file subset.

[0024] According to the machine learning algorithm based on the first risk type screening and matching provided by the present invention, a first learning algorithm is obtained, and risk assessment is performed on each bidding document in the second classification document set based on the first learning algorithm, including:

[0025] Step 31: Filter the machine learning algorithm corresponding to the first risk type to obtain an initial learning algorithm set corresponding to the current second classification file subset;

[0026] Step 32: Obtain the first label of each first bidding document in the current second classification document subset;

[0027] Step 33: determining the degree of consistency between the label type and label parameters of the first label and each initial learning algorithm in the initial learning algorithm set, thereby obtaining a first degree of consistency set;

[0028] Step 34: sorting each first degree of fit in the first degree of fit set, so as to use the initial learning algorithm with the highest first degree of fit in the first degree of fit set as the first reference algorithm, and use the initial learning algorithm with the second highest first degree of fit in the first degree of fit set as the second reference algorithm;

[0029] Step 35: Obtaining a first algorithm stability of the first reference algorithm and a second algorithm stability of the second reference algorithm;

[0030] Step 36: combining the stability of the first algorithm with the first degree of fit of the first reference algorithm to obtain a first reference value, and combining the stability of the second algorithm with the first degree of fit of the second reference algorithm to obtain a second reference value;

[0031] Step 37: Compare the first reference value and the second reference value, and use the reference algorithm corresponding to the larger reference value between the first reference value and the second reference value as the first learning algorithm of the target bidding document;

[0032] Step 38: Input the file parameters of the first bidding file into the corresponding first learning algorithm to obtain the risk assessment result of the current first bidding file, thereby obtaining the risk assessment result of each first bidding file in the second classified file set.

[0033] According to the present invention, the first reference value is obtained by combining the stability of the first algorithm with the first degree of fit of the first reference algorithm, including:

[0034] Combining the stability of the first algorithm with the first degree of fit of the first reference algorithm to obtain a first reference value T;

[0035] Among them, T is the first reference value, α 1 is the reference weight of the stability of the first reference algorithm, α 2 is the reference weight of the first consistency between the first reference algorithm and the label type and label parameters of the first label, δ 1 is the stability conversion coefficient, δ 2 is the conversion coefficient of goodness of fit, A s is the standard algorithm stability value of the first reference algorithm, A i is the historical algorithm stability value of the first reference algorithm in the target bidding system for the i-th historical processing process, μ i is the stability conversion factor of the i-th historical processing process of the first reference algorithm in the target bidding system, ρ is the error influence factor of the historical algorithm stability value, H is the first degree of fit between the label type and label parameters of the first reference algorithm and the first label, and n is the number of historical processing times of the first reference algorithm in the target bidding system.

[0036] According to the present invention, the first risk type is analyzed in combination with the corresponding risk assessment result to obtain a risk warning scheme for the corresponding bidding document, thereby performing a corresponding risk warning, including:

[0037] Step 41: extracting a risk assessment sub-database corresponding to the first risk type from the risk-assessment database based on the first risk type;

[0038] Step 42: Perform string matching on the risk assessment result of the current first bidding document and each risk assessment result in the risk assessment sub-database, so as to obtain the first risk assessment result with the highest degree of consistency with the current risk assessment result;

[0039] Step 43: Obtain the corresponding first risk level based on the first risk assessment result, and obtain the risk warning scheme of the first bidding document requiring risk warning from the risk assessment sub-database based on the first risk level, so as to issue a risk warning to the corresponding bidding document.

[0040] According to the present invention, the first risk level corresponding to the first risk assessment result is obtained, and the risk warning scheme for the first bidding document requiring risk warning is obtained from the risk assessment sub-database based on the first risk level, so as to perform risk warning on the corresponding bidding document, including:

[0041] comparing the first risk assessment result with each risk assessment range in the assessment-level database, thereby determining a first risk level of the first risk assessment result;

[0042] If the first risk level is lower than the preset standard risk level, there is no need to issue a risk warning for the first bidding document corresponding to the first risk assessment result;

[0043] Otherwise, based on the first risk level, a risk warning scheme of the corresponding level is obtained from the risk assessment sub-database, the compatibility of the risk warning scheme with the corresponding first bidding document is determined, and a risk warning is issued to the corresponding first bidding document based on the adapted risk warning scheme.

[0044] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention provides a bidding document risk warning method based on machine learning, which classifies bidding documents according to bidding types, thereby determining the risk type of each bidding type, and then matching the corresponding machine learning algorithm for risk assessment, thereby performing risk warning based on the risk assessment results, which can improve the efficiency of risk assessment and enable risk warning to be performed more quickly and accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 It is a flow chart of a bidding document risk early warning method based on machine learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] Embodiment 1:

[0049] The embodiment of the present invention provides a bidding document risk early warning method based on machine learning, such as Figure 1 As shown, including:

[0050] Step 1: obtaining real-time bidding documents, and classifying documents based on the real-time bidding documents to obtain a first classified document set, and processing each bidding document in the first classified document set to obtain a second classified document set;

[0051] Step 2: Determine the risk keywords and corresponding risk rules corresponding to the second classification document subset based on the bidding type of the target bidding document to obtain the first risk type;

[0052] Step 3: Screening a matching machine learning algorithm based on the first risk type to obtain a first learning algorithm, and performing risk assessment on each bidding document in the second classification document set based on the first learning algorithm;

[0053] Step 4: Analyze the first risk type in combination with the corresponding risk assessment results to obtain a risk warning plan for the corresponding bidding document, thereby conducting corresponding risk warnings.

[0054] In this embodiment, the real-time bidding documents are all bidding documents contained in the target bidding system.

[0055] In this embodiment, the first classification file set refers to classifying the real-time bidding files based on different bidding types, thereby obtaining the first classification file set, wherein each classification file subset in the first classification file set corresponds to the same bidding type.

[0056] In this embodiment, the second classification file set refers to a classification file set obtained by removing first classification files whose file integrity is lower than a preset integrity in the first classification file set and then performing standardization processing on the remaining files.

[0057] In this embodiment, the bidding types include goods bidding documents, engineering bidding documents, and service bidding documents, among which engineering bidding documents are further divided into types such as construction, installation, integration, and after-sales maintenance.

[0058] In this embodiment, risk keywords include technical indicator keywords, information accuracy keywords, information compliance keywords, etc.

[0059] In this embodiment, the risk rules include budget and quotation rules, compliance rules, legal rules, document arrangement rules, etc.

[0060] In this embodiment, the first risk type includes technical risk, economic risk, legal risk, market and competition risk, operational risk, information risk, etc.

[0061] In this embodiment, the first learning algorithm is a machine learning algorithm extracted from the machine learning algorithm and consistent with the current first risk type.

[0062] In this embodiment, the risk assessment result is the output result obtained after the bidding document is input into the corresponding first learning algorithm.

[0063] In this embodiment, the risk warning is to issue a risk warning to the bidding documents that need risk warning based on the risk assessment results, wherein the content of the risk warning includes the risk level and risk type of the bidding documents.

[0064] The beneficial effect of the above technical solution is: by classifying the bidding documents according to the bidding type, the risk type of each bidding type is determined, and the corresponding machine learning algorithm is matched for risk assessment, and risk warning is issued based on the risk assessment results, which can improve the efficiency of risk assessment and enable faster and more accurate risk warning.

[0065] Embodiment 2:

[0066] Based on Example 1, a second classification file set is obtained, including:

[0067] Step 11: obtaining real-time bidding documents, and classifying the real-time bidding documents based on bidding types to obtain a first classified document set;

[0068] Step 12: Perform a file integrity check on each first classification file in the first classification file set, and remove first classification files whose integrity is lower than a preset integrity;

[0069] Step 13: Perform data cleaning and data standardization on the remaining first classification files in the first classification file set to obtain the first bidding file, thereby constructing the second classification file set.

[0070] In this embodiment, the real-time bidding documents are all bidding documents contained in the target bidding system.

[0071] In this embodiment, the bidding types include goods bidding documents, engineering bidding documents, and service bidding documents, among which engineering bidding documents are further divided into types such as construction, installation, integration, and after-sales maintenance.

[0072] In this embodiment, the first classification file set refers to classifying the real-time bidding files based on different bidding types, thereby obtaining the first classification file set, wherein each classification file subset in the first classification file set corresponds to the same bidding type.

[0073] In this embodiment, the second classification file set refers to a classification file set obtained by removing first classification files whose file integrity is lower than a preset integrity in the first classification file set and then performing data cleaning and standardization on the remaining files.

[0074] The beneficial effect of the above technical solution is: by classifying the real-time bidding documents, thereby determining the risk type of each classification result and matching the corresponding machine learning algorithm for risk assessment, the screening efficiency of the machine learning algorithm can be improved, thereby improving the risk assessment efficiency.

[0075] Embodiment 3:

[0076] Based on Example 1, the first risk type is obtained, including:

[0077] Step 21: determining the standard risk keyword and the standard risk rule corresponding to the current bidding type based on the bidding type corresponding to each second classification document subset in the second classification document set, and obtaining the initial risk type corresponding to the second classification document subset, wherein each second classification document subset contains at least one first bidding document;

[0078] Step 22: Obtain historical bidding documents of the bidding system corresponding to the real-time bidding document, and select the first historical bidding document corresponding to the current bidding type from the historical bidding documents;

[0079] Step 23: Obtain historical risk keywords and historical risk rules of the first historical bidding document to obtain historical risk types, thereby optimizing the initial risk types to obtain the first risk type corresponding to the current second classification document subset.

[0080] In this embodiment, the standard risk keywords and standard risk rules are standardized contents of risk keywords and risk rules determined based on the system risk requirements of the target bidding system.

[0081] In this embodiment, the initial risk type is the risk type of technical risk, economic risk, legal risk, market and competition risk, operational risk, information risk, etc. of the target bidding system determined based on standard risk keywords and standard risk rules.

[0082] In this embodiment, the historical bidding documents refer to the bidding documents of the target bidding system in the historical bidding process.

[0083] In this embodiment, the first historical bidding document refers to a historical bidding document corresponding to the current bidding type among the historical bidding documents.

[0084] In this embodiment, the historical risk type refers to the document risk type corresponding to each first historical bidding document.

[0085] In this embodiment, the first risk type is a risk type obtained by comparing the historical risk type with the initial risk type, thereby optimizing the initial risk type based on the historical risk type.

[0086] The beneficial effect of the above technical solution is: by determining the risk type of each second classification subset, and then matching the corresponding machine learning algorithm for risk assessment, the screening efficiency of the machine learning algorithm can be improved, thereby improving the risk assessment efficiency.

[0087] Embodiment 4:

[0088] Based on Example 3, the first risk type corresponding to the current second classification file subset is obtained, including:

[0089] Step 231: Obtain historical risk keywords and historical risk rules of the first historical bidding document, and construct corresponding historical risk types based on the historical risk keywords and historical risk rules at the same historical moment, thereby obtaining a historical risk type set;

[0090] Step 232: performing a corresponding comparison based on each historical risk type in the historical risk type set, thereby extracting the median of the historical risk type in the historical risk type set, thereby obtaining a reference risk type;

[0091] Step 233: Compare the reference risk type with the initial risk type, and extract the sub-risk type with errors based on the comparison result;

[0092] Step 234: Optimize the sub-risk types corresponding to the initial risk types based on the difference values ​​of the sub-risk types, so as to obtain the first risk type corresponding to the current second classified file subset.

[0093] In this embodiment, the historical risk type set is to match the historical risk keywords and historical risk rules of the first historical bidding document according to historical moments to obtain the historical risk type of each historical moment, thereby constructing a historical risk type set based on each historical risk type.

[0094] In this embodiment, the reference risk type is obtained by making a corresponding comparison with each historical risk type in the historical risk type set, so as to determine the median of the historical risk types in the historical risk type set as the reference risk type.

[0095] In this embodiment, the difference value refers to the risk difference value between the reference risk type and each sub-risk type in the initial risk type. For example, the budget contained in the reference risk type and the budget of the quotation rule are 10.2w, and the budget contained in the initial risk type and the budget of the quotation rule are 10w, then the corresponding difference value is 0.2w.

[0096] The beneficial effect of the above technical solution is: by determining the risk type of each second classification subset, and then matching the corresponding machine learning algorithm for risk assessment, the screening efficiency of the machine learning algorithm can be improved, thereby improving the risk assessment efficiency.

[0097] Embodiment 5:

[0098] Based on Example 1, a matching machine learning algorithm is selected based on the first risk type to obtain a first learning algorithm, and a risk assessment is performed on each bidding document in the second classification document set based on the first learning algorithm, including:

[0099] Step 31: Filter the machine learning algorithm corresponding to the first risk type to obtain an initial learning algorithm set corresponding to the current second classification file subset;

[0100] Step 32: Obtain the first label of each first bidding document in the current second classification document subset;

[0101] Step 33: determining the degree of consistency between the label type and label parameters of the first label and each initial learning algorithm in the initial learning algorithm set, thereby obtaining a first degree of consistency set;

[0102] Step 34: sorting each first degree of fit in the first degree of fit set, so as to use the initial learning algorithm with the highest first degree of fit in the first degree of fit set as the first reference algorithm, and use the initial learning algorithm with the second highest first degree of fit in the first degree of fit set as the second reference algorithm;

[0103] Step 35: Obtaining a first algorithm stability of the first reference algorithm and a second algorithm stability of the second reference algorithm;

[0104] Step 36: combining the stability of the first algorithm with the first degree of fit of the first reference algorithm to obtain a first reference value, and combining the stability of the second algorithm with the first degree of fit of the second reference algorithm to obtain a second reference value;

[0105] Step 37: Compare the first reference value and the second reference value, and use the reference algorithm corresponding to the larger reference value between the first reference value and the second reference value as the first learning algorithm of the target bidding document;

[0106] Step 38: Input the file parameters of the first bidding file into the corresponding first learning algorithm to obtain the risk assessment result of the current first bidding file, thereby obtaining the risk assessment result of each first bidding file in the second classified file set.

[0107] In this embodiment, the initial learning algorithm set is a set of machine learning algorithms extracted from the machine learning algorithms that can perform risk judgment under the first risk type.

[0108] In this embodiment, the first tag refers to a tag consisting of key information of each first bidding document, wherein the first tag includes a tag type and a tag parameter.

[0109] In this embodiment, the tag type refers to the information type of key information of the first bidding document, for example, the tag type includes the bidding number, qualification certificate, commercial terms, etc.

[0110] In this embodiment, the tag parameter refers to the key information parameter of the first bidding document.

[0111] In this embodiment, the algorithm consistency refers to the consistency between each initial learning algorithm in the initial learning algorithm set and the first label of the current first bidding document, wherein the algorithm consistency is determined based on the application degree and influence degree of the label type and label parameters of the first label in the reference algorithm.

[0112] In this embodiment, the first matching degree set is a set consisting of the matching degree between each initial learning algorithm in the initial learning algorithm set and the first label of the current first bidding document.

[0113] In this embodiment, the first reference algorithm refers to sorting each first degree of fit in the first degree of fit set, thereby taking the initial learning algorithm with the highest first degree of fit in the first degree of fit set as the first reference algorithm, and the second reference algorithm refers to taking the initial learning algorithm with the second highest first degree of fit in the first degree of fit set as the second reference algorithm.

[0114] In this embodiment, the first algorithm stability refers to the assessment stability of the first reference algorithm for risk assessment, wherein the first algorithm stability is comprehensively determined based on the standard algorithm stability of the first reference algorithm and the historical algorithm stability of the first reference algorithm in the historical processing process of the target bidding system.

[0115] In this embodiment, the second algorithm stability refers to the evaluation stability of the second reference algorithm for risk assessment, wherein the second algorithm stability is comprehensively determined based on the standard algorithm stability of the second reference algorithm and the historical algorithm stability of the second reference algorithm in the historical processing process of the target bidding system.

[0116] In this embodiment, the first reference value is determined by combining the stability of the first algorithm and the first degree of fit of the first reference algorithm.

[0117] In this embodiment, the second reference value is determined by combining the stability of the second algorithm with the first degree of fit of the second reference algorithm.

[0118] In this embodiment, the first learning algorithm is a learning algorithm of the target bidding document obtained by extracting the reference algorithm corresponding to the larger reference value between the first reference value and the second reference value.

[0119] In this embodiment, the risk assessment result is the output result obtained after the bidding document is input into the corresponding first learning algorithm.

[0120] The beneficial effect of the above technical solution is: by extracting the label of each first bidding document, determining the reference algorithm in the initial learning algorithm set corresponding to the first risk type, and screening the reference algorithm in combination with the algorithm stability, thereby determining the first learning algorithm, the risk assessment efficiency of the bidding document can be faster and more accurate.

[0121] Embodiment 6:

[0122] Based on Example 5, the first reference value is obtained by combining the stability of the first algorithm with the first degree of fit of the first reference algorithm, including:

[0123] Combining the stability of the first algorithm with the first degree of fit of the first reference algorithm to obtain a first reference value T;

[0124] Among them, T is the first reference value, α 1 is the reference weight of the stability of the first reference algorithm, α 2 is the reference weight of the first consistency between the first reference algorithm and the label type and label parameters of the first label, δ 1 is the stability conversion coefficient, δ 2 is the conversion coefficient of goodness of fit, A s is the standard algorithm stability value of the first reference algorithm, A i is the historical algorithm stability value of the first reference algorithm in the target bidding system for the i-th historical processing process, μ i is the stability conversion factor of the i-th historical processing process of the first reference algorithm in the target bidding system, ρ is the error influence factor of the historical algorithm stability value, H is the first degree of fit between the label type and label parameters of the first reference algorithm and the first label, and n is the number of historical processing times of the first reference algorithm in the target bidding system.

[0125] The beneficial effect of the above technical solution is: by determining the first reference value of the first reference algorithm and comparing it with the second reference value, the determined first learning algorithm can better meet the risk assessment requirements of the current first bidding document, thereby improving the accuracy of risk assessment.

[0126] Embodiment 7:

[0127] Based on Example 1, corresponding risk warning is performed, including:

[0128] Step 41: extracting a risk assessment sub-database corresponding to the first risk type from the risk-assessment database based on the first risk type;

[0129] Step 42: Perform string matching on the risk assessment result of the current first bidding document and each risk assessment result in the risk assessment sub-database, so as to obtain the first risk assessment result with the highest degree of consistency with the current risk assessment result;

[0130] Step 43: Obtain the corresponding first risk level based on the first risk assessment result, and obtain the risk warning scheme of the first bidding document requiring risk warning from the risk assessment sub-database based on the first risk level, so as to issue a risk warning to the corresponding bidding document.

[0131] In this embodiment, the risk-assessment database includes risk types, risk levels, and risk assessment results, and the risk assessment sub-database is the risk level and risk assessment results related to the current first risk type extracted from the risk-assessment database.

[0132] In this embodiment, string matching refers to performing a corresponding comparison of corresponding strings between the risk assessment result in the risk assessment sub-database and the risk assessment result of the current first bidding document.

[0133] In this embodiment, the first risk assessment result refers to the risk assessment result in the risk assessment sub-database that has the highest degree of consistency with the risk assessment result of the current first bidding document.

[0134] In this embodiment, the first risk level refers to the risk level in the risk assessment sub-database corresponding to the first risk assessment result.

[0135] The beneficial effect of the above technical solution is: by analyzing the first risk type in combination with the corresponding risk assessment result, the risk level of the first risk assessment result is determined, and risk warnings are issued for bidding documents with high risk levels, which can make risk warnings more accurate and effective.

[0136] Embodiment 8:

[0137] Based on Example 7, risk warning is performed on the corresponding bidding documents, including:

[0138] comparing the first risk assessment result with each risk assessment range in the assessment-level database, thereby determining a first risk level of the first risk assessment result;

[0139] If the first risk level is lower than the preset standard risk level, there is no need to issue a risk warning for the first bidding document corresponding to the first risk assessment result;

[0140] Otherwise, based on the first risk level, a risk warning scheme of the corresponding level is obtained from the risk assessment sub-database, the compatibility of the risk warning scheme with the corresponding first bidding document is determined, and a risk warning is issued to the corresponding first bidding document based on the adapted risk warning scheme.

[0141] In this embodiment, the standard risk level is determined based on the security performance of the target bidding system. The higher the security performance of the target bidding system, the smaller the corresponding standard risk level.

[0142] In this embodiment, adaptation means that the risk warning scheme is used to correspond to whether there is a warning deviation in the first bidding document or whether the warning type exists in the first bidding document. If so, it is adapted; if not, it is not adapted.

[0143] The beneficial effect of the above technical solution is: by judging the risk level of the first risk assessment result, risk warnings can be issued for bidding documents that require risk warnings, making risk warnings more accurate and effective.

[0144] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A bidding document risk early warning method based on machine learning, characterized in that: include: Step 1: obtaining real-time bidding documents, and classifying documents based on the real-time bidding documents to obtain a first classified document set, and processing each bidding document in the first classified document set to obtain a second classified document set; Step 2: Determine corresponding risk keywords and corresponding risk rules based on the bidding type of the target bidding document to obtain a first risk type; Step 3: Screening a matching machine learning algorithm based on the first risk type to obtain a first learning algorithm, and performing risk assessment on each bidding document in the second classification document set based on the first learning algorithm; Step 4: Analyze the first risk type in combination with the corresponding risk assessment results to obtain a risk warning plan for the corresponding bidding document, thereby conducting corresponding risk warnings.

2. The method for early warning of bidding document risks based on machine learning according to claim 1, characterized in that: Acquiring real-time bidding documents, classifying documents based on the real-time bidding documents to obtain a first classified document set, and processing each bidding document in the first classified document set to obtain a second classified document set, including: Step 11: obtaining real-time bidding documents, and classifying the real-time bidding documents based on bidding types to obtain a first classified document set; Step 12: Perform a file integrity check on each first classification file in the first classification file set, and remove first classification files whose integrity is lower than a preset integrity; Step 13: Perform data cleaning and data standardization on the remaining first classification files in the first classification file set to obtain the first bidding file, thereby constructing the second classification file set.

3. The bidding document risk early warning method based on machine learning according to claim 1 is characterized in that: Determine corresponding risk keywords and corresponding risk rules based on the bidding type of the target bidding document to obtain a first risk type, including: Step 21: determining the standard risk keyword and the standard risk rule corresponding to the current bidding type based on the bidding type corresponding to each second classification document subset in the second classification document set, and obtaining the initial risk type corresponding to the second classification document subset, wherein each second classification document subset contains at least one first bidding document; Step 22: Obtain historical bidding documents of the bidding system corresponding to the real-time bidding document, and select the first historical bidding document corresponding to the current bidding type from the historical bidding documents; Step 23: Obtain historical risk keywords and historical risk rules of the first historical bidding document to obtain historical risk types, thereby optimizing the initial risk types to obtain the first risk type corresponding to the current second classification document subset.

4. The method for early warning of bidding document risks based on machine learning according to claim 3 is characterized in that: The historical risk keywords and historical risk rules of the first historical bidding document are obtained to obtain the historical risk type, thereby optimizing the initial risk type to obtain the first risk type corresponding to the current second classification document subset, including: Step 231: Obtain historical risk keywords and historical risk rules of the first historical bidding document, and construct corresponding historical risk types based on the historical risk keywords and historical risk rules at the same historical moment, thereby obtaining a historical risk type set; Step 232: performing a corresponding comparison based on each historical risk type in the historical risk type set, thereby extracting the median of the historical risk type in the historical risk type set, thereby obtaining a reference risk type; Step 233: Compare the reference risk type with the initial risk type, and extract the sub-risk type with errors based on the comparison result; Step 234: Optimize the sub-risk types corresponding to the initial risk types based on the difference values ​​of the sub-risk types, so as to obtain the first risk type corresponding to the current second classified file subset.

5. The bidding document risk early warning method based on machine learning according to claim 1 is characterized in that: The matching machine learning algorithm is selected based on the first risk type to obtain a first learning algorithm, and a risk assessment is performed on each bidding document in the second classification document set based on the first learning algorithm, including: Step 31: Filter the machine learning algorithm corresponding to the first risk type to obtain an initial learning algorithm set corresponding to the current second classification file subset; Step 32: Obtain the first label of each first bidding document in the current second classification document subset; Step 33: determining the degree of consistency between the label type and label parameters of the first label and each initial learning algorithm in the initial learning algorithm set, thereby obtaining a first degree of consistency set; Step 34: sorting each first degree of fit in the first degree of fit set, so as to use the initial learning algorithm with the highest first degree of fit in the first degree of fit set as the first reference algorithm, and use the initial learning algorithm with the second highest first degree of fit in the first degree of fit set as the second reference algorithm; Step 35: Obtaining a first algorithm stability of the first reference algorithm and a second algorithm stability of the second reference algorithm; Step 36: combining the stability of the first algorithm with the first degree of fit of the first reference algorithm to obtain a first reference value, and combining the stability of the second algorithm with the first degree of fit of the second reference algorithm to obtain a second reference value; Step 37: Compare the first reference value and the second reference value, and use the reference algorithm corresponding to the larger reference value between the first reference value and the second reference value as the first learning algorithm of the target bidding document; Step 38: Input the file parameters of the first bidding file into the corresponding first learning algorithm to obtain the risk assessment result of the current first bidding file, thereby obtaining the risk assessment result of each first bidding file in the second classified file set.

6. The method for early warning of bidding document risks based on machine learning according to claim 5, characterized in that: Combining the stability of the first algorithm with the first degree of agreement of the first reference algorithm to obtain a first reference value includes: Combining the stability of the first algorithm with the first degree of fit of the first reference algorithm to obtain a first reference value T; Wherein, T is the first reference value, α1 is the reference weight of the stability of the first reference algorithm, α2 is the reference weight of the first consistency of the label type and label parameters of the first reference algorithm and the first label, δ1 is the stability conversion coefficient, δ2 is the consistency conversion coefficient, and A s is the standard algorithm stability value of the first reference algorithm, A i is the historical algorithm stability value of the first reference algorithm in the target bidding system for the i-th historical processing process, μ i is the stability conversion factor of the i-th historical processing process of the first reference algorithm in the target bidding system, ρ is the error influence factor of the historical algorithm stability value, H is the first degree of fit between the label type and label parameters of the first reference algorithm and the first label, and n is the number of historical processing times of the first reference algorithm in the target bidding system.

7. The bidding document risk early warning method based on machine learning according to claim 1 is characterized in that: Analyze the first risk type in combination with the corresponding risk assessment results to obtain a risk warning plan for the corresponding bidding document, thereby performing corresponding risk warnings, including: Step 41: extracting a risk assessment sub-database corresponding to the first risk type from the risk-assessment database based on the first risk type; Step 42: Perform string matching on the risk assessment result of the current first bidding document and each risk assessment result in the risk assessment sub-database, so as to obtain the first risk assessment result with the highest degree of consistency with the current risk assessment result; Step 43: Obtain the corresponding first risk level based on the first risk assessment result, and obtain the risk warning scheme of the first bidding document requiring risk warning from the risk assessment sub-database based on the first risk level, so as to issue a risk warning to the corresponding bidding document.

8. The method for early warning of bidding document risks based on machine learning according to claim 7 is characterized in that: Obtaining a corresponding first risk level based on the first risk assessment result, and obtaining a risk warning scheme for a first bidding document requiring risk warning from a risk assessment sub-database based on the first risk level, thereby performing risk warning on the corresponding bidding document, including: comparing the first risk assessment result with each risk assessment range in the assessment-level database, thereby determining a first risk level of the first risk assessment result; If the first risk level is lower than the preset standard risk level, there is no need to issue a risk warning for the first bidding document corresponding to the first risk assessment result; Otherwise, based on the first risk level, a risk warning scheme of the corresponding level is obtained from the risk assessment sub-database, the compatibility of the risk warning scheme with the corresponding first bidding document is determined, and a risk warning is issued to the corresponding first bidding document based on the adapted risk warning scheme.

Images