Big data-based bidding risk assessment method and system

Abstract

The application belongs to the technical field of bidding risk assessment, and specifically provides a bidding risk assessment method and system based on big data, which mainly comprises the following steps: collecting basic identification data of bidders, historical bidding record data, real-time behavior log data and external environment index data, and generating an integrated initial data set through multi-source data integration processing; performing data cleaning and standardization processing on the integrated initial data set to obtain purified standard data; performing risk feature extraction on the purified standard data to form feature risk indexes; and performing risk linkage analysis by using the feature risk indexes to construct a comprehensive risk profile. The application can realize efficient integration and in-depth analysis of multi-dimensional data of bidders, accurately identify risk abnormalities, highlight key risk signals, reduce the probability of decision-making misjudgment, and provide intuitive and visual risk assessment results to assist decision-making.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of bidding risk assessment, and particularly relates to a bidding risk assessment method and system based on big data. BACKGROUND

[0002] In bidding risk assessment, current work is mostly carried out by combining artificial auditing with traditional statistical models. Artificial auditing focuses on static data such as the qualification documents and historical performance of bidders, and traditional statistical models select fixed-dimension data such as financial indicators and performance records to complete risk judgment through preset thresholds or simple weighting. Some methods may include industry level, policies and regulations and other external data to assist in assessment.

[0003] The current assessment method is limited by data integration and analysis mode, and it is difficult to realize real-time correlation processing of dynamic changing data, which leads to difficulty in accurately distinguishing normal business fluctuations from potential risk abnormalities in the analysis of bidder behavior, especially when multiple types of data change simultaneously. Isolated index analysis cannot capture the influence between factors, so that key risk signals are easily covered by regular data fluctuations, ultimately increasing the probability of misjudgment in the decision-making process. SUMMARY

[0004] The application provides a bidding risk assessment method and system based on big data, which effectively solves the problems of insufficient real-time correlation of dynamic data, difficulty in distinguishing normal fluctuations from risk abnormalities, uncaught influence of multiple data linkage, and high risk signal coverage, high decision-making error rate in the prior art, realizes efficient integration and deep analysis of multi-dimensional data of bidders, accurately identifies risk abnormalities, highlights key risk signals, reduces the probability of decision-making errors, and provides intuitive and visual risk assessment results to assist decision-making.

[0005] To achieve the above purpose, the application adopts the following technical solutions:

[0006] In a first aspect, the application provides a bidding risk assessment method based on big data, comprising:

[0007] Collecting basic identification data of bidders, historical bidding record data, real-time behavior log data and external environment indicator data, and generating an integrated initial data set through multi-source data integration processing.

[0008] Performing data cleaning and standardization processing on the integrated initial data set to obtain purified standard data.

[0009] Performing risk feature extraction on the purified standard data to form feature risk indicators.

[0010] Performing risk linkage analysis using the feature risk indicators to construct a comprehensive risk profile.

[0011] A risk quantification evaluation is performed on the comprehensive risk profile to generate a normalized risk value.

[0012] The normalized risk value is outputted and a response report is generated to produce a visual response report.

[0013] Further, basic identification data of the bidding party, historical bidding record data, real-time behavior log data, and external environment index data are collected, and an integrated initial data set is generated through multi-source data integration processing, including:

[0014] The basic identification data of the bidding party is obtained, and unique verification processing is performed to generate valid identification data.

[0015] The historical bidding record data is called based on the valid identification data, and a missing value completion processing is performed to output complete historical records.

[0016] Real-time behavior log data is collected based on the complete historical records, and time axis calibration processing is performed to form time series behavior data.

[0017] Based on the time series behavior data, external environment index data is obtained, and unit normalization processing is performed to obtain the integrated initial data set.

[0018] Further, data cleaning and standardization processing is performed on the integrated initial data set to obtain purified standard data, including:

[0019] For the bidding party financial attribute data of the integrated initial data set, abnormal data filtering processing is performed to obtain screened financial data.

[0020] The behavior operation data in the screened financial data is subjected to pattern feature filtering processing to generate purified behavior data.

[0021] The social reputation data in the purified behavior data is extracted for credibility classification processing to output a credible reputation index.

[0022] The legal compliance data of the credible reputation index is subjected to specification consistency verification to finally generate the purified standard data.

[0023] Further, risk feature extraction is performed on the purified standard data to form a feature risk index, including:

[0024] The historical failure case data in the purified standard data is extracted for feature vector extraction to obtain a failure feature set.

[0025] Real-time abnormal signals in the failure feature set are subjected to mutation point positioning processing to generate an abnormal event identifier.

[0026] According to the abnormal event identification, trend intensity quantification is performed on economic fluctuation data to form an economic risk trend.

[0027] According to the economic risk trend, competitive pressure assessment is performed on competitive bidding data to obtain the characteristic risk indicator.

[0028] Further, the characteristic risk indicator is used to perform risk linkage analysis to construct a comprehensive risk profile, including:

[0029] Multi-factor correlation analysis is performed on behavior abnormal data in the characteristic risk indicator to construct a correlation risk graph.

[0030] From the correlation risk graph, a collaborative correlation rule is extracted, and a threshold trigger monitoring operation is performed in combination with an environmental mutation indicator to generate an environmental abnormal signal.

[0031] Based on the environmental abnormal signal, a rule deviation measurement operation is performed on the collaborative correlation rule to output a linkage error value.

[0032] The linkage error value is used to perform a topological structure reorganization process on the correlation risk graph to form a reconstructed correlation model.

[0033] The reconstructed correlation model and the characteristic risk indicator are subjected to data logic fitting verification operation to output a verification pass signal.

[0034] Based on the verification pass signal, a multi-source dynamic collaboration operation is performed on external environment indicator data, time series behavior data, and a failure feature set to generate the comprehensive risk profile.

[0035] Further, based on the verification pass signal, a multi-source dynamic collaboration operation is performed on external environment indicator data, time series behavior data, and a failure feature set to generate the comprehensive risk profile, including:

[0036] According to the verification pass signal, a dynamic parameter adaptation operation is performed on external environment indicator data to generate an adaptive risk model.

[0037] The adaptive risk model is subjected to multi-mode fusion with time series behavior data to form a fusion risk mode.

[0038] Based on the fusion risk mode, a backtracking data integration operation is performed on historical dimension data in the failure feature set to finally construct the comprehensive risk profile.

[0039] Further, risk quantification evaluation is performed on the comprehensive risk profile to generate a normalized risk value, including:

[0040] For potential loss data of the comprehensive risk profile, a probability interval calculation process is performed to generate a loss probability distribution.

[0041] Based on the loss probability distribution, a multi-dimensional space mapping operation is performed in combination with competitive pressure data in the feature risk indicator to form a risk influence matrix.

[0042] Using the risk influence matrix, an effect aggregation analysis operation is performed on the control strategy features in the fusion risk mode, and a risk mitigation factor is output.

[0043] The risk mitigation factor is numerically normalized with the quantitative evaluation features in the feature risk indicator and input into a risk scoring model to generate the normalized risk value.

[0044] Further, the normalized risk value is subjected to output and response generation operation to produce a visual response report, including:

[0045] According to the normalized risk value, a warning template data is called to implement a template dynamic filling process to generate a risk warning report.

[0046] The preset warning template data is called, and a template dynamic filling operation is performed based on the normalized risk value to generate a risk warning report.

[0047] A set of preset logical rules is extracted, and a rule logic matching operation is performed based on the risk warning report to form a recommended action plan.

[0048] Historical audit data is obtained, and an audit trajectory iteration operation is performed based on the recommended action plan to output an updated audit log.

[0049] An interactive template of a user interface configuration library is loaded, and a visual format conversion operation is performed based on the updated audit log to finally obtain the visual response report.

[0050] Further, a data logic fitting verification operation is performed on the reconstructed association model and the feature risk indicator, including:

[0051] The node relationship of the reconstructed association model is mapped to the data structure of the feature risk indicator, and a conflict logic detection operation is performed to output a conflict node identifier.

[0052] Based on the conflict node identifier, in combination with a historical verification rule library of risk linkage analysis, a conflict level grading operation is performed to generate a conflict level label.

[0053] According to the conflict level label, a topology structure optimization operation is performed on the reconstructed association model to output a verification pass signal.

[0054] In a second aspect, the present application provides a big data-based bidding risk assessment system, which comprises:

[0055] The data collection integration module collects basic identification data of the bidding party, historical bidding record data, real-time behavior log data and external environment index data, and generates an integrated initial data set through multi-source data integration processing.

[0056] The data cleaning and standardization module performs data cleaning and standardization processing on the integrated initial data set to obtain standardized data after cleaning.

[0057] The risk feature extraction module extracts risk features from the standardized data after cleaning to form feature risk indicators.

[0058] The risk linkage analysis module performs risk linkage analysis using the feature risk indicators to construct a comprehensive risk profile.

[0059] The risk quantification evaluation module performs risk quantification evaluation on the comprehensive risk profile to generate a normalized risk value.

[0060] The response output report module outputs and generates a visual response report based on the normalized risk value.

[0061] In a third aspect, the present application provides a big data-based bidding risk assessment device, which comprises a memory and a processor; the memory is used to store a computer program; and the processor is used to execute the computer program to realize the steps of the big data-based bidding risk assessment method according to the first aspect.

[0062] In a fourth aspect, the present application provides a storage medium, which stores computer program instructions; when the computer program instructions are read and run by a processor, the steps of the big data-based bidding risk assessment method according to the first aspect are executed.

[0063] The present application has the following advantages:

[0064] The present application realizes full-process data processing and risk assessment by using the scheme of multi-source data integration, cleaning and standardization, risk feature extraction, linkage analysis to construct a profile, quantification evaluation to generate a normalized risk value and visual output, effectively solves the problems of insufficient real-time correlation of dynamic data, difficulty in distinguishing normal fluctuations from risk abnormalities, failure to capture the influence of multiple data linkage, resulting in risk signals being masked, high decision-making error rate, realizes efficient integration and deep analysis of multi-dimensional data of the bidding party, accurately identifies risk abnormalities, highlights key risk signals, reduces the probability of decision-making errors, and provides intuitive and visual risk assessment results to assist decision-making.

[0065] Other features and advantages of the present application will be set forth in the description that follows, and in part will be apparent from the description, or can be learned by practice of the application. The purposes and other advantages of the present application will be realized and attained by the structure particularly pointed out in the description and appended claims. BRIEF DESCRIPTION OF DRAWINGS

[0066] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0067] Figure 1 The flowchart of the big data-based bidding risk assessment method of the present application is shown.

[0068] Figure 2 The module diagram of the big data-based bidding risk assessment system of the present application is shown. DETAILED DESCRIPTION

[0069] In order to solve the problems raised in the background art, the present application adopts the scheme of multi-source data integration, cleaning standardization, risk feature extraction, linkage analysis contour construction, quantitative evaluation normalized risk value generation and visual output, to realize the whole-process data processing and risk assessment.

[0070] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0071] In some embodiments, as shown in Figure 1 The present application provides a big data-based bidding risk assessment method, which comprises:

[0072] S1. Collecting bidding party basic identification data, historical bidding record data, real-time behavior log data and external environment index data, and generating an integrated initial data set through multi-source data integration processing.

[0073] S2. Data cleaning and standardization processing is performed on the integrated initial data set to obtain purified standard data.

[0074] S3. Risk feature extraction is performed on the purified standard data to form feature risk indicators.

[0075] S4. Perform risk linkage analysis using characteristic risk indicators to construct a comprehensive risk profile.

[0076] S5. Perform risk quantification assessment on the comprehensive risk profile to generate normalized risk values.

[0077] S6. Implement output and response generation operations on the normalized risk values to produce a visual response report.

[0078] In some embodiments, the basic identification data in S1 represents the basic information that uniquely identifies the bidding subject, including enterprise name, unified social credit code, legal representative, registered address, etc.

[0079] The historical bidding record data represents the structured record of the bidding party's past participation in bidding projects, including project name, bidding time, bid result, and bid cancellation reason.

[0080] The real-time behavior log data represents the dynamic behavior flow in the bidding operation process, including file modification timestamp, access IP address, and operation type.

[0081] The external environment indicator data represents the quantitative value of external factors that affect bidding risk, including the number of policy and regulation changes, raw material price volatility, and the number of competitors.

[0082] Collect the bidding party's basic identification data, historical bidding record data, real-time behavior log data, and external environment indicator data, and generate an integrated initial data set through multi-source data integration processing, including:

[0083] S11. Obtain the bidding party's basic identification data and implement unique verification processing to generate valid identification data.

[0084] Use database primary key constraint technology to implement hash value comparison on the unified social credit code field. If the hash value is repeated, trigger the manual review process, and finally output the unique enterprise identification data, marked as valid identification data.

[0085] For example, when inputting three enterprises with the same credit code, only the first record is retained and marked as valid identification data.

[0086] S12. Call the historical bidding record data based on the valid identification data and perform missing value completion processing to output complete historical records.

[0087] After identifying the null fields of the valid identification data, fill in the historical bidding average value of similar enterprises in the same industry and size, and finally output the complete historical records.

[0088] For example, if the "bid amount" field of an enterprise is missing, fill in the median of the bid amount of similar enterprises in the same industry in the past three years.

[0089] S13. Collect real-time behavior log data according to complete historical records, perform time axis calibration processing to form time series behavior data.

[0090] All operation events in the complete historical records are unified by time axis calibration processing, and time series behavior data is formed.

[0091] Exemplarily, the operation record of Shanghai user at 09:00:00 is calibrated to UTC time 01:00:00.

[0092] S14. Obtain external environment index data based on time series behavior data, and obtain integrated initial data set by unit standardization processing.

[0093] External environment index data such as the number of policy and regulation changes and raw material price index is obtained by using time series behavior data, and data compatibility is realized by unit standardization processing, and integrated initial data set is obtained.

[0094] In some embodiments, the data cleaning and standardization processing of the integrated initial data set in S2 obtains the purified standard data, including:

[0095] S21. For the bidder financial attribute data of the integrated initial data set, perform abnormal data filtering processing to obtain screened financial data.

[0096] The principle of three standard deviations is used to identify outliers, that is, the mean and standard deviation of each financial indicator are calculated, and data points outside the range of mean plus or minus three standard deviations are removed to generate screened financial data.

[0097] Exemplarily, if the industry average asset-liability ratio is sixty percent, and the standard deviation is ten percent, data below thirty percent or above ninety percent is removed.

[0098] S22. Implement mode feature filtering processing on the behavior operation data in the screened financial data to generate purified behavior data.

[0099] High-frequency mechanical operation modes in the screened financial data are identified, such as more than ten operations per second, which are determined as machine crawler behavior and removed to generate purified behavior data.

[0100] Exemplarily, a certain account submits fifteen times of bid file modification requests per second, which is identified as non-human operation and filtered.

[0101] S23. Extract social reputation data in the purified behavior data for credibility classification processing, and output credible reputation indicators.

[0102] Specifically, social reputation data is extracted from the purification behavior data, and weights are assigned according to the authority of the data source, such as government official data weight 0.9, third-party platform data weight 0.7, and the comprehensive reputation score is calculated by weighting, and the reliable reputation index is output.

[0103] For example, the penalty announcement score of a certain enterprise official website is 80 points, and the commercial public opinion report score is 60 points, and the comprehensive reputation index is: 80 points x 0.9 + 60 points x 0.7 = 108 points.

[0104] S24. Perform specification consistency verification on the legal compliance data of the reliable reputation index, and finally generate the purified standard data.

[0105] The specific operation of performing specification consistency verification is: comparing the data of the national enterprise credit information public system, verifying the validity of the enterprise state and qualification, and generating purified standard data after passing.

[0106] For example, the business license of a certain enterprise shows "survival", but the qualification certificate is expired, and the verification is not passed and marked as a risk.

[0107] In some embodiments, the risk feature extraction of the purified standard data in S3 forms a feature risk index, including:

[0108] S31. Extract historical failure case data from the purified standard data, perform feature vector extraction, and obtain a failure feature set.

[0109] Extract historical failure case data from the purified standard data, such as reasons for cancellation of bidding and number of times of cancellation of bidding, and convert it into a structured feature set through feature vector extraction processing.

[0110] Specifically, principal component analysis technology can be used for dimensionality reduction processing, which maps multi-dimensional case attributes (such as technical scheme defects and bid price deviation) to low-dimensional feature vectors to generate a failure feature set.

[0111] For example, in the historical failure case of a certain enterprise, "technical scheme not up to standard" accounts for 70%, and "bid price exceeds budget" accounts for 30%, and after vectorization, the feature value [0.7, 0.3] is output.

[0112] S32. Perform mutation point positioning processing on real-time abnormal signals in the failure feature set to generate an abnormal event identifier.

[0113] Specifically, the cumulative sum control chart algorithm can be applied, and when the signal change amplitude in the failure feature set exceeds the preset threshold, it is marked as a mutation point, and an abnormal event identifier is generated.

[0114] For example, a bidder's bid price drops from 1 million to 800,000 (a drop of 20%) in the last hour before the deadline, triggering an abnormal identifier.

[0115] S33. Perform trend intensity quantification processing on the economic fluctuation data according to the abnormal event identification, and form an economic risk trend.

[0116] Specifically, the fluctuation acceleration index can be calculated, and the fluctuation acceleration index is marked as the economic risk trend. The reference formula is: ; wherein, represents the fluctuation acceleration, represents the price fluctuation rate of the current monitoring period, represents the price fluctuation rate of the last monitoring period, represents the time interval of the monitoring period.

[0117] For example, the steel price fluctuation rate last month is 5%, the fluctuation rate this month is 8%, and the time interval is 1 month, then the fluctuation acceleration is: (8%-5%) ÷ 1 = 3%.

[0118] S34. Perform competitive pressure assessment processing on the competitive bidding data using the economic risk trend, and obtain a characteristic risk index.

[0119] Specifically, the fluctuation acceleration index can be calculated, and the fluctuation acceleration index is marked as the characteristic risk index. The reference formula is: ; wherein, represents the competitive pressure value, represents the actual number of participating bidding enterprises, represents the average number of bidding enterprises in the industry, represents the price sensitivity coefficient, which can be obtained by historical data regression.

[0120] In some embodiments, the use of the characteristic risk index in S4 to perform risk linkage analysis and construct a comprehensive risk profile includes:

[0121] S41. Perform multi-factor correlation analysis on the behavior abnormal data in the characteristic risk index, and construct a correlation risk graph.

[0122] Using graph neural network technology, the bidding party is taken as a node, and the logical relationship between abnormal behaviors is taken as an edge, to construct a correlation risk graph.

[0123] The edge weight in the correlation risk graph represents the correlation strength, which is calculated by the historical behavior co-occurrence frequency.

[0124] For example, a bidding party continuously modifies the bid price 5 times and withdraws the bid within 10 minutes, forming a “frequent modification withdrawal” correlation edge with a weight of 0.9, which is a high-frequency event.

[0125] S42. Extract the collaborative correlation rules from the correlation risk graph, and perform threshold trigger monitoring operation combined with the environmental mutation index to generate an environmental abnormal signal.

[0126] Identify node pairs connected by high-weight edges in the association risk graph, such as weight ≥ 0.7, convert the node pairs into logical rules, and obtain the collaborative association rules.

[0127] Extract collaborative association rules from the association risk graph, such as the "policy change Abnormal fluctuations in the offer" rule, and monitor environmental mutation indicators, such as the number of policy changes, and when the environmental mutation indicator exceeds the threshold, such as the number of policy changes ≥ 2, generate an environmental anomaly signal.

[0128] For example, 3 new bidding regulations are issued on the same day, triggering an environmental anomaly signal.

[0129] S43. Based on the environmental anomaly signal, perform rule bias measurement operation on the collaborative association rules, and output the linkage error value.

[0130] Calculate the bias value of the current collaborative association rule and the historical benchmark rule, using the formula: ; Where, represents the linkage error value, represents the weight of the current rule, represents the weight of the historical benchmark rule, represents the environmental factor influence coefficient, which can be determined by industry expert evaluation, and n represents the number of independent dimensions for weight difference comparison between the current collaborative association rule and the historical benchmark rule.

[0131] For example, when the policy changes Take 1.2 to enhance the impact, and in the regular period Take 0.8.

[0132] S44. Use the linkage error value to perform topological structure reorganization processing on the association risk graph, forming a reconstructed association model.

[0133] Adjust the structure of the association risk graph according to the linkage error value, if E is greater than the preset linkage error threshold, remove the changes in weight that exceed the preset weight change threshold, such as , remove the edges whose weight changes exceed ±40%, and add new association edges derived from environmental anomalies, such as the "raw material price Abandoned bid" edge.

[0134] The linkage error threshold can be determined by model performance inflection point analysis method, and the weight change threshold can be determined based on the model stability optimization principle.

[0135] S45. Perform data logic fitting verification operation on the reconstructed association model and the characteristic risk indicators, and output the verification pass signal.

[0136] S46. Based on the verification pass signal, generate a comprehensive risk profile by performing multi-source dynamic collaborative operation on external environment indicator data, timing behavior data, and failure feature set.

[0137] In some embodiments, the data logic fitting verification operation in S45 between the reconstructed correlation model and the feature risk indicator outputs a verification pass signal, including:

[0138] S451. Map the node relationship of the reconstructed correlation model to the data structure of the feature risk indicator, perform conflict logic detection operation, and output conflict node identification.

[0139] By comparing the node relationship in the reconstructed correlation model with the data structure of the feature risk indicator, detect the logic contradiction points.

[0140] For each rule edge in the correlation model, check the statistical characteristics (such as mean, variance) of the corresponding data item in the feature risk indicator. If the model prediction risk direction (such as policy change pushing up the risk of quotation) conflicts with the actual trend of the indicator (such as the decline of quotation volatility), it is marked as a conflict node.

[0141] Specifically, the conflict determination principle is: when the absolute difference between the predicted value in the reconstructed correlation model and the actual observed value of the feature risk indicator exceeds the historical standard deviation of the feature indicator, it is determined as a logic conflict.

[0142] Output conflict node identification records all contradiction point positions and conflict types.

[0143] S452. Based on the conflict node identification, combine the historical verification rule library of risk linkage analysis to perform conflict level grading operation, and generate conflict level label.

[0144] Specifically, search the historical cases similar to the current conflict in the rule library, calculate the conflict score according to the contradiction amplitude and the case occurrence frequency : ; wherein, represents the absolute difference between the model prediction value and the actual value of the indicator, represents the historical standard deviation of the indicator, represents the number of occurrences of similar conflicts, , represents the industry experience coefficient, such as , .

[0145] Then perform grading output, if is less than or equal to 0.5, output low-level conflict, if is greater than 0.5 and less than or equal to 0.8, output medium-level conflict, and other cases output high-level conflict.

[0146] The conflict level label identifies the severity of each conflict node.

[0147] S453. According to the conflict level label, a topology optimization operation is performed on the reconstructed association model, and a verification pass signal is output.

[0148] According to the conflict level label, the reconstructed association model is adjusted. If it is a low-level conflict, the original topology is maintained; if it is a medium-level conflict, the association edge weight is reduced, such as by 20%; if it is a high-level conflict, the conflict edge is removed and a compensation edge is added, such as a new rule derived from environmental changes.

[0149] After the optimization is completed, if the high-level conflict is eliminated and the overall conflict node proportion is less than a certain value, such as 5%, a verification pass signal is output.

[0150] In some embodiments, based on the verification pass signal in S46, a comprehensive risk profile is generated by performing multi-source dynamic collaboration on external environmental indicator data, time series behavior data, and a failure feature set, including:

[0151] S461. According to the verification pass signal, a dynamic parameter adaptation operation is performed on the external environmental indicator data to generate an adaptive risk model.

[0152] The external environmental indicator data is continuously monitored. When the environmental indicators experience a mutation, such as when the number of policy changes in a single day is greater than or equal to 2, the weight of the indicator is increased by 20%; when the environmental indicators experience stable fluctuations, such as when the price index changes by less than 5%, the baseline weight is maintained.

[0153] Based on the dynamically adjusted parameter weight, an adaptive risk assessment model is constructed: , thereby obtaining an adaptive risk model that stores the dynamic weight; wherein, represents the environmental risk score, represents the policy weight, represents the normalized value of policy change intensity, represents the economic weight, and E represents the normalized value of price volatility.

[0154] For example, it is monitored that 3 new regulations are issued on the same day, which exceeds the threshold, and the policy and regulation weight is increased from 0.6 to 0.72, and the reconstructed model obtains an adaptive risk model.

[0155] S462. The adaptive risk model is fused with the time series behavior data to form a fused risk pattern.

[0156] The adaptive risk model is called, external environmental indicator data is input, and an environmental risk score is output. .

[0157] Analyze the time-series behavior data, specifically, use time-series clustering algorithm to identify behavior patterns, calculate intensity coefficient , Calculated by the average frequency of historical similar behaviors.

[0158] Use time-series clustering techniques to identify high-frequency behavior patterns, such as intensive modification of the tender within the last 1 hour.

[0159] Calculate the risk value for each behavior pattern: ; wherein, represents the environmental risk value, represents the environmental risk score, output by the adaptive risk model, represents the behavior pattern intensity coefficient, calculated by the average frequency of historical similar behaviors.

[0160] Integrate the superimposed risk values of all behavior patterns to generate a fusion risk pattern by weighted average: ; wherein, represents the behavior risk value, represents the risk value of the i-th behavior pattern in a specific environment, represents the intensity coefficient of the i-th behavior pattern, and n represents the number of independent behavior patterns identified by clustering.

[0161] S463. Based on the fusion risk pattern, perform a backtracking data integration operation on the historical dimension data in the failure feature set, and finally construct a comprehensive risk profile.

[0162] Retrieve historical cases similar to the current fusion risk pattern in the failure feature set, such as historical cases with a similarity greater than 80%.

[0163] Recalculate the risk value of the failure factors of the matching cases according to the current weight: ; wherein, represents the historical risk value, represents the new risk value after recalculation, represents the latest policy weight generated by S461, represents the original weight of the historical case.

[0164] The comprehensive risk profile includes environmental risk values , behavior risk values , and new risk values after recalculation .

[0165] In some embodiments, the risk quantification assessment for the comprehensive risk profile in S5 generates a normalized risk value, including:

[0166] S51. For the potential loss data of the comprehensive risk profile, implement a probability interval calculation process to generate a loss probability distribution.

[0167] The loss probability distribution is generated by Monte Carlo simulation method, such as calculating the loss value probability interval by 100000 random sampling.

[0168] For example, the potential loss range of a certain bidding project is 1 million to 5 million yuan, and the probability distribution generated by simulation calculation shows that the probability of loss exceeding 300 million yuan is 15%.

[0169] S52. Based on the loss probability distribution, the multi-dimensional space mapping operation is performed combined with the competitive pressure data in the characteristic risk index to form a risk impact matrix.

[0170] A three-dimensional coordinate system of loss probability, loss intensity and competitive pressure intensity is established, and the data points are mapped to the space grid to generate the risk impact matrix. The mapping rule can use linear interpolation method.

[0171] For example, the loss probability is 15% (dimension X), the loss intensity is 300 million yuan (dimension Y), and the competitive pressure value is 0.8 (dimension Z), which is mapped to the risk impact matrix coordinates (15, 300, 0.8).

[0172] S53. Using the risk impact matrix, the effect aggregation analysis operation is performed on the control strategy characteristics in the fusion risk mode, and the risk mitigation factor is output.

[0173] The strategy effect attenuation coefficient k is calculated, ; wherein k represents the risk mitigation factor, , represents the industry attenuation constant, such as 0.2 for the construction industry and 0.3 for the manufacturing industry, and I represents the strategy implementation intensity.

[0174] For example, increase the margin to 120%, I=1.2, the construction industry =0.2, and is calculated.

[0175] S54. The risk mitigation factor is input into the risk score model after numerical normalization processing with the quantitative evaluation characteristics in the characteristic risk index, and the normalized risk value is generated.

[0176] The quantitative evaluation characteristics in the characteristic risk index are extracted, such as credit score; the risk mitigation factor is linearly superimposed with the quantitative evaluation characteristics: ; wherein represents the superimposed value, , represents the industry weight coefficient, which is determined by regression analysis of industry historical data set, represents the quantitative evaluation characteristic normalized value.

[0177] The superimposed value V is input into the risk score model, such as FICO credit score model, to obtain the normalized risk value.

[0178] In some embodiments, the normalized risk value output in S6 is implemented in response generation operations to produce a visual response report, including:

[0179] S61. According to the normalized risk value, call the early warning template data, implement the template dynamic filling process to generate the risk early warning report.

[0180] According to the risk range of the risk value, match the preset template type, and the range threshold value can be determined by the statistical characteristics of the historical risk event frequency. During the filling process, the project number, timestamp and other fields are injected into the template placeholder to generate the risk early warning report.

[0181] S62. Call the preset early warning template data, and execute the template dynamic filling operation based on the normalized risk value to generate the risk early warning report.

[0182] On the basis of the risk early warning report, an anti-risk person field filling mechanism is added, the responsibility allocation threshold is determined according to the responsibility weight critical value set by the organization management specification, the generated risk early warning report contains an audit tracking identifier, and the response traceability is realized.

[0183] S63. Extract the preset logical rule set, execute the rule logic matching operation based on the risk early warning report, and form the recommended action plan.

[0184] The matching operation adopts a feature field weighted comparison algorithm, and the rule weight is determined through expert evaluation combined with historical execution effectiveness verification. The output includes the recommended action plan containing control measures, responsible subjects and time nodes.

[0185] S64. Obtain historical audit data, execute the audit trajectory iteration operation based on the recommended action plan, and output the updated audit log.

[0186] Calculate the scheme optimization coefficient through the effect optimization model, wherein the time decay factor is determined according to the audit information life cycle management standard, and the updated audit log with version identifier is output to realize the knowledge base continuous learning.

[0187] S65. Load the interactive template of the user interface configuration library, execute the visual format conversion operation based on the updated audit log, and finally obtain the visual response report.

[0188] According to the man-machine interaction interface design specification, the structured data is mapped into chart elements, the conversion rule is automatically adapted according to the display terminal type, and the visual response report supporting multi-platform access is output.

[0189] In some embodiments, as shown in Figure 2 The present application provides a big data-based bidding risk assessment system, which comprises:

[0190] Data collection integration module: collect the basic identification data of the bidding party, historical bidding record data, real-time behavior log data and external environment index data, and generate an integrated initial data set after multi-source data integration processing.

[0191] Data cleaning and standardization module: the integrated initial data set is subjected to data cleaning and standardization processing to obtain the purified standard data.

[0192] Risk feature extraction module: the purified standard data is subjected to risk feature extraction to form a feature risk index.

[0193] Risk linkage analysis module: the feature risk index is used to perform risk linkage analysis to construct a comprehensive risk profile.

[0194] Risk quantification evaluation module: the comprehensive risk profile is subjected to risk quantification evaluation to generate a normalized risk value.

[0195] Response output report module: the normalized risk value is output and response is generated to produce a visual response report.

[0196] In some embodiments, the present application provides a big data-based bidding risk assessment device, which comprises a memory and a processor; the memory is used to store a computer program; and the processor is used to execute the computer program to realize the steps of the big data-based bidding risk assessment method.

[0197] In some embodiments, the present application provides a storage medium, which stores computer program instructions; when the computer program instructions are read and run by a processor, the steps of the big data-based bidding risk assessment method are executed.

[0198] Any reference to memory, storage, database or other medium used in the embodiments of the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory.

[0199] It should be noted that, in the present document, relational terms such as "first" and "second", and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0200] Although the present application has been described in detail with reference to the foregoing embodiments, it should be understood that modifications can be made to the foregoing embodiments, or additional implementations can be implemented, without departing from the spirit and scope of the inventive subject matter. Accordingly, the present application is not limited to the implementations described herein, but is intended to cover all modifications and equivalents falling within the spirit and scope of the inventive subject matter.

Claims

1. A method for risk assessment of bidding and tendering based on big data, characterized in that, Comprise: Collect basic identification data of bidding party, historical bidding record data, real-time behavior log data and external environment index data, and generate integrated initial data set through multi-source data integration processing; Data cleaning and standardization processing is performed on the integrated initial data set to obtain purified standard data; Risk feature extraction is performed on the purified standard data to form feature risk indicators; Risk linkage analysis is performed using the feature risk indicators to construct a comprehensive risk profile, including: Multi-factor correlation analysis is performed on behavior abnormal data in the feature risk indicators to construct a correlation risk graph; From the correlation risk graph, extract the collaborative correlation rules, combine the environmental mutation indicators to perform threshold trigger monitoring operation, and generate environmental abnormal signal; Based on the environmental abnormal signal, rule bias measurement operation is performed on the collaborative correlation rules, and linkage error value is output; The linkage error value is used to perform topological structure reorganization processing on the correlation risk graph to form a reconstructed correlation model; Data logic fitting verification operation is performed on the reconstructed correlation model and feature risk indicators to output verification pass signal, including: Map the node relationship of the reconstructed correlation model to the data structure of the feature risk indicators, perform conflict logic detection operation, and output conflict node identification; Based on the conflict node identification, combined with the historical verification rule library of risk linkage analysis, conflict level grading operation is performed to generate conflict level label; According to the conflict level label, topological structure optimization operation is performed on the reconstructed correlation model to output the verification pass signal; Based on the verification pass signal, through multi-source dynamic collaboration operation on external environment index data, time series behavior data and failure feature set, the comprehensive risk profile is generated, including: According to the verification pass signal, dynamic parameter adaptation operation is performed on the external environment index data to generate an adaptive risk model; Multi-mode fusion is performed between the adaptive risk model and the time series behavior data to form a fusion risk mode; Based on the fusion risk mode, the historical dimension data in the failure feature set is subjected to backtracking data integration operation to finally construct the comprehensive risk profile; Risk quantification evaluation is performed on the comprehensive risk profile to generate normalized risk value; The normalized risk value is subjected to output and response generation operation to produce a visual response report. 2.The big data-based bidding risk assessment method according to claim 1, characterized in that, Collect basic identification data of bidding party, historical bidding record data, real-time behavior log data and external environment index data, and generate integrated initial data set through multi-source data integration processing, including: Obtain bidding party basic identification data, and perform uniqueness verification processing to generate valid identification data; Call historical bidding record data based on the valid identification data, and perform missing value completion processing to output complete historical record; According to the complete historical record, collect real-time behavior log data, and perform time axis calibration processing to form time series behavior data; Based on the time series behavior data, obtain external environment index data, and obtain the integrated initial data set through unit standardization processing. 3.The big data-based bidding risk assessment method according to claim 1, characterized in that, Data cleaning and standardization processing is performed on the integrated initial data set to obtain purified standard data, including: The bidding party financial attribute data of the integrated initial data set is subjected to an abnormal data filtering process to obtain screened financial data; The behavior operation data in the screened financial data is subjected to a mode feature filtering process to generate purified behavior data; The social reputation data in the purified behavior data is subjected to a credibility classification process to output a credible reputation index; The legal compliance data of the credible reputation index is subjected to a specification consistency verification to finally generate the purified standard data. 4.The big data-based bidding risk assessment method according to claim 2, characterized in that, The purified standard data is subjected to risk feature extraction to form a feature risk index, including: The historical failure case data in the purified standard data is subjected to feature vector extraction to obtain a failure feature set; The real-time abnormal signal in the failure feature set is subjected to a mutation point positioning process to generate an abnormal event identifier; The economic fluctuation data is subjected to a trend intensity quantification process according to the abnormal event identifier to form an economic risk trend; The competitive bidding data is subjected to a competitive pressure evaluation process using the economic risk trend to obtain the feature risk index. 5.The big data based bidding risk assessment method according to claim 1, wherein, The integrated risk profile is subjected to risk quantification evaluation to generate a normalized risk value, including: The potential loss data of the integrated risk profile is subjected to a probability interval calculation process to generate a loss probability distribution; Based on the loss probability distribution, a multi-dimensional space mapping operation is performed on the competitive pressure data in the feature risk index to form a risk influence matrix; The risk influence matrix is used to perform an effect aggregation analysis operation on the control strategy features in the fusion risk mode to output a risk mitigation factor; The risk mitigation factor and the quantification evaluation features in the feature risk index are subjected to numerical normalization processing and then input into a risk scoring model to generate the normalized risk value. 6.The big data-based bidding risk assessment method according to claim 1, wherein, The normalized risk value is subjected to output and response generation operations to produce a visual response report, including: An early warning template data is called according to the normalized risk value, and a template dynamic filling process is performed to generate a risk early warning report; A preset early warning template data is called, and a template dynamic filling operation is performed based on the normalized risk value to generate a risk early warning report; A set of preset logical rules is extracted, and a rule logic matching operation is performed based on the risk early warning report to form a recommended action plan; Historical audit data is obtained, and an audit trajectory iteration operation is performed based on the recommended action plan to output an updated audit log; An interactive template of a user interface configuration library is loaded, and a visual format conversion operation is performed based on the updated audit log to finally obtain the visual response report.

7. A big data-based bidding risk assessment system, characterized in that, It includes: A data collection and integration module: collects bidding party basic identification data, historical bidding record data, real-time behavior log data, and external environment index data, and generates an integrated initial data set through multi-source data integration processing; A data cleaning and standardization module: the integrated initial data set is subjected to data cleaning and standardization processing to obtain purified standard data; A risk feature extraction module: the purified standard data is subjected to risk feature extraction to form a feature risk index; A risk linkage analysis module: the feature risk index is used to perform risk linkage analysis to construct an integrated risk profile, including: Perform multi-factor correlation analysis on the behavior anomaly data in the feature risk indicators, and build a correlation risk graph; Extract the synergistic correlation rules from the correlation risk graph, combine the environmental mutation indicators to perform threshold trigger monitoring operations, and generate environmental abnormality signals; Based on the environmental abnormality signals, perform rule bias measurement operations on the synergistic correlation rules, and output linkage error values; Use the linkage error values to perform topological structure reorganization processing on the correlation risk graph, forming a reconstructed correlation model; Perform data logic fitting verification operations on the reconstructed correlation model and the feature risk indicators, and output a verification pass signal, including: Map the node relationship of the reconstructed correlation model to the data structure of the feature risk indicators, perform conflict logic detection operations, and output conflict node identifiers; Based on the conflict node identifiers, combine the historical verification rule library of risk linkage analysis to perform conflict level grading operations, generating conflict level tags; According to the conflict level tags, perform topological structure optimization operations on the reconstructed correlation model, and output a verification pass signal; Based on the verification pass signal, perform multi-source dynamic synergy operations on external environmental indicator data, time series behavior data, and failure feature sets to generate the comprehensive risk profile, including: According to the verification pass signal, perform dynamic parameter adaptation operations on the external environmental indicator data to generate an adaptive risk model; Perform multi-mode fusion on the adaptive risk model and the time series behavior data to form a fusion risk mode; Based on the fusion risk mode, perform backtracking data integration operations on the historical dimension data in the failure feature set to finally build the comprehensive risk profile; Risk quantification evaluation module: perform risk quantification evaluation on the comprehensive risk profile to generate normalized risk values; Response output report module: perform output and response generation operations on the normalized risk values to produce a visual response report.

Images