Intelligent supply chain bid evaluation optimization method and device based on machine learning

Abstract

The invention provides a supply chain intelligent bid evaluation optimization method and device based on machine learning, and the method comprises the steps: obtaining historical behavior data related to a supplier, carrying out the standardization and structuring processing of the historical behavior data, and obtaining a supply chain intelligent bid evaluation optimization result; generating a multi-dimensional feature vector set containing the performance behaviors of the suppliers; the multi-dimensional feature vector set comprises a plurality of features; based on the risk influence degree of each dimension of data in the multi-dimensional feature vector set, distributing an initial weight value for each feature, and generating an information weight matrix corresponding to the multi-dimensional feature vector set; and identifying and extracting risk factors from the historical behavior data according to the information weight matrix. According to the invention, risk missing detection caused by information asymmetry is effectively reduced, the reliability of the bid evaluation result is improved, and the problem of inaccurate risk control evaluation of the supply chain in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of supply chain technology, and in particular to a supply chain intelligent bid evaluation optimization method and device based on machine learning. Background Art

[0002] With the development of artificial intelligence and machine learning technologies, intelligent evaluation systems have been gradually introduced into supply chain management. In the supplier evaluation stage, intelligent evaluation technology can help purchasers judge the supplier's performance capabilities and potential risks more quickly and accurately. However, in the traditional evaluation system, supplier risk assessment mainly relies on manual operations and static scoring methods. Evaluators need to score and classify suppliers based on information on contract execution, performance and payment records. This is not only time-consuming and labor-intensive, but also involves many subjective factors, which can easily lead to inconsistent evaluation results and affect the accuracy and objectivity of the evaluation.

[0003] In the current bidding evaluation system, some companies have begun to introduce automatic scoring models or risk prediction systems based on big data analysis to evaluate the credit and performance capabilities of suppliers through simple statistical analysis or logistic regression. However, the existing technology has exposed some defects in practical applications: First, the bid evaluation model based on static data analysis finds it difficult to dynamically track changes in supplier behavior and cannot promptly reflect the supplier's potential risks. A supplier may perform well in certain stages of fulfillment but face risks of default or delayed payment in later stages, and existing technologies are unable to detect and respond effectively in a timely manner.

[0004] Secondly, traditional scoring methods usually lack the introduction of game theory and risk game models, and only evaluate suppliers based on fixed scoring criteria, ignoring the possible concealment of suppliers and the cost investment of buyers in identifying risks.

[0005] In addition, when processing supplier historical data, existing intelligent bid evaluation systems often use simple similarity matching or fixed weight methods to score, and fail to score weights based on the specific performance of each supplier. This leads to distortion of the scoring results, which is more obvious in complex environments involving a large number of suppliers and multi-dimensional risk factors. Traditional risk assessment methods usually lack sufficient flexibility and cannot be flexibly adjusted according to the supplier's risk performance and the different business needs of buyers, resulting in difficulty in achieving good results when dealing with uncertainty risks in the supply chain.

[0006] There is currently no effective solution to the problem of inaccurate risk control evaluation of the supply chain in existing related technologies. Summary of the invention

[0007] The present invention provides a supply chain intelligent bid evaluation optimization method and device based on machine learning, which are used to solve the defect of inaccurate risk control evaluation of the supply chain in the prior art.

[0008] In a first aspect, the present invention provides a supply chain intelligent bid evaluation optimization method based on machine learning, comprising: Acquire historical behavior data related to the supplier, and perform standardization and structure processing on the historical behavior data to generate a multidimensional feature vector set containing the supplier's performance behavior; the multidimensional feature vector set includes a number of features; Based on the risk impact degree of each dimension of data in the multidimensional feature vector set, an initial weight value is assigned to each feature, and an information weight matrix corresponding to the multidimensional feature vector set is generated; identifying and extracting risk factors from the historical behavior data according to the information weight matrix; Constructing an asymmetric risk control game model based on the risk factors, and generating a determination strategy for each of the risk factors through the asymmetric risk control game model; Iteratively update and optimize the determination strategy of the risk factor, dynamically adjust each initial weight value corresponding to the risk factor, and determine the risk level of the supplier based on the final weight value.

[0009] According to a supply chain intelligent bid evaluation optimization method based on machine learning provided by the present invention, historical behavior data related to a supplier is obtained, and the historical behavior data is standardized and structured to generate a multidimensional feature vector set containing the supplier's performance behavior, including: Obtaining historical behavior data related to the supplier from a supplier performance database; Performing data cleaning and standardization on the historical behavior data, and converting the historical behavior data into a dimensionless form to generate a historical behavior matrix; The historical behavior matrix is ​​structured to generate a multidimensional feature vector set containing the supplier's contract performance behavior.

[0010] According to a supply chain intelligent bid evaluation optimization method based on machine learning provided by the present invention, based on the risk impact degree of each dimensional data in the multidimensional feature vector set, an initial weight value is assigned to each of the features, and an information weight matrix corresponding to the multidimensional feature vector set is generated, including: Analyzing the risk impact of each feature in the multidimensional feature vector set on the supplier based on the historical risk data; Based on the risk impact degree of each feature in the multidimensional feature vector set on the supplier, an initial weight value is assigned to each preset risk impact coefficient, and an initial weight vector is generated; the risk impact coefficient is used to represent the potential impact degree of each feature on the supplier risk; Based on the initial weight vector, the information weight matrix is ​​constructed.

[0011] According to a supply chain intelligent bid evaluation optimization method based on machine learning provided by the present invention, risk factors are identified and extracted from the historical behavior data according to the information weight matrix, including: Aggregating the data in the historical behavior data according to the features in the multidimensional feature vector set to generate a feature vector set; Determine the risk contribution value of each feature in the feature vector set by using the feature vector set and the weight values ​​in the information weight matrix; Based on the risk contribution value, risk factors highly associated with risk are identified and extracted from the historical behavior data.

[0012] According to a supply chain intelligent bid evaluation optimization method based on machine learning provided by the present invention, data in the historical behavior data is aggregated according to the features in the multidimensional feature vector set, including: Aggregating the performance records, contract execution records and payment records in the historical behavior data according to the features in the multidimensional feature vector set; The risk factors include performance deviation, contract default rate and payment delay rate.

[0013] According to a supply chain intelligent bid evaluation optimization method based on machine learning provided by the present invention, an asymmetric risk control game model is constructed based on the risk factors, including: According to the risk factors, an asymmetric risk control game model is constructed between the buyer and the supplier; in the asymmetric risk control game model, the buyer is the party with information advantage, and the supplier is the party with information disadvantage.

[0014] According to a supply chain intelligent bid evaluation optimization method based on machine learning provided by the present invention, a determination strategy for each risk factor is generated through the asymmetric risk control game model, including: In the asymmetric risk control game model, a benefit function of the supplier's risk hiding behavior and a cost function of the buyer's risk identification behavior are defined; constructing the supplier's optimal hiding strategy according to the revenue function, and constructing the buyer's optimal identification strategy according to the cost function; Obtaining an initial determination strategy for each of the risk factors through Nash equilibrium of the optimal hiding strategy and the optimal identification strategy; Optimizing the expected benefits of the supplier and the buyer and the cost function to generate an optimal strategy combination under each risk factor; the optimal strategy combination is used to characterize the optimal hiding probability of the supplier and the optimal identification probability of the buyer; A determination strategy for each of the risk factors is constructed according to the optimal strategy combination.

[0015] According to a supply chain intelligent bid evaluation optimization method based on machine learning provided by the present invention, the determination strategy of the risk factor is iteratively updated and optimized, and each initial weight value corresponding to the risk factor is dynamically adjusted, including: In the asymmetric risk control game model, based on the initial weight values ​​in the information weight matrix, the expected benefits of the buyer and the expected benefits of the supplier under different optimal strategy combinations are determined; Based on the expected profit of the buyer and the expected profit of the supplier, gradually adjust the weight value of each risk factor in the information weight matrix; The buyer's decision strategy and the weight values ​​in the information weight matrix are updated in a cyclic and iterative manner until a convergence condition is met, thereby obtaining a final combination of the information weight matrix and the buyer's decision strategy.

[0016] According to a supply chain intelligent bid evaluation optimization method based on machine learning provided by the present invention, the risk level of the supplier is determined based on the final weight value, including: Determine the overall risk score of each of the suppliers based on the weight values ​​in the dynamically adjusted information weight matrix and the buyer determination strategy combination; Thresholds are applied to the overall risk scores of the suppliers, and the suppliers are classified into different risk levels.

[0017] In a second aspect, the present invention further provides a supply chain intelligent bid evaluation and optimization device based on machine learning, comprising: An acquisition module, used to acquire historical behavior data related to the supplier, and perform standardization and structure processing on the historical behavior data to generate a multidimensional feature vector set containing the supplier's performance behavior; the multidimensional feature vector set includes a number of features; An allocation module, configured to allocate an initial weight value to each of the features based on the risk impact degree of each dimension of data in the multidimensional feature vector set, and generate an information weight matrix corresponding to the multidimensional feature vector set; An extraction module, used to identify and extract risk factors from the historical behavior data according to the information weight matrix; A processing module, configured to construct an asymmetric risk control game model based on the risk factors, and generate a determination strategy for each of the risk factors through the asymmetric risk control game model; The optimization module is used to iteratively update and optimize the determination strategy of the risk factor, dynamically adjust each initial weight value corresponding to the risk factor, and determine the risk level of the supplier based on the final weight value.

[0018] In a third aspect, the present invention further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the supply chain intelligent bid evaluation and optimization method based on machine learning as described in the first aspect above is implemented.

[0019] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the supply chain intelligent bid evaluation optimization method based on machine learning as described in the first aspect above.

[0020] In a fifth aspect, the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the supply chain intelligent bid evaluation and optimization method based on machine learning as described in the first aspect above.

[0021] Compared with the prior art, the present invention has the following beneficial effects: 1. The supply chain intelligent bid evaluation optimization method based on machine learning provided by the present invention takes into account the cost factors of suppliers hiding potential risk factors and buyers identifying risks, allowing buyers to dynamically monitor the potential risks of suppliers through game strategies under conditions of information asymmetry. Traditional bid evaluation methods fail to fully utilize the information differences between buyers and sellers. This method ensures the objectivity and accuracy of the evaluation results by adjusting the weight values ​​and judgment strategies of risk factors, effectively reduces the risk omissions caused by information asymmetry, improves the reliability of bid evaluation results, and solves the problem of inaccurate risk control evaluation of the supply chain in existing related technologies.

[0022] 2. In the dynamic adjustment step of the present invention, the information weight matrix will be continuously iterated and updated as the risk factors are identified to achieve adaptive adjustment of the weights. Different from the traditional static weight method, the present invention gradually optimizes the weight of each risk factor according to the buyer's expected return and the supplier's expected return through a cyclic iterative algorithm, effectively improving the flexibility of the system in a multi-risk factor environment, allowing the system to adjust the contribution of each factor according to the real-time risk situation, thereby improving the accuracy of risk assessment and dynamic response capabilities, and significantly enhancing the system's adaptability to supplier risk fluctuations.

[0023] 3. The present invention constructs an overall risk scoring formula including adjustment coefficients, smoothing factors and nonlinear risk factor weights, which can comprehensively consider the supplier's performance on multiple risk factors and accurately calculate its risk level. It introduces the parameter structure of exponential decay and nonlinear adjustment to smooth the impact of different risk factors and avoid scoring distortion. At the same time, through the threshold division mechanism, the present invention realizes the automatic classification of risks into three levels: low, medium and high, providing a clear risk reference for bid evaluation personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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.

[0025] Figure 1 It is a flow chart of the supply chain intelligent bid evaluation optimization method based on machine learning provided by the present invention; Figure 2 is a schematic diagram of constructing an asymmetric risk game model in an embodiment of the present invention; Figure 3 It is a structural block diagram of a supply chain intelligent bid evaluation and optimization device based on machine learning provided by the present invention; Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0026] 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.

[0027] The present invention provides a supply chain intelligent bid evaluation optimization method based on machine learning. Figure 1 is a flow chart of a supply chain intelligent bid evaluation optimization method based on machine learning provided by the present invention, such as Figure 1 As shown, the method comprises the following steps: Step S101, obtaining historical behavior data related to the supplier, and standardizing and structuring the historical behavior data to generate a multidimensional feature vector set containing the supplier's contract performance behavior; the multidimensional feature vector set includes a number of features.

[0028] Step S102, based on the risk impact degree of each dimensional data in the multidimensional feature vector set, an initial weight value is assigned to each feature, and an information weight matrix corresponding to the multidimensional feature vector set is generated.

[0029] Step S103, identifying and extracting risk factors from historical behavior data according to the information weight matrix.

[0030] Step S104: construct an asymmetric risk control game model based on the risk factors, and generate a determination strategy for each risk factor through the asymmetric risk control game model.

[0031] Step S105, iteratively update and optimize the risk factor determination strategy, dynamically adjust each initial weight value corresponding to the risk factor, and determine the risk level of the supplier based on the final weight value.

[0032] In this method, first, historical behavior data related to the supplier is obtained. The historical behavior data can characterize the supplier's transaction behavior and transaction habits in the past transaction process, so that the buyer can better judge the supplier. The historical behavior data is standardized and structured to generate a multidimensional feature vector set, which is mainly used to characterize the supplier's performance behavior. Then, based on the risk impact of the data in each dimension in the multidimensional feature vector set, an initial weight value is assigned to each feature, and an information weight matrix is ​​generated. The weight value is set based on the risk performance of the historical behavior data and the actual business needs. Each element of the information weight matrix corresponds to the importance of different features in risk control. Then, according to the information weight matrix, all historical behavior data provided by the supplier are analyzed to identify and extract risk factors. Then, based on the risk factors, an asymmetric risk control game model of buyers and suppliers is constructed, and a judgment strategy for each risk factor is generated through the asymmetric risk control game model. Finally, the judgment strategy corresponding to the optimized risk factor is updated, the weight value is adjusted, and the risk level of the supplier is determined according to the adjusted weight value.

[0033] In the above process, the cost factors of suppliers hiding potential risk factors and buyers identifying risks are taken into account, allowing buyers to dynamically monitor potential risks of suppliers through game strategies under conditions of information asymmetry. Traditional bid evaluation methods fail to fully utilize the information differences between buyers and sellers. This method ensures the objectivity and accuracy of the evaluation results by adjusting the weight values ​​and judgment strategies of risk factors, effectively reduces the risk omissions caused by information asymmetry, improves the reliability of bid evaluation results, and solves the problem of inaccurate risk control evaluation of the supply chain in existing related technologies.

[0034] In some of the embodiments, step S101, obtaining historical behavior data related to the supplier, and standardizing and structuring the historical behavior data to generate a multidimensional feature vector set containing the supplier's performance behavior, includes: obtaining historical behavior data related to the supplier from a supplier performance database; cleaning and standardizing the historical behavior data, and converting the historical behavior data into a dimensionless form to generate a historical behavior matrix; structuring the historical behavior matrix to generate a multidimensional feature vector set containing the supplier's performance behavior.

[0035] For example, a supplier's historical behavior dataset D is obtained from the supplier performance database. The historical behavior dataset D includes performance records R, contract execution records C, payment records P, and supplier background information B. The expression of the historical behavior dataset is as follows:

[0036] Where N is the total number of suppliers, is the set of recorded time of supplier i, , , and Respectively represent the performance record, contract execution record, payment record and background information of supplier i at time t. Then, the historical behavior dataset D is cleaned and standardized to remove outliers and noise data, and the data is converted into dimensionless form to generate the historical behavior matrix D'. Finally, the standardized historical behavior matrix D' is structured to form a multidimensional feature vector set V containing the supplier's performance behavior.

[0037] In some of the embodiments, step S102, based on the risk impact of each dimensional data in the multidimensional feature vector set, assigns an initial weight value to each feature, and generates an information weight matrix corresponding to the multidimensional feature vector set, including: analyzing the risk impact of each feature in the multidimensional feature vector set on the supplier based on historical risk data; based on the risk impact of each feature in the multidimensional feature vector set on the supplier, assigning an initial weight value to each preset risk impact coefficient, and generating an initial weight vector; the risk impact coefficient is used to represent the potential impact of each feature on the supplier risk; based on the initial weight vector, constructing an information weight matrix.

[0038] For example, first, based on historical risk data, the impact of each feature in the multidimensional feature vector set V on the supplier risk is analyzed. The multidimensional feature vector set V contains multiple feature values ​​of the supplier's performance behavior, and the risk impact coefficient is defined as , represents the potential impact of feature k on supplier risk. Then, according to actual business needs, the risk impact coefficients Assign initial weight values ​​and generate initial weight vector W. Construct information weight matrix M, the elements of which are It represents the importance measurement of feature i and feature j to risk control under different feature combinations. The specific expression is as follows:

[0039] in, Represents the product of the weight value of feature i and the risk impact coefficient of feature j.

[0040] In some of the embodiments, step S103, identifying and extracting risk factors from historical behavior data according to the information weight matrix, includes: aggregating the data in the historical behavior data according to the features in the multidimensional feature vector set to generate a feature vector set; determining the risk contribution value of each feature in the feature vector set using the feature vector set and the weight value in the information weight matrix; and identifying and extracting risk factors that are highly correlated with risks from the historical behavior data based on the risk contribution values.

[0041] Furthermore, the data in the historical behavior data are aggregated according to the features in the multidimensional feature vector set, including: aggregating the performance records, contract execution records and payment records in the historical behavior data according to the features in the multidimensional feature vector set. Correspondingly, the risk factors include performance deviation, contract default rate and payment delay rate.

[0042] For example, first, the supplier's historical behavior data set D is analyzed based on the information weight matrix M. The elements of the information weight matrix M are Represents the weight of feature i and feature j in risk control. Then, the performance records R, contract execution records C, and payment records P in the historical behavior data are aggregated according to the features to generate a feature vector set , where each feature Represents the specific behavior of the supplier and is associated with a corresponding weight value Then use the feature vector set F and the weight values ​​in the information weight matrix M to calculate the risk contribution value of each feature , the specific expression is as follows:

[0043] in, Representation characteristics Contribution to the overall risk assessment of the supplier. Finally, based on the risk contribution value Identify and extract risk factors that are highly correlated with risk , where risk factors include performance deviation , Contract Default Rate and payment delay rate , the specific expression is as follows:

[0044] in, represents the risk assessment value of the i-th risk factor, is the basic weight coefficient of the risk factor, which is set according to the risk characteristics of the supply chain bid evaluation. is the eigenvalue in the eigenvector set, is the risk factor offset, is the adjustment coefficient of the risk factor, is an exponential decay factor.

[0045] Figure 2 Schematic diagram of constructing an asymmetric risk game model in an embodiment of the present invention, such as Figure 2 As shown, in some of the embodiments, step S104, constructing an asymmetric risk control game model based on risk factors, includes: constructing an asymmetric risk control game model between the buyer and the supplier according to the risk factors; in the asymmetric risk control game model, the buyer is the party with information advantage, and the supplier is the party with information disadvantage.

[0046] On this basis, a judgment strategy for each risk factor is generated through an asymmetric risk control game model, including: in the asymmetric risk control game model, defining the benefit function of the supplier's risk hiding behavior and the cost function of the buyer's risk identification; constructing the supplier's optimal hiding strategy according to the benefit function, and constructing the buyer's optimal identification strategy according to the cost function; obtaining the initial judgment strategy for each risk factor through the Nash equilibrium of the optimal hiding strategy and the optimal identification strategy; optimizing the expected benefits and cost functions of the supplier and the buyer to generate the optimal strategy combination under each risk factor; the optimal strategy combination is used to characterize the supplier's optimal hiding probability and the buyer's optimal identification probability; and constructing a judgment strategy for each risk factor based on the optimal strategy combination.

[0047] For example, according to the risk factor set Construct an asymmetric risk control game model between buyers and suppliers. In the asymmetric risk control game model, the buyer is set as the party with information advantage and has more information about the supplier's historical behavior data, and the supplier is set as the party with information disadvantage and hides potential risk factors. Then, define the profit function of the supplier's hidden risk behavior in the asymmetric risk control game model and the buyer's cost function for identifying risk , the supplier’s revenue function The expression is as follows:

[0048] in, is the risk assessment value of the ith risk factor, indicating the supplier’s risk level in terms of performance deviation, contract default rate, and payment delay rate. is the probability of the buyer identifying the i-th risk factor, ranging from , reflecting the buyer's emphasis on this risk factor, is the expected profit that the supplier obtains by successfully hiding the i-th risk factor, reflecting the contractual benefits that the supplier obtains after concealing the risk. is the cost paid by the supplier to hide the i-th risk factor, The cost sensitivity coefficient of the supplier's risk-hiding behavior reflects the supplier's sensitivity to hidden costs. The buyer's cost function The expression is as follows:

[0049] in, The basic cost required to identify the i-th risk factor for the buyer, The risk sensitivity coefficient of the buyer's identification cost. In the risk control game model, the supplier's optimal hiding strategy and the buyer's optimal identification strategy are established based on the benefits and costs of both parties. By solving the Nash equilibrium, the judgment strategy for each risk factor is obtained, and the supplier's expected profit function is for:

[0050] in, is the supplier's hiding probability of the i-th risk factor, indicating the supplier's tendency to choose to conceal the risk factor. Then, by optimizing the expected benefits and cost functions of the supplier and the buyer, the optimal strategy combination under each risk factor is solved. , the supplier's optimal hiding probability for:

[0051] The buyer's optimal identification probability satisfy:

[0052] Finally, according to the optimal strategy combination Establish a determination strategy for each risk factor.

[0053] On the basis of the above embodiment, step S105 iteratively updates the judgment strategy for optimizing risk factors, and dynamically adjusts each initial weight value corresponding to the risk factor, including: in an asymmetric risk control game model, based on the initial weight value in the information weight matrix, determining the buyer's expected profit and the supplier's expected profit under different optimal strategy combinations; based on the buyer's expected profit and the supplier's expected profit, gradually adjusting the weight value of each risk factor in the information weight matrix; cyclically iteratively updating the buyer's judgment strategy and the weight value in the information weight matrix until the convergence condition is met, and obtaining the final information weight matrix and buyer's judgment strategy combination.

[0054] For example, in the risk control game model, based on the initial weight value of the information weight matrix M, the buyer's optimal strategy combination in different games is calculated. Expected return under and suppliers’ revenue expectations :

[0055] in, is the expected return that the buyer will receive after identifying risk factor i. Based on the expected return function and , gradually adjust the weight value of each risk factor in the information weight matrix M , the specific expression is as follows:

[0056] in, and denote the weight of risk factor i at iteration t+1 and t respectively, is the learning rate, which controls the step size of weight adjustment. is the increase in the buyer’s expected return, is the expected increment of the supplier's profit. Then, the buyer's decision strategy and the weight value in the information weight matrix are continuously updated through cyclic iteration until the convergence condition is met, that is, the buyer's expected profit and supplier revenue expectations Reach a stable value and get the final information weight matrix and buyer decision strategy combination .

[0057] Furthermore, based on the final weight value, the risk level of the supplier is determined, including: determining the overall risk score of each supplier based on the weight value in the dynamically adjusted information weight matrix and the buyer's judgment strategy combination; and performing threshold division on the overall risk score of the supplier to classify the supplier into different risk levels.

[0058] Exemplarily, based on the dynamically adjusted information weight matrix The optimized weight value and buyer decision strategy combination in , calculate the overall risk score for each supplier , the specific expression is as follows:

[0059] in, Provide an overall risk score for suppliers to assess their comprehensive performance under multiple risk factors. is the optimization weight of risk factor i, indicating its contribution to the overall score of the supplier in the final risk assessment. is the risk assessment value of the ith risk factor, which is used to quantify the risk level of the supplier in terms of risk factors. is the supplier’s specific performance score on risk factor i, is the adjustment coefficient, is the adjustment parameter of risk factor i, is the risk contribution value of the ith risk factor, is a smoothing factor used to control the nonlinear impact of supplier performance on overall risk.

[0060] Then, the calculated overall risk score Perform threshold division and classify suppliers into different risk levels: define low risk threshold and high risk threshold ,in , the risk level of the supplier Determined according to the following rules:

[0061] In addition, the overall risk score of the supplier can be and their corresponding risk levels Stored in the supplier risk assessment database.

[0062] In order to verify the effectiveness of the above method, the following is based on a procurement bid evaluation project of a large manufacturing enterprise, which aims to optimize the intelligent evaluation and risk control of suppliers. Manufacturing enterprises need to purchase a large amount of raw materials and equipment every year, involving hundreds of suppliers. Therefore, the evaluation and screening of suppliers become key links. Traditional bid evaluation methods are mostly manual screening and static scoring, which are not only time-consuming and labor-intensive but also have information asymmetry problems. In order to improve the efficiency and accuracy of bid evaluation, the enterprise decided to introduce a supply chain intelligent bid evaluation algorithm optimization method based on machine learning into its bid evaluation system to apply the present invention to achieve more efficient supplier risk management.

[0063] In April 2024, the manufacturing company released its annual equipment procurement bidding plan, planning to purchase a series of high-precision manufacturing equipment in batches within the next year. To ensure quality and delivery time, the company invited about 300 suppliers to participate in the bidding. The suppliers submitted their historical performance records, contract execution performance and payment records. Due to the large number of suppliers, the limitations of traditional bid evaluation methods have become increasingly prominent. Manual review of supplier data and risk assessment has become a time-consuming and inefficient task. Therefore, the company decided to use the method of the present invention to achieve a comprehensive evaluation of each supplier through an intelligent algorithm, so as to more objectively analyze its performance capabilities and potential risks.

[0064] First, the company extracted the historical behavior data of each supplier participating in the bidding from the supplier performance database, including performance records, contract execution records, payment records and related background information. For example, the performance record of supplier A showed that its on-time delivery rate was 90%, the contract execution compliance rate was 95%, and the payment delay rate was 5% in the past year. Similarly, the data of each supplier was stored in the form of a multi-dimensional feature vector for further analysis.

[0065] This method cleans and standardizes the data, removes abnormal data, and forms a multidimensional data set containing the supplier's performance. The standardized data set ensures that all indicators of all suppliers are comparable in the same dimension, which is convenient for subsequent analysis.

[0066] Next, the standardized supplier data is converted into a multidimensional feature vector. For supplier A, the feature vector includes its on-time delivery rate, contract compliance rate, and payment delay rate indicators. The data of all suppliers is analyzed, and based on the risk performance in historical data, the initial weights of each feature are assigned and an information weight matrix is ​​generated. For example, the weight of on-time delivery rate is 0.4, the contract compliance rate is 0.3, and the payment delay rate is 0.3. The weight distribution is based on the statistical analysis of past procurement data, so that important features receive higher attention in the evaluation.

[0067] This method introduces an asymmetric risk control game model between buyers and suppliers. Taking supplier B as an example, its historical data shows that 5% of its orders are delayed in delivery. The buyer, as the party with information advantage, can predict the risk factors that the supplier may conceal through the game model, such as the risk of delivery delay. The model balances the buyer's investment in identifying risk factors and the benefits of the supplier's attempt to conceal risks by setting the profit function of the supplier's concealment behavior and the buyer's identification cost function.

[0068] Combined with the above-mentioned asymmetric risk control game model system, various risk factors are dynamically adjusted. During the initial scoring, it was found that the contract compliance rate weight of supplier C was low. Under the game balance, the weight value of the contract compliance rate was automatically adjusted from 0.3 to 0.35 to ensure that this factor is more important in the risk score. This process is continuously iterated, and optimization is carried out after each round of evaluation based on the balance of benefits and costs between buyers and sellers.

[0069] Ultimately, the comprehensive risk score of each supplier is calculated through a complex overall risk scoring formula and divided into low risk, medium risk and high risk levels. For example, supplier A’s overall score is 0.62, which is at a medium risk level; supplier B’s score is 0.85, which is classified as high risk; supplier C’s score is 0.45, which is rated as low risk. The calculation of the risk score combines the weights of multiple factors, historical data and the results of the game model to ensure the comprehensiveness of the final evaluation.

[0070] In this embodiment, the enterprise conducted large-scale simulation training on the historical data of 300 suppliers, and compared this method with the traditional static scoring method. The simulation data included important risk factors such as supplier performance deviation, contract default rate and payment delay rate. During the evaluation period, the system automatically adjusted the information weight according to the game model, so that the scoring results dynamically responded to risk changes. The specific data comparison is shown in the following table: Table 1 Comparison of evaluation performance data of different methods

[0071] It can be seen from the data results in the above table that this method significantly shortens the evaluation time and can be completed in only 5 hours, while the traditional method takes up to 18 hours. In addition, the evaluation accuracy of the present invention is improved to 91%, and the risk omission rate is reduced to 4%, which is significantly higher than the 68% recognition rate of the traditional method, proving the feasibility and effectiveness of the present invention in improving the efficiency and accuracy of supply chain intelligent bidding evaluation.

[0072] Further experimental results show that the method of the present invention can effectively adapt to changes in supplier data in the process of dynamically adjusting weights and scores. When the simulated supplier D's fulfillment rate drops by 5%, the system adjusts the weight of this factor in real time, causing D's risk score to increase from 0.58 to 0.76, jumping from a medium risk level to a high risk level, reminding bid evaluators to pay special attention to this supplier. This flexibility cannot be achieved by traditional static scoring methods.

[0073] It can be seen from the above application process that the supply chain intelligent bidding evaluation method of the present invention has strong advantages in dealing with complex supplier relationships and risk assessment, and can provide scientific and accurate decision-making support for corporate procurement, greatly improving the intelligence level of supply chain management.

[0074] The present invention constructs an asymmetric risk game model between buyers and suppliers in the bid evaluation method, introduces the cost factors of suppliers hiding potential risk factors and buyers identifying risks into the risk assessment model, and allows buyers to dynamically monitor suppliers' potential risks through game strategies under conditions of information asymmetry. Traditional bid evaluation methods fail to fully utilize the information differences between buyers and sellers, while the present invention automatically adjusts the weights and strategies of risk factors through the benefit function of suppliers' hiding behavior and the buyer's identification cost function, thereby ensuring the objectivity and accuracy of the evaluation results, effectively reducing risk omissions caused by information asymmetry, and improving the reliability of bid evaluation results.

[0075] In the dynamic adjustment step of the present invention, the information weight matrix will be continuously iterated and updated as the risk factors are identified to achieve adaptive adjustment of the weights. Different from the traditional static weight method, the present invention gradually optimizes the weight of each risk factor according to the buyer's expected return and the supplier's expected return through a cyclic iterative algorithm, effectively improving the flexibility of the system in a multi-risk factor environment, allowing the system to adjust the contribution of each factor according to the real-time risk situation, thereby improving the accuracy of risk assessment and dynamic response capabilities, and significantly enhancing the system's adaptability to supplier risk fluctuations.

[0076] The present invention constructs an overall risk scoring formula including adjustment coefficients, smoothing factors and nonlinear risk factor weights, which can comprehensively consider the supplier's performance on multiple risk factors and accurately calculate its risk level. It introduces the parameter structure of exponential decay and nonlinear adjustment to smooth the impact of different risk factors and avoid scoring distortion. At the same time, through the threshold division mechanism, the present invention realizes the automatic classification of risks into three levels: low, medium and high, providing a clear risk reference for bid evaluation personnel.

[0077] The present invention also provides a supply chain intelligent bid evaluation and optimization device based on machine learning. The supply chain intelligent bid evaluation and optimization device based on machine learning provided by the present invention is described below. The supply chain intelligent bid evaluation and optimization device based on machine learning described below and the supply chain intelligent bid evaluation and optimization method based on machine learning described above can be referenced to each other. Figure 3 is a structural block diagram of a supply chain intelligent bid evaluation and optimization device based on machine learning provided by the present invention, such as Figure 3 As shown, the device comprises: The acquisition module 301 is used to acquire historical behavior data related to the supplier, and perform standardization and structure processing on the historical behavior data to generate a multidimensional feature vector set containing the supplier's performance behavior; the multidimensional feature vector set includes several features; An allocation module 302 is used to allocate an initial weight value to each feature based on the risk impact degree of each dimension of data in the multidimensional feature vector set, and generate an information weight matrix corresponding to the multidimensional feature vector set; Extraction module 303, used to identify and extract risk factors from historical behavior data according to the information weight matrix; The processing module 304 is used to construct an asymmetric risk control game model based on the risk factors, and generate a determination strategy for each risk factor through the asymmetric risk control game model; The optimization module 305 is used to iteratively update the determination strategy of the optimized risk factor, dynamically adjust each initial weight value corresponding to the risk factor, and determine the risk level of the supplier based on the final weight value.

[0078] When the device is in use, first, the acquisition module 301 acquires historical behavior data related to the supplier. The historical behavior data can characterize the supplier's transaction behavior and transaction habits in the past transaction process, so that the buyer can better judge the supplier. The historical behavior data is standardized and structured to generate a multidimensional feature vector set, which is mainly used to characterize the supplier's performance behavior. Then, the allocation module 302 assigns an initial weight value to each feature based on the risk impact degree of the data in each dimension in the multidimensional feature vector set, and generates an information weight matrix. The setting of the weight value is based on the risk performance of the historical behavior data and the actual business needs. Each element of the information weight matrix corresponds to the importance of different features in risk control. Then, the extraction module 303 analyzes all the historical behavior data provided by the supplier according to the information weight matrix, identifies and extracts the risk factors. The processing module 304 then constructs an asymmetric risk control game model for the buyer and the supplier based on the risk factors, and generates a judgment strategy for each risk factor through the asymmetric risk control game model. Finally, the optimization module 305 updates the determination strategy corresponding to the optimization risk factor, adjusts the weight value, and determines the risk level of the supplier according to the adjusted weight value.

[0079] In the above process, the cost factors of suppliers hiding potential risk factors and buyers identifying risks are taken into account, allowing buyers to dynamically monitor potential risks of suppliers through game strategies under conditions of information asymmetry. Traditional bid evaluation methods fail to fully utilize the information differences between buyers and sellers. The device ensures the objectivity and accuracy of the evaluation results by adjusting the weight values ​​and judgment strategies of risk factors, effectively reduces the risk omissions caused by information asymmetry, improves the reliability of bid evaluation results, and solves the problem of inaccurate risk control evaluation of the supply chain in existing related technologies.

[0080] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4As shown, the electronic device may include: a processor 401, a communications interface 402, a memory 403 and a communication bus 404, wherein the processor 401, the communications interface 402 and the memory 403 communicate with each other via the communication bus 404. The processor 401 may call the logic instructions in the memory 403 to execute the supply chain intelligent bid evaluation optimization method based on machine learning, and the method includes: Acquire historical behavior data related to suppliers, standardize and structure the historical behavior data, and generate a multidimensional feature vector set containing supplier performance behaviors; the multidimensional feature vector set includes several features; Based on the risk impact of each dimension of data in the multidimensional feature vector set, an initial weight value is assigned to each feature, and an information weight matrix corresponding to the multidimensional feature vector set is generated; Identify and extract risk factors from historical behavior data based on the information weight matrix; Construct an asymmetric risk control game model based on risk factors, and generate a decision strategy for each risk factor through the asymmetric risk control game model; Iterate and update the risk factor determination strategy to optimize the risk factors, dynamically adjust the initial weight value corresponding to each risk factor, and determine the supplier's risk level based on the final weight value.

[0081] In addition, the logic instructions in the above-mentioned memory 403 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a disk or an optical disk.

[0082] On the other hand, the present invention further provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the supply chain intelligent bid evaluation optimization method based on machine learning provided by the above methods, which includes: Acquire historical behavior data related to suppliers, standardize and structure the historical behavior data, and generate a multidimensional feature vector set containing supplier performance behaviors; the multidimensional feature vector set includes several features; Based on the risk impact of each dimension of data in the multidimensional feature vector set, an initial weight value is assigned to each feature, and an information weight matrix corresponding to the multidimensional feature vector set is generated; Identify and extract risk factors from historical behavior data based on the information weight matrix; Construct an asymmetric risk control game model based on risk factors, and generate a decision strategy for each risk factor through the asymmetric risk control game model; Iterate and update the risk factor determination strategy to optimize the risk factors, dynamically adjust the initial weight value corresponding to each risk factor, and determine the supplier's risk level based on the final weight value.

[0083] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to execute the supply chain intelligent bid evaluation optimization method based on machine learning provided by the above methods, the method comprising: Acquire historical behavior data related to suppliers, standardize and structure the historical behavior data, and generate a multidimensional feature vector set containing supplier performance behaviors; the multidimensional feature vector set includes several features; Based on the risk impact of each dimension of data in the multidimensional feature vector set, an initial weight value is assigned to each feature, and an information weight matrix corresponding to the multidimensional feature vector set is generated; Identify and extract risk factors from historical behavior data based on the information weight matrix; Construct an asymmetric risk control game model based on risk factors, and generate a decision strategy for each risk factor through the asymmetric risk control game model; Iterate and update the risk factor determination strategy to optimize the risk factors, dynamically adjust the initial weight value corresponding to each risk factor, and determine the supplier's risk level based on the final weight value.

[0084] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0085] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. 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 supply chain intelligent bid evaluation optimization method based on machine learning, characterized in that: include: Acquire historical behavior data related to the supplier, and perform standardization and structure processing on the historical behavior data to generate a multi-dimensional feature vector set containing the supplier's performance behavior; The multidimensional feature vector set includes several features; Based on the risk impact degree of each dimension of data in the multidimensional feature vector set, an initial weight value is assigned to each feature, and an information weight matrix corresponding to the multidimensional feature vector set is generated; identifying and extracting risk factors from the historical behavior data according to the information weight matrix; Constructing an asymmetric risk control game model based on the risk factors, and generating a determination strategy for each of the risk factors through the asymmetric risk control game model; Iteratively update and optimize the determination strategy of the risk factor, dynamically adjust each initial weight value corresponding to the risk factor, and determine the risk level of the supplier based on the final weight value.

2. The supply chain intelligent bid evaluation optimization method based on machine learning according to claim 1 is characterized in that: Acquire historical behavior data related to the supplier, and perform standardization and structured processing on the historical behavior data to generate a multidimensional feature vector set containing the supplier's performance behavior, including: Obtaining historical behavior data related to the supplier from a supplier performance database; Performing data cleaning and standardization on the historical behavior data, and converting the historical behavior data into a dimensionless form to generate a historical behavior matrix; The historical behavior matrix is ​​structured to generate a multidimensional feature vector set containing the supplier's contract performance behavior.

3. The supply chain intelligent bid evaluation optimization method based on machine learning according to claim 1 is characterized in that: Based on the risk impact degree of each dimension of data in the multidimensional feature vector set, an initial weight value is assigned to each feature, and an information weight matrix corresponding to the multidimensional feature vector set is generated, including: Analyzing the risk impact of each feature in the multidimensional feature vector set on the supplier based on the historical risk data; Based on the risk impact degree of each feature in the multidimensional feature vector set on the supplier, an initial weight value is assigned to each preset risk impact coefficient, and an initial weight vector is generated; the risk impact coefficient is used to represent the potential impact degree of each feature on the supplier risk; Based on the initial weight vector, the information weight matrix is ​​constructed.

4. The supply chain intelligent bid evaluation optimization method based on machine learning according to claim 1 is characterized in that: According to the information weight matrix, identifying and extracting risk factors from the historical behavior data includes: Aggregating the data in the historical behavior data according to the features in the multidimensional feature vector set to generate a feature vector set; Determine the risk contribution value of each feature in the feature vector set by using the feature vector set and the weight values ​​in the information weight matrix; Based on the risk contribution value, risk factors highly associated with risk are identified and extracted from the historical behavior data.

5. The supply chain intelligent bid evaluation optimization method based on machine learning according to claim 4 is characterized in that: Aggregating the data in the historical behavior data according to the features in the multidimensional feature vector set includes: Aggregating the performance records, contract execution records and payment records in the historical behavior data according to the features in the multidimensional feature vector set; The risk factors include performance deviation, contract default rate and payment delay rate.

6. The supply chain intelligent bid evaluation optimization method based on machine learning according to claim 1 is characterized in that: An asymmetric risk control game model is constructed based on the risk factors, including: According to the risk factors, an asymmetric risk control game model is constructed between the buyer and the supplier; in the asymmetric risk control game model, the buyer is the party with information advantage, and the supplier is the party with information disadvantage.

7. The supply chain intelligent bid evaluation optimization method based on machine learning according to claim 6 is characterized in that: Generating a determination strategy for each of the risk factors through the asymmetric risk control game model includes: In the asymmetric risk control game model, a benefit function of the supplier's risk hiding behavior and a cost function of the buyer's risk identification behavior are defined; constructing the supplier's optimal hiding strategy according to the revenue function, and constructing the buyer's optimal identification strategy according to the cost function; Obtaining an initial determination strategy for each of the risk factors through Nash equilibrium of the optimal hiding strategy and the optimal identification strategy; Optimizing the expected benefits of the supplier and the buyer and the cost function to generate an optimal strategy combination under each risk factor; the optimal strategy combination is used to characterize the optimal hiding probability of the supplier and the optimal identification probability of the buyer; A determination strategy for each of the risk factors is constructed according to the optimal strategy combination.

8. The supply chain intelligent bid evaluation optimization method based on machine learning according to claim 7 is characterized in that: Iteratively updating and optimizing the determination strategy of the risk factor, and dynamically adjusting each initial weight value corresponding to the risk factor, including: In the asymmetric risk control game model, based on the initial weight values ​​in the information weight matrix, the expected benefits of the buyer and the expected benefits of the supplier under different optimal strategy combinations are determined; Based on the expected profit of the buyer and the expected profit of the supplier, gradually adjust the weight value of each risk factor in the information weight matrix; The buyer's decision strategy and the weight values ​​in the information weight matrix are updated in a cyclic and iterative manner until a convergence condition is met, thereby obtaining a final combination of the information weight matrix and the buyer's decision strategy.

9. The supply chain intelligent bid evaluation optimization method based on machine learning according to claim 8 is characterized in that: Based on the final weighted value, the risk level of the supplier is determined, including: Determine the overall risk score of each of the suppliers based on the weight values ​​in the dynamically adjusted information weight matrix and the buyer determination strategy combination; Thresholds are applied to the overall risk scores of the suppliers, and the suppliers are classified into different risk levels.

10. A supply chain intelligent bid evaluation and optimization device based on machine learning, characterized in that: include: An acquisition module, used to acquire historical behavior data related to the supplier, and perform standardization and structure processing on the historical behavior data to generate a multi-dimensional feature vector set containing the supplier's performance behavior; The multidimensional feature vector set includes several features; An allocation module, configured to allocate an initial weight value to each of the features based on the risk impact degree of each dimension of data in the multidimensional feature vector set, and generate an information weight matrix corresponding to the multidimensional feature vector set; An extraction module, used to identify and extract risk factors from the historical behavior data according to the information weight matrix; A processing module, configured to construct an asymmetric risk control game model based on the risk factors, and generate a determination strategy for each of the risk factors through the asymmetric risk control game model; The optimization module is used to iteratively update and optimize the determination strategy of the risk factor, dynamically adjust each initial weight value corresponding to the risk factor, and determine the risk level of the supplier based on the final weight value.

Images