Farm-related enterprise credit scoring method, equipment and medium

Abstract

The invention discloses an agricultural enterprise credit scoring method and device, and a medium, and relates to the technical field of credit risk assessment. The method comprises the following steps: acquiring multi-source heterogeneous data of an enterprise; based on the historical price data of the agricultural product, determining the current price cycle phase of the agricultural product; obtaining a weight corresponding to the operating characteristic in the price cycle phase from the characteristic weight matrix; and inputting the basic credit features, the operation features and the weights into a credit scoring model to obtain the credit score of the enterprise output by the credit scoring model. Through the mode, the problems of risk identification lagging and insufficient response to periodic market fluctuation of an existing method can be solved.

Description

Technical Field

[0001] This application relates to the field of credit risk assessment technology, and in particular to a credit scoring method, equipment and medium for agricultural enterprises. Background Technology

[0002] In the field of credit risk assessment, traditional credit scoring models mainly employ machine learning methods such as logistic regression, support vector machines, or decision trees, modeling based on static structured features such as the borrower's historical repayment records, asset and liability status, and income level. These methods exhibit good classification performance in scenarios with complete financial data and stable market environments. However, for credit assessment of agricultural business entities, their risk is significantly affected by external market conditions, particularly agricultural cycle fluctuations, making it difficult for traditional models to effectively capture such dynamic systemic risks.

[0003] Existing technologies primarily rely on the static credit characteristics of borrowers for scoring, with fixed model parameters that fail to reflect the dynamic impact of changes in the external market environment on repayment ability. This is particularly true in the agricultural sector, where agricultural product prices exhibit significant nonlinear and non-stationary cyclical fluctuations, and the risk implications of the same financial indicator differ significantly across different cyclical phases. Furthermore, traditional methods fail to effectively integrate time-varying information from price cycles, leading to delayed risk warnings during price downturns and potential underestimation of bubble risks during upturns. Summary of the Invention

[0004] This application provides a credit scoring method, device, and medium for agricultural enterprises to address the following technical problem: how to solve the problems of lagging risk identification and insufficient response to cyclical market fluctuations in existing methods.

[0005] In a first aspect, embodiments of this application provide a credit scoring method for agricultural enterprises. The method includes: acquiring multi-source heterogeneous data of the enterprise, wherein the enterprise is an agricultural enterprise, and the multi-source heterogeneous data includes: the enterprise's basic credit characteristics, operational characteristics, and the type of agricultural product corresponding to the enterprise and historical price data of the agricultural product; determining the current price cycle phase of the agricultural product based on the historical price data of the agricultural product; obtaining the weights corresponding to the operational characteristics under the price cycle phase from a feature weight matrix, wherein the feature weight matrix includes at least one weight corresponding to the operational characteristics under at least one phase; inputting the basic credit characteristics, the operational characteristics, and the weights into a credit scoring model to obtain the credit score of the enterprise output by the credit scoring model, wherein the credit scoring model uses a dual-channel fusion structure to score the enterprise, the first channel processes the basic credit characteristics, and the second channel uses the weights to weight the operational characteristics.

[0006] Secondly, embodiments of this application also provide a credit scoring device for agricultural enterprises, the device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform an agricultural enterprise credit scoring method as described in the first aspect above.

[0007] Thirdly, embodiments of this application also provide a computer storage medium storing computer-executable instructions, which, when executed, implement a credit scoring method for agricultural enterprises as described in the first aspect above.

[0008] The credit scoring method, equipment, and medium for agricultural enterprises provided in this application have the following beneficial effects: In this embodiment, multi-source heterogeneous data of an enterprise can be acquired. Then, based on the historical price data of the enterprise's agricultural products, the current price cycle phase of the agricultural products is determined. Furthermore, the weights of the operating characteristics corresponding to the aforementioned price cycle phase are obtained from the feature weight matrix. Finally, the basic credit characteristics, operating characteristics, and weights are input into the credit scoring model to obtain the enterprise's credit score. This approach overcomes the limitations of existing static scoring models. By identifying the price cycle phase of agricultural products and determining the weights of economic characteristics accordingly, dynamic adjustment of the weights can be achieved, making the scoring results forward-looking and adaptive. This addresses the problems of delayed risk identification and insufficient response to cyclical market fluctuations. Attached Figure Description

[0009] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart of a credit scoring method for agricultural enterprises provided in this application embodiment; Figure 2 This is a schematic diagram of the evaluation results of a model that does not incorporate agricultural cyclical fluctuations. Figure 3 A schematic diagram illustrating the evaluation results of a model incorporating agricultural cyclical fluctuation characteristics, provided as an embodiment of this application; Figure 4 Flowchart of another credit scoring method for agricultural enterprises provided in this application embodiment; Figure 5 This is a schematic diagram of the internal structure of a credit scoring device for agricultural enterprises provided in an embodiment of this application. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0011] This application provides a credit scoring scheme for agricultural enterprises. The technical solution proposed in this application will be described in detail below with reference to the accompanying drawings.

[0012] Figure 1 This is a flowchart illustrating a credit scoring method for agricultural enterprises, provided as an embodiment of this application. Figure 1 As shown in the embodiment of this application, a credit scoring method for agricultural enterprises specifically includes the following steps: Step 101: Obtain multi-source heterogeneous data from the enterprise.

[0013] The enterprise is an agricultural enterprise, and the multi-source heterogeneous data includes: the enterprise's basic credit characteristics, operating characteristics, the type of agricultural products corresponding to the enterprise, and the historical price data of the agricultural products.

[0014] In this embodiment, multi-source heterogeneous data on the enterprise can be obtained first, including the enterprise's basic credit characteristics, operational characteristics, and the types of agricultural products corresponding to the enterprise, as well as the historical price data of those agricultural products. In practical applications, basic credit characteristics can refer to the enterprise's basic credit information, such as its establishment year, registered capital, historical repayment records, guarantee methods, and credit reports. Operational characteristics can refer to the enterprise's operational data, such as its output, inventory, feed costs, profit margin, and cash flow status, with data granularity ranging from monthly to quarterly. The types of agricultural products corresponding to the enterprise refer to the types of agricultural products that the enterprise mainly produces, which are generally only one type, such as pig farming or potato cultivation. This provides a solid data foundation for subsequent credit scoring.

[0015] Step 102: Based on the historical price data of the agricultural product, determine the current price cycle phase of the agricultural product.

[0016] In practical applications, agricultural product price cycles directly impact corporate revenue. For example, if hog prices are in an upward cycle, a company's sales revenue is expected to increase, and its repayment ability will improve; conversely, if prices are in a downward cycle, the company needs to be wary of potential cash flow pressure. Therefore, in this embodiment of the application, when conducting credit scoring for agricultural enterprises, the current phase of the price cycle of the agricultural product can be determined based on historical price data, which facilitates a more accurate credit assessment of the enterprise in the future.

[0017] Step 103: Obtain the weights corresponding to the operating characteristics under the price cycle phase from the feature weight matrix.

[0018] The feature weight matrix includes at least one weight corresponding to the operating feature under each phase.

[0019] In this embodiment of the application, the weight of the operating feature corresponding to the above-mentioned price cycle phase can be obtained from the feature weight matrix. The feature weight matrix is ​​pre-constructed based on historical data and includes the weight of the operating feature under at least one phase; wherein, under the same phase, the operating feature and the weight correspond one-to-one.

[0020] In practical applications, traditional credit assessment is typically based on fixed-weight models. However, price cycle phases (such as rising, peak, falling, and trough periods) can significantly alter the impact of operational characteristics (such as inventory levels, cost structure, and cash flow stability) on risk. By extracting the weights of operational characteristics corresponding to the current price cycle phase from the feature weight matrix—that is, dynamically adjusting the weights—more accurate credit assessments can be achieved.

[0021] Step 104: Input the basic credit characteristics, the business characteristics and the weights into the credit scoring model to obtain the credit score of the enterprise output by the credit scoring model.

[0022] The credit scoring model employs a dual-channel fusion structure to score the enterprise. The first channel processes the basic credit characteristics, while the second channel uses the weights to weight the operational characteristics.

[0023] In this embodiment, the credit scoring model differs from traditional models. Instead of relying solely on basic credit characteristics, it integrates a company's basic credit characteristics with weighted operational characteristics. For example, the credit score is calculated as: Credit Score = f(Static Basic Characteristics, ω(Price Cycle Phase) * Dynamic Operational Characteristics). This model can provide completely different risk assessments for the same company with the same set of basic credit and operational characteristics depending on the cycle phase. This allows for a shift in credit scoring from static to dynamic, and from reactive to forward-looking approaches, significantly improving risk warning capabilities during periods of price decline.

[0024] In this embodiment, multi-source heterogeneous data of an enterprise can be acquired. Then, based on the historical price data of the enterprise's agricultural products, the current price cycle phase of the agricultural products is determined. Furthermore, the weights of the operating characteristics corresponding to the aforementioned price cycle phase are obtained from the feature weight matrix. Finally, the basic credit characteristics, operating characteristics, and weights are input into the credit scoring model to obtain the enterprise's credit score. This approach overcomes the limitations of existing static scoring models. By identifying the price cycle phase of agricultural products and determining the weights of economic characteristics accordingly, dynamic adjustment of the weights can be achieved, making the scoring results forward-looking and adaptive. This addresses the problems of delayed risk identification and insufficient response to cyclical market fluctuations.

[0025] In one possible implementation, acquiring the enterprise's multi-source heterogeneous data includes: The type of agricultural product corresponding to the enterprise is determined by the enterprise's declaration data or industry classification labels; Historical price data of the agricultural products are collected in the open market over a historical period, wherein the length of the historical period is greater than a preset number of years.

[0026] In practical applications, a company's core business category, i.e., the main agricultural product type, can be determined through enterprise declarations or industry classification labels, thus ensuring the accuracy of the agricultural product type. Simultaneously, historical price data for this agricultural product can be collected from the open market, with a data frequency of daily or monthly, and a time span exceeding a preset period, such as no less than 10 years, to ensure the statistical validity of period identification.

[0027] In practical applications, a mapping relationship between "enterprise - main agricultural products" can be established to form the basic data structure of enterprise profile, providing a basis for subsequent matching of external cyclical signals.

[0028] In one possible implementation, determining the current price cycle phase of the agricultural product based on its historical price data includes: The historical price series of the agricultural product is adaptively decomposed using Integrated Empirical Mode Decomposition (EEMD) to obtain a set of intrinsic mode function (IMF) components and a trend term. Identify the principal period components that dominate price fluctuations from the IMF components, and extract the analytical signals corresponding to the principal period components through Hilbert transform; Based on the analyzed signal, the instantaneous phase and instantaneous amplitude at the current moment are determined; Based on the changing trend of the instantaneous phase and the relative level of the instantaneous amplitude, the current price cycle phase is determined according to the period phase division rules.

[0029] In practical applications, for the historical price series of each agricultural product, adaptive signal decomposition is performed using Ensemble Empirical Mode Decomposition (EEMD). The specific process can be as follows: add white noise of a specific amplitude to the original price signal, perform multiple EMD decompositions, and then perform ensemble averaging on the results to effectively suppress mode aliasing and obtain a set of stable intrinsic mode function (IMF) components and a trend term.

[0030] Then, the dominant cycle component driving price fluctuations can be selected from the decomposition results. The selection criteria are: the average cycle length of the IMF most closely resembles the industry-recognized cycle pattern, and its energy proportion is significantly higher than other components. Subsequently, a Hilbert transform is performed on the selected dominant cycle IMF component to obtain its analytical signal. z(t) = x(t) + jH[x(t)] Where x(t) is the principal period IMF, and H[x(t)] is its Hilbert transform result.

[0031] Therefore, the instantaneous phase θ(t) = arg(z(t)) and the instantaneous amplitude A(t) = |z(t)| can be calculated.

[0032] In this way, EEMD effectively suppresses mode aliasing by adding white noise and decomposing it multiple times, adaptively extracting intrinsic mode functions (IMFs) from nonlinear and non-stationary signals. The Hilbert transform further calculates the instantaneous frequency and amplitude of each IMF, revealing the multi-scale time-frequency characteristics of price fluctuations. If traditional methods like Fourier transform are used, agricultural product prices are affected by multiple factors such as supply and demand, policy, and climate, exhibiting nonlinear and non-stationary characteristics (e.g., a sudden drought causing a short-term price surge). Fourier transform cannot capture these time-varying frequency characteristics and is prone to spectral leakage or spurious frequencies. Furthermore, if price data contains outliers caused by extreme weather, Fourier transform may misclassify noise as periodic components. However, EEMD, by integrating averaging to suppress noise, calculates instantaneous characteristics on pure IMFs using the Hilbert transform, resulting in more stable results and improved accuracy in phase determination.

[0033] In one possible implementation, the rule for dividing the periodic phase is as follows: when the instantaneous phase is continuously rising and the amplitude is in the growth stage, it is determined to be the rising period; when the amplitude reaches a local maximum and the phase change slows down, it is determined to be the peak period; when the phase is continuously falling and the amplitude is at a high level, it is determined to be the falling period; when the amplitude is at a local minimum and the phase tends to be stable, it is determined to be the trough period.

[0034] In practical applications, the instantaneous frequency f(t) can be obtained by taking the derivative of the instantaneous phase. Combining instantaneous amplitude and first-order difference trend, the cycle is divided into four phases: when the instantaneous frequency is high and the amplitude is rising, it is determined as the "rising phase"; when the amplitude reaches its peak and the change slows down, it is the "peak phase"; when the amplitude decreases and the frequency remains high, it is the "falling phase"; and when the amplitude is at a trough and tends to stabilize, it is the "trough phase". In practical applications, the number of phases can be determined according to the actual situation; the above is only an example. Thus, the current price cycle phase can be determined according to the above rules. It should be noted that all these calculations are based on historical data and are purely "measurements," not "predictions." In one possible implementation, the feature weight matrix is ​​constructed in the following way: Group the historical samples according to their period phase; The prediction model was trained on each set of data, where the dependent variable was whether a default event occurred, and the independent variables were basic credit characteristics and business characteristics. Once the training of the prediction model is complete, the regression coefficients of each feature in the prediction model corresponding to each set of data are extracted. The regression coefficients are arranged to obtain the feature weight matrix.

[0035] In the above embodiments, the feature weight matrix can be established by statistically analyzing the correlation strength between each feature and default events at different phases in historical data. In practical applications, it can be stored in the form of a weight matrix or function. In practical applications, the above prediction model can be a Cox proportional hazards model or a logistic regression model. In the above embodiments, the prediction model is a statistical tool used to discover patterns from historical data, that is, which features are truly important at different cycle phases, and to what extent. The essence of the feature weight matrix W is to solidify and store the patterns (regression coefficients) learned by the prediction model in tabular form. Within each group, all relevant basic credit features and operational features (xi, xi, xi) can be included. i and y j () is used as the independent variable.

[0036] In practical applications, the dependent variable in the Cox proportional hazards model is the "time from the observation point to the occurrence of default." It is more granular, not only determining whether a default has occurred but also characterizing the speed at which a default occurs. During a downturn, high-risk firms may default more quickly, and the Cox proportional hazards model can capture this "time risk." The Cox proportional hazards model can output a coefficient for each feature. This coefficient reflects the impact of that feature on the "instantaneous default risk rate." The coefficient can be represented by the value at the corresponding position in the feature weight matrix W.

[0037] The dependent variable in a logistic regression model is "whether the company will default within a fixed time window (e.g., 12 months)." The task of a logistic regression model is to train a model using historical data within each periodic phase group to predict "whether the company will default within a fixed period (e.g., 12 months)." After the logistic regression model is trained, a coefficient can be calculated for each feature. This coefficient is the "weight" of that feature in that specific phase, making it simpler and more intuitive compared to the Cox proportional hazards model.

[0038] In this way, by training the model and obtaining the regression coefficients of each feature variable under different phases, a feature weight matrix can be formed. In this matrix, 4 represents four cyclical phases, and n is the feature dimension. An element W[i,j] represents the contribution (regression coefficient) of a single unit change in the j-th feature (e.g., j=5 representing "inventory turnover") to the company's default risk when the market is in the i-th phase (e.g., i=3 representing a decline). The values ​​(weights) of the same feature can be completely different in different rows. The differences in the weights of operational features across rows are generally very large, which is the source of the model's "dynamic nature." In practical applications, during periods of rising prices, the value of inventory increases, and its weight can be appropriately reduced (because asset appreciation buffers risk); while during periods of falling prices, the weight of inventory needs to be increased (because the risk of price declines intensifies). During the peak of the cycle, companies have ample cash flow and can reduce the weight of the current ratio, focusing instead on the potential for capacity expansion; during the trough, the weight of cash flow needs to be strengthened to assess solvency.

[0039] In one possible implementation, the credit scoring model uses the following formula:

[0040] Where Pb is the probability of default, and σ(·) is the Sigmoid function. The intercept term of the credit scoring model is... The fixed weights are those of the i-th basic credit feature. The i-th basic credit feature, ω j (phase) represents the weight corresponding to the j-th operational characteristic. This refers to the j-th operational characteristic.

[0041] In practical applications, the above formula clearly embodies the dual-channel fusion concept of the credit scoring model. The first channel, the static credit channel, uses linear weighted summation to represent the inherent, stable risk fundamentals of a company across cycles. Regardless of market conditions, companies with a high history of defaults, high debt, and weak guarantees tend to have higher base risk scores. Moreover, It is globally fixed and does not change with the cycle, which reflects the "rigid" part of the basic credit risk.

[0042] The second channel, namely the dynamic operation channel, can employ linear weighting with table lookup modulation, ω j The phase () can be a function that takes the current phase as input and outputs the weight value of the j-th business characteristic under that phase, obtained from a table in the weight matrix W. This makes the contribution of the business characteristic to the final score dynamically change with the period phase. This term quantifies the additional risk inherent in the company's business behavior under the current market environment. Finally, the static risk and dynamic risk are added together to obtain a comprehensive risk score, which is then converted into a default probability Pb ​​between 0 and 1 using the Sigmoid function. This is more in line with the intuition of financial risk control and is easier to link with risk pricing (e.g., PD is used to calculate the expected loss EL).

[0043] In one possible implementation, after obtaining the credit score of the enterprise output by the credit scoring model, the method further includes: Based on the credit score, the company's cyclical risk level and cyclical sensitivity indicators are determined; Based on the credit score, the periodic risk level, and the periodic sensitive item prompts, a credit analysis report for the enterprise is generated.

[0044] In the above embodiments, the cyclical risk level and cyclical sensitivity indicators of an enterprise can be determined based on the credit score obtained from the credit assessment model, thereby generating a credit analysis report. In practical applications, a visual report can be generated, showing the enterprise's position in the cyclical phase diagram, historical score change trends, and risk warning suggestions corresponding to the cyclical sensitivity indicators, enhancing the interpretability and user trust of the credit scoring model.

[0045] In one possible implementation, after generating the credit analysis report for the enterprise, the method further includes: Obtain the default events that occurred in the enterprise within a preset time period; Based on the default event, the feature weight matrix and credit scoring model are retrained.

[0046] In practical applications, a periodic update mechanism can be set up to rerun the EEMD-Hilbert analysis monthly, updating the current phase of each agricultural product. Simultaneously, newly occurring default events are collected as feedback data to retrain the model and optimize the feature weight matrix W and the credit scoring model. When a company changes its main product, the cyclical signal source can be automatically switched. This closed-loop process of "perception—scoring—early warning—validation—optimization" ensures that the model continuously adapts to market changes and improves long-term predictive performance.

[0047] Experiments show that, compared to traditional models (such as...), Figure 2 As shown, on the axis corresponding to the horizontal axis 0.2, the line above indicates the proportion of bad samples, and the line below indicates the proportion of good samples. The credit scoring method for agricultural enterprises proposed in this application embodiment (such as...) Figure 3 As shown, on the axis corresponding to the horizontal axis 0.2, the line above indicates the proportion of bad samples, and the line below indicates the proportion of good samples. On average, the KS index is improved by more than 25%, the false rejection rate of good customers is reduced by 12%, and the model's discrimination ability is significantly improved.

[0048] Figure 4 This diagram illustrates the credit scoring process for agricultural enterprises in one application scenario provided by this application. Figure 4 As shown, the credit scoring of agricultural enterprises in this embodiment may further include the following execution process: Step 1: Obtain price data and external variable data (operating characteristics) for agricultural products; Step 2: Based on the EEMD-Hilbert transform and the price data of agricultural products, determine the current price cycle phase of the agricultural products. Step 3: Construct a periodic phase-dependent feature weight matrix based on historical samples; The current phase label is used to identify the price cycle phase described in the historical samples.

[0049] Step 4: Construct a credit scoring model. The credit scoring model needs to include the basic credit characteristics and economic characteristics of the enterprise, as well as the weights of the economic characteristics corresponding to the price cycle phase in the feature weight matrix. Step 5: Input the company's basic credit characteristics, economic characteristics, and the weights of the economic characteristics corresponding to the current price cycle phase into the credit scoring model to obtain the credit score and risk interpretation. Step 6: Obtain feedback data from subsequent enterprises and optimize the feature weight matrix and credit scoring model.

[0050] The above are embodiments of the method proposed in this application. Based on the same inventive concept, embodiments of this application also provide a credit scoring device for agricultural enterprises, the structure of which is as follows: Figure 5 As shown.

[0051] Figure 5 This is a schematic diagram of the internal structure of a credit scoring device for agricultural enterprises provided in an embodiment of this application. Figure 5 As shown, the device includes: At least one processor 501; And a memory 502 that is communicatively connected to at least one processor; The memory 502 stores instructions that can be executed by at least one processor. The instructions are executed by at least one processor 501 so that at least one processor 501 can: execute the above-mentioned credit scoring method for agricultural enterprises.

[0052] Some embodiments of this application provide corresponding to Figure 1 A non-volatile computer storage medium stores computer-executable instructions, which are configured to execute the aforementioned credit scoring method for agricultural enterprises.

[0053] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for IoT devices and media are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0054] The systems, media, and methods provided in this application are one-to-one correspondences. Therefore, the systems and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0055] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0056] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0057] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0058] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0059] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0060] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0061] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0062] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0063] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of this application should be included within the scope of the claims of this application.

Claims

1. A credit scoring method for agricultural enterprises, characterized in that, include: Acquire multi-source heterogeneous data of enterprises, wherein the enterprises are agricultural enterprises, and the multi-source heterogeneous data includes: the basic credit characteristics and operating characteristics of the enterprises, the types of agricultural products corresponding to the enterprises, and the historical price data of the agricultural products; Based on the historical price data of the agricultural product, determine the current price cycle phase of the agricultural product; From the feature weight matrix, obtain the weights corresponding to the operating characteristics under the price cycle phase, wherein the feature weight matrix includes weights corresponding to the operating characteristics under at least one phase; The basic credit features, the business features, and the weights are input into the credit scoring model to obtain the credit score of the enterprise output by the credit scoring model. The credit scoring model adopts a dual-channel fusion structure to score the enterprise. The first channel processes the basic credit features, and the second channel is used to use the weights to weight the business features.

2. The method according to claim 1, characterized in that, The acquisition of multi-source heterogeneous data from enterprises includes: The type of agricultural product corresponding to the enterprise is determined by the enterprise's declaration data or industry classification labels; Historical price data of the agricultural products are collected in the open market over a historical period, wherein the length of the historical period is greater than a preset number of years.

3. The method according to claim 1, characterized in that, Determining the current price cycle phase of the agricultural product based on its historical price data includes: The historical price series of the agricultural product is adaptively decomposed using Integrated Empirical Mode Decomposition (EEMD) to obtain a set of intrinsic mode function (IMF) components and a trend term. Identify the principal period components that dominate price fluctuations from the IMF components, and extract the analytical signals corresponding to the principal period components through Hilbert transform; Based on the analyzed signal, the instantaneous phase and instantaneous amplitude at the current moment are determined; Based on the changing trend of the instantaneous phase and the relative level of the instantaneous amplitude, the current price cycle phase is determined according to the period phase division rules.

4. The method according to claim 3, characterized in that, The rules for dividing the periodic phase are as follows: when the instantaneous phase continues to rise and the amplitude is in the growth stage, it is determined to be the rising phase; when the amplitude reaches a local maximum and the phase change slows down, it is determined to be the peak phase; when the phase continues to fall and the amplitude is at a high level, it is determined to be the falling phase; when the amplitude is at a local minimum and the phase tends to be stable, it is determined to be the trough phase.

5. The method according to claim 1, characterized in that, The feature weight matrix is ​​constructed in the following way: Group the historical samples according to their period phase; The prediction model was trained on each set of data, where the dependent variable was whether a default event occurred, and the independent variables were basic credit characteristics and business characteristics. Once the training of the prediction model is complete, the regression coefficients of each feature in the prediction model corresponding to each set of data are extracted. The regression coefficients are arranged to obtain the feature weight matrix.

6. The method according to claim 1, characterized in that, The formula used in the credit scoring model is as follows: Where Pb is the probability of default, and σ(·) is the Sigmoid function. The intercept term of the credit scoring model is... The fixed weights are those of the i-th basic credit feature. The i-th basic credit feature, ω j (phase) represents the weight corresponding to the j-th operational characteristic. This refers to the j-th operational characteristic.

7. The method according to claim 1, characterized in that, After obtaining the credit score of the enterprise output by the credit scoring model, the method further includes: Based on the credit score, the company's cyclical risk level and cyclical sensitivity indicators are determined; Based on the credit score, the periodic risk level, and the periodic sensitive item prompts, a credit analysis report for the enterprise is generated.

8. The method according to claim 7, characterized in that, After generating the credit analysis report for the enterprise, the method further includes: Obtain the default events that occurred in the enterprise within a preset time period; Based on the default event, the feature weight matrix and credit scoring model are retrained.

9. A credit scoring device for agricultural enterprises, characterized in that, The device includes: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform a credit scoring method for agricultural enterprises as described in any one of claims 1-8.

10. A computer storage medium storing computer-executable instructions, characterized in that, When the computer-executable instructions are executed, they implement a credit scoring method for agricultural enterprises as described in any one of claims 1-8.

Images