Supply chain collaboration-oriented fund flow anomaly monitoring and early warning method

Abstract

The present application relates to the technical field of supply chain collaboration and fund flow monitoring, in particular to a fund flow abnormality monitoring and early warning method for supply chain collaboration, receiving logistics performance data, transaction data and fund settlement data of multiple node enterprises in the supply chain, calculating logistics performance uncertainty features and performing multi-scale decomposition processing, screening low-frequency component data and calculating to generate logistics uncertainty features; extracting credit behavior features from the transaction data, inputting the logistics uncertainty features into a time delay coupling analysis model to obtain a time delay response relationship; constructing a supply chain credit network model containing node weights and edge weight relationships according to the time delay response relationship, and performing structure parameter extraction, change detection processing and phase change determination processing on the supply chain credit network model to generate a supply chain credit network state result, and finally generating fund flow abnormality early warning information. The present application can realize early identification and early warning of fund flow abnormalities in the supply chain, and improve the accuracy and forward-looking nature of fund flow risk monitoring.

Description

Technical Field

[0001] This invention relates to the field of supply chain collaboration and cash flow monitoring technology, specifically a method for monitoring and early warning of abnormal cash flow in supply chain collaboration. Background Technology

[0002] In business, a supply chain is a network-like collaborative system composed of upstream and downstream enterprises, with cash flow being the cash transfer process that accompanies it, including payments for goods, accounts receivable and payable, and financing settlements. By gaining insights into the patterns of the flow of people, goods, and money, and using data models to track key indicators such as cash recovery, payment defaults, payment term deviations, and liquidity gaps in real time, we can accurately identify risks such as funding disruptions, credit deterioration, or fraudulent transactions and issue early warnings to ensure the stability and sustainability of supply chain collaboration.

[0003] In existing technologies, anomaly monitoring of supply chain cash flow typically relies on analysis of transaction data itself. This includes statistical modeling or rule-based judgment of transaction amounts, frequencies, paths, and account behavior characteristics to identify abnormal transaction behaviors. Furthermore, some solutions incorporate inter-enterprise transaction relationships, analyzing the flow paths of funds across different entities to improve the accuracy of anomaly identification. These methods primarily depend on cash flow data or transaction information directly related to cash flow for identifying and issuing warnings of abnormal behavior.

[0004] However, in actual supply chain operations, changes in cash flow are not only affected by the transactions themselves but also by external factors during the supply chain fulfillment process. Among these, the instability of logistics fulfillment has a significant impact on the cash flow structure. Specifically, when the logistics system in the supply chain transitions from a stable to an unstable state—for example, when transportation times fluctuate significantly or delivery cycles become unpredictable—even if a single delay is not significant, the uncertainty of the overall fulfillment process will gradually accumulate, thereby affecting the trust relationship between supply chain entities. In this situation, companies typically do not immediately cease transactions but instead respond by adjusting their credit strategies, such as shortening payment terms, increasing prepayment ratios, or reducing credit transactions.

[0005] As multiple supply chain entities respond to similar uncertainties in fulfillment, changes in individual credit behavior will have a ripple effect on the supply chain network, leading to systemic changes in the cash flow structure, such as an increase in the proportion of advance settlement and a decrease in the proportion of credit transactions. Ultimately, this may trigger an overall contraction of the supply chain credit system. This process of change has obvious non-linear characteristics, meaning that the changes are relatively slow in the initial stage, but a comprehensive and sudden structural transformation will occur after certain conditions are met.

[0006] Existing technologies primarily focus on the characteristics of individual transactions or localized cash flows, lacking modeling of external disturbances such as logistical fulfillment uncertainty. They also fail to reveal the impact path of changes in logistical stability on supply chain credit behavior, and are unable to identify the resulting structural abrupt changes in cash flows. Therefore, during the transition from a stable to an unstable logistics system, existing methods struggle to capture the evolutionary trends of supply chain credit relationships in a timely manner, resulting in an inability to provide early warnings of potential systemic risks, leading to monitoring lags or insufficient identification. Summary of the Invention

[0007] This application provides a method for monitoring and early warning of abnormal cash flow in supply chain collaboration, which solves the problem that abnormal cash flow in the supply chain is difficult to identify and warn in a timely manner in the existing technology.

[0008] To achieve the above objectives, this application discloses the following technical solution:

[0009] This solution discloses a method for monitoring and early warning of abnormal cash flow in supply chain collaboration, including the following steps: Step 1: Receive logistics fulfillment data, transaction data, and fund settlement data from multiple nodes in the supply chain. The logistics fulfillment data includes transportation time series, delivery completion records, and node delay records. Perform time series processing on the logistics fulfillment data to calculate the logistics fulfillment uncertainty characteristics. Then, perform multi-scale decomposition processing on the logistics fulfillment uncertainty characteristics to obtain data of different frequency band components. Based on the data of different frequency band components, filter out low-frequency component data with frequencies below a preset threshold, and calculate the fluctuation amplitude and trend of the low-frequency component data to generate logistics uncertainty characteristics that characterize the stability changes of the logistics system. The logistics performance uncertainty features calculated in step 1 are decomposed by multi-scale, screened for low-frequency components and enhanced for stability, and then used to generate logistics uncertainty features for subsequent analysis. These features serve as the input to the time-delay coupling analysis model in step 2. Step 2: Perform feature extraction processing on the transaction data to generate credit behavior features. Input the logistics uncertainty features obtained in Step 1 into the preset time-delay coupling analysis model and perform correlation calculation with the credit behavior features to obtain the time-delay response relationship between logistics changes and credit behavior. The credit behavior features include payment period change data, prepayment ratio data, and credit ratio data. Step 3: Construct a supply chain credit network model based on the time-delay response relationship. The supply chain credit network model includes node weights and edge weight relationships. The node weights are determined by fund settlement data and logistics uncertainty characteristics, while the edge weight relationships are determined by credit behavior characteristics. The time-delay response relationship is used to align the dynamic correlation between credit behavior characteristics and logistics uncertainty characteristics in time, so that when constructing the supply chain credit network model, the node edge weights can reflect the credit transmission strength after time-delay adjustment. Step 4: Extract structural parameters from the supply chain credit network model, perform change detection processing based on the structural parameters to generate network structure change features, and perform phase transition determination processing based on the network structure change features to generate the supply chain credit network state result. Step 5: Generate early warning information on abnormal cash flow based on the status results of the supply chain credit network.

[0010] By introducing multi-scale decomposition of the uncertainty characteristics of logistics fulfillment, the original time series data is decoupled into different frequency band components at different time scales, thereby avoiding the averaging or masking of fluctuation information at a single time scale. Among them, low-frequency component data can reflect the long-term trend evolution characteristics of the system, while high-frequency component data corresponds to random disturbances and short-term fluctuations. By screening low-frequency component data and calculating its fluctuation amplitude and trend, the ability to mathematically characterize the changes in system stability can be enhanced, enabling subsequent time-delay coupling analysis models to perform correlation modeling based on more stable feature inputs, thereby improving the robustness and discriminative ability of the model.

[0011] Furthermore, in order to overcome the problem that traditional methods based on a single statistical measure are insufficient to characterize complex logistics fluctuations, the calculation of logistics performance uncertainty features introduces fluctuation indicators and dispersion indicators for joint characterization, and further quantifies the degree of disorder of the system through entropy calculation. This enables the constructed logistics performance uncertainty features to not only reflect the magnitude of numerical fluctuations, but also the structural complexity of the time series, thereby providing more information-expressive feature inputs for subsequent coupled analysis between logistics and credit behavior.

[0012] Furthermore, in order to effectively separate high-frequency information caused by random disturbances from low-frequency information reflecting structural changes in logistics fulfillment data, this scheme decomposes the original data through frequency domain transformation and performs random disturbance filtering on the high-frequency components, thereby reducing the impact of random noise on the analysis results. At the same time, stability enhancement calculations are performed on the low-frequency components to make them more concentrated in reflecting the long-term evolution trend of the system, thereby ensuring that the data input to the time-delay coupling analysis model has stronger structural consistency and stability.

[0013] Furthermore, considering the significant time lag effect between logistics changes and credit behavior, this scheme captures the dynamic response relationship between the two by calculating the correlation within multiple time windows, and further achieves time alignment of the causal relationship by calculating the optimal time lag parameter. The mapping relationship established in this way can avoid the misjudgment problem caused by time mismatch in traditional synchronous analysis methods, making the correlation between logistics uncertainty characteristics and credit behavior characteristics more consistent with the delay transmission mechanism in actual supply chain operation.

[0014] Furthermore, in order to map the relationship between logistics and credit into an analyzable structured expression, this solution generates credit connection relationships between nodes based on credit behavior characteristics, and calculates node weights by combining fund settlement data and logistics uncertainty characteristics, so that node weights can reflect the relative influence of enterprises in the fund flow network. By combining credit connection relationships with node weights, a weighted network structure is constructed, so that the supply chain credit network model can simultaneously represent the strength of relationships and the importance of nodes, thereby improving the ability to characterize the overall structure of the system.

[0015] Furthermore, in order to capture the structural evolution of the supply chain credit network in a timely manner during operation, this scheme periodically samples the network structure parameters and continuously analyzes their changing trends to identify abrupt change points in the structural evolution process. These abrupt change points, as key indicators of network structure changes, can reflect the transition process of the system from a stable state to an unstable state, thus providing key input basis for subsequent phase transition determination.

[0016] Furthermore, to avoid the risk of misjudgment caused by relying on a single indicator for state determination, this solution introduces a spectral feature analysis method in addition to threshold judgment. By extracting network structure feature vectors and detecting changes in their feature values, the stability of the network can be characterized from the algebraic structure level. When the network structure change feature and the spectral feature change simultaneously meet the preset conditions, it can more reliably determine that the supply chain credit network is in a phase transition state, thereby improving the robustness and credibility of the judgment results.

[0017] Furthermore, in order to achieve effective expression and transmission of risk information, this solution calculates risk level parameters based on the status results of the supply chain credit network and identifies key node enterprises by combining network structure analysis. This enables the early warning information to not only include the overall risk level, but also to indicate the key propagation path of risk in the network, thereby improving the pertinence and decision support capabilities of the early warning information.

[0018] Furthermore, to improve the reliability of anomaly identification, this solution simulates the stable state of logistics by constructing a benchmark model and compares the actual cash flow data with the simulation results to obtain deviation data. By processing the deviation data for anomaly judgment, the system can be supplemented from the perspective of "deviation from normal operating state", so that the early warning judgment no longer depends on the output of a single model, but combines the results of counterfactual analysis to improve the robustness of the overall judgment.

[0019] Furthermore, in order to achieve early identification of the critical state of the system, this scheme constructs a potential function model to quantitatively characterize the stability of the supply chain credit network and uses the recovery capability change parameter to characterize the system's self-recovery capability. When the recovery capability change parameter continues to rise and its value increases, it indicates that the system's recovery capability is declining, thereby generating a critical approach indicator. This indicator is used to trigger early warning of abnormal cash flow, enabling the early warning mechanism to be upgraded from "outcome-driven" to "trend-driven", thus enhancing the system's forward-looking risk identification capability.

[0020] This invention addresses the problem in existing technologies that rely solely on transaction data analysis, which struggles to promptly capture structural changes in supply chain credit behavior caused by uncertainties in logistics fulfillment. This leads to delayed identification of systemic risks such as funding disruptions and credit deterioration. The invention employs time-series processing and multi-scale decomposition of logistics fulfillment data to generate logistics uncertainty features characterizing changes in the stability of the logistics system. These features, along with credit behavior features extracted from transaction data, are input into a time-lag coupling analysis model to establish a time-lag response relationship between logistics changes and credit behavior. Furthermore, a supply chain credit network model incorporating node weights and edge weights is constructed. Based on changes in structural parameters and phase transitions, the model identifies abrupt changes in the network structure, thus depicting the nonlinear evolution of cash flow from dispersion to contraction. Finally, based on the supply chain credit network state results, anomaly warning information for cash flow is generated. Combined with counterfactual analysis and critical state detection, this enables early identification and warning of systemic risks in the supply chain, effectively solving the problems of monitoring lag and insufficient structural risk identification in existing technologies. Attached Figure Description

[0021] Figure 1 This is a flowchart of a method according to an embodiment of the present invention; Figure 2 This is a multi-scale decomposition process according to an embodiment of the present invention; Figure 3 This is a diagram of the time-delay coupling analysis model according to an embodiment of the present invention; Figure 4 This is a diagram illustrating the construction of a supply chain credit network model according to an embodiment of the present invention. Figure 5 This is a diagram illustrating the phase transition determination process in an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. In the following description, numerous specific details are set forth to provide a comprehensive understanding of the present invention. The present invention may be practiced without some or all of these specific details. In other instances, well-known processes have not been described in detail to avoid unnecessarily obscuring the present invention.

[0023] When used in conjunction with the terms "comprising," "method comprising," or similar language in this specification and appended claims, the singular forms "a," "some," and "the" include plural references unless the context clearly indicates otherwise. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0024] Terminology definition: "Logistics fulfillment uncertainty characteristics" refers to the calculation of second-order or higher moments of physical flow data such as transportation time and delivery nodes, which is used to measure the degree of dispersion and unpredictability of the logistics system deviating from the predetermined stable state.

[0025] The "time-delay coupling analysis model" is an algorithmic structure specifically designed to process signal sequences with time-difference correlations. It can identify the specific delay period by which disturbances in the logistics system are transmitted to cash flow credit strategies (such as payment term adjustments) after being decided by the enterprise's management.

[0026] The "supply chain credit network model" abstracts the node enterprises in the supply chain as vertices in graph theory, and abstracts the credit transactions between enterprises, such as credit sales and prepayments, as weighted edges, thereby constructing a global dynamic topology.

[0027] "Credit behavior characteristics" refers to a set of characteristic parameters extracted from transaction data to characterize a company's credit strategy and funding arrangements. These parameters include at least data on changes in payment terms, prepayment ratios, and credit ratios, and are used to reflect a company's credit expansion or contraction behavior during transactions.

[0028] "Credit network phase transition state" refers to the critical state in which the network, represented by the model, transitions from a stable state to an unstable state when the structural parameters and spectral characteristics of the supply chain credit network model undergo abrupt changes.

[0029] Example 1

[0030] Step 1: Receive logistics fulfillment data, transaction data, and fund settlement data from multiple nodes in the supply chain. Logistics fulfillment data includes transportation time series, delivery completion records, and node delay records. Perform time series processing on the logistics fulfillment data to calculate logistics fulfillment uncertainty characteristics, and then perform multi-scale decomposition processing on these characteristics to obtain data from different frequency bands. Perform frequency domain transformation processing on the logistics fulfillment data to decompose it into high-frequency and low-frequency components. Perform random perturbation filtering on the high-frequency components to obtain the filtered data. Perform stability enhancement calculations on the low-frequency components to obtain structural fluctuation data. Then, process the structural fluctuation data... Data is input into the time-delay coupling analysis model as a characteristic of logistics uncertainty. Based on data from different frequency bands, low-frequency component data with frequencies below a preset threshold is selected, and the fluctuation amplitude and trend of the low-frequency component data are calculated to generate logistics uncertainty characteristics that characterize the stability changes of the logistics system. The preset threshold is a frequency boundary used to distinguish between low-frequency component data and high-frequency component data, which can be set according to the time scale, sampling period, or historical spectrum distribution characteristics of the logistics fulfillment data. Preferably, the preset threshold is adaptively determined according to the dominant frequency distribution or energy concentration range of the logistics fulfillment data, so that the selected low-frequency component data can reflect the long-term trend of the logistics system.

[0031] Receive time-series data from logistics fulfillment data; calculate volatility and dispersion indices for the time-series data; combine the volatility and dispersion indices to obtain basic uncertainty characteristics; perform entropy calculation on the basic uncertainty characteristics to obtain fulfillment entropy values; fuse the fulfillment entropy values ​​with the basic uncertainty characteristics to generate logistics fulfillment uncertainty characteristics; and output the logistics fulfillment uncertainty characteristics to subsequent processing steps.

[0032] In this embodiment, the system receives logistics fulfillment data, transaction data, and fund settlement data from the supply chain management platform or blockchain nodes through a standard data interface, and performs time series processing on the transportation time series, delivery completion records, and node delay records. Subsequently, the time series data is de-trended and normalized, and the volatility index and dispersion index are calculated based on a sliding window. The basic uncertainty characteristics are obtained by combining them.

[0033] Considering that the logistics fluctuation indicators (reflecting the intensity of change) and dispersion indicators (reflecting the breadth of distribution) affecting supply chain stability are not simply linear superpositions, but rather have a mutually reinforcing "stress effect," this embodiment adopts a weighted coupling mapping relationship based on the Minkowski paradigm to construct the basic uncertainty characteristic calculation equation as follows: ; in, This represents the fundamental uncertainty characteristic value, used to characterize the overall deviation of the current logistics fulfillment status from historical benchmarks; This indicator represents the fluctuation of logistics fulfillment data within the current time window. It is obtained through statistical calculation of transportation time series data, preferably the standard deviation of transportation time or delivery duration within the time window. This represents the historical benchmark volatility index, which is obtained by statistically averaging the volatility index of historical logistics fulfillment data during the stable operation phase. This index represents the dispersion of logistics fulfillment data within the current time window, and it is obtained by calculating the coefficient of variation or coefficient of dispersion based on the transportation time series. This represents the historical baseline dispersion index, which is obtained by statistically averaging the dispersion index during historical stable periods. and The weighting coefficients for the volatility index and the dispersion index are preset parameters that can be obtained through statistical analysis of historical logistics fulfillment data. In one implementation, the mean or variance of the volatility index and the dispersion index can be calculated based on the degree of fluctuation in transportation time and the stability of delivery for node enterprises within a historical period, and the weighting coefficients can be normalized based on their relative proportions, thereby... and The sum is 1; k represents the sensitivity adjustment parameter, which is a preset parameter used to adjust the amplification of abnormal fluctuations. Its value is not less than 2. In one implementation, k is an integer between 2 and 4. By conducting multiple sets of parameter comparison experiments on historical data, the value that makes the logistics performance uncertainty characteristics respond most obviously to abnormal events is selected. All the above parameters are derived from logistics performance data and historical statistical data.

[0034] By calculating the fundamental uncertainty eigenvalues This allows for the acquisition of quantitative indicators reflecting the degree of change in the stability of the logistics fulfillment system. An increase in the value indicates a significant deviation of the current logistics performance status from the historical benchmark. This value serves as the basic input for subsequent calculation of performance entropy and generation of logistics performance uncertainty characteristics, thereby providing stable and interpretable data support for the time-delay coupling analysis model.

[0035] Further entropy calculations are performed on the fundamental uncertainty characteristics to obtain the performance entropy value; To quantify the degree of disorder in the distribution of logistics fulfillment uncertainty characteristics over time and to characterize the increasing randomness during the evolution of the logistics system from a stable to an unstable state, fulfillment entropy is introduced as a metric. By calculating the information entropy of the probability distribution of basic uncertainty characteristics, a quantitative characterization of the complexity of logistics fulfillment states is achieved, thus providing statistically significant input characteristics for subsequent time-delay coupling analysis models. This embodiment constructs a fulfillment entropy calculation function based on a normalized probability distribution: ; in, It represents the performance entropy value, used to characterize the overall degree of disorder in the uncertainty characteristics of logistics performance; This represents the basic uncertainty characteristic value obtained statistically within the i-th interval. This basic uncertainty characteristic value is obtained by calculating the volatility index and dispersion index and combining them after processing the logistics performance data through time series. This represents the normalized weight of the fundamental uncertainty eigenvalue in the i-th interval, which is obtained by normalizing the fundamental uncertainty eigenvalue in each interval. This represents the sum of all basic uncertainty characteristic values ​​within the current time window, used for probability normalization; n represents the total number of intervals after dividing the basic uncertainty characteristic values ​​into intervals. This interval division is obtained by discretizing the historical distribution range of logistics fulfillment data according to a preset resolution. The preset resolution is the interval division granularity set according to the distribution characteristics of historical data. Discretization processing includes equal-interval division, equal-frequency division, or adaptive division based on data density. All the above parameters are derived from the results of time series processing, statistical calculation, and interval division processing of logistics fulfillment data.

[0036] By calculating the performance entropy value This allows us to obtain a distribution complexity index of logistics fulfillment uncertainty characteristics within the current time window. When the value increases, it indicates that the logistics fulfillment status is becoming more disordered and volatile, reflecting a decline in the stability of the supply chain logistics system; when When the value decreases, it indicates that the logistics fulfillment status tends to be concentrated and stable. Therefore, this indicator can serve as an important input in the subsequent time-delay coupling analysis model to judge the degree of impact of logistics changes on credit behavior, and can be used to support the identification of risk evolution process and early warning of abnormal cash flow in the supply chain credit network model.

[0037] Subsequently, the performance entropy value is fused with the basic uncertainty characteristics to generate logistics performance uncertainty characteristics. Based on this, frequency domain transformation processing is performed on the logistics performance data to decompose it into high-frequency components and low-frequency components. The high-frequency components are used to characterize random disturbances, while the low-frequency components are used to characterize structural trends. Random disturbance filtering processing is performed on the high-frequency components, which can be achieved by moving average or probabilistic filtering methods. Stability enhancement calculation is performed on the low-frequency components to extract structural fluctuation data, and low-frequency component data with frequencies below a preset threshold are selected and their fluctuation amplitude and trend are analyzed to generate logistics uncertainty characteristics that characterize the stability changes of the logistics system.

[0038] By decoupling logistics data from both the time and frequency domains, this step effectively distinguishes between occasional disturbances and long-term trend changes. The introduction of entropy values ​​quantifies the unpredictability of the logistics system, thereby solving the problem that traditional statistical methods cannot characterize complex fluctuation structures. This results in the extracted logistics uncertainty features having stronger stability and interpretability.

[0039] Step 2: Perform feature extraction processing on the transaction data to generate credit behavior features. Input the logistics uncertainty features into the preset time-delay coupling analysis model and perform correlation calculation with the credit behavior features to obtain the time-delay response relationship between logistics changes and credit behavior. The time lag coupling analysis model is a time lag analysis model built based on the correlation calculation between the characteristics of logistics uncertainty and the characteristics of credit behavior. It is used to calculate the correlation between the two in multiple time windows and determine the optimal time lag parameters. The credit behavior characteristics include data on changes in payment terms, prepayment ratios, and credit sales ratios; it receives logistics uncertainty characteristics and credit behavior characteristics; it calculates the correlation between logistics uncertainty characteristics and credit behavior characteristics within multiple time windows; it calculates the optimal time lag parameter based on the correlation; it establishes a mapping relationship between logistics uncertainty characteristics and credit behavior characteristics based on the optimal time lag parameter; and it outputs the time lag response relationship to construct a supply chain credit network model.

[0040] The system extracts key fields such as payment period length, prepayment ratio, and credit ratio from transaction data and constructs a credit behavior feature vector. Then, it inputs the logistics performance uncertainty features and credit behavior features into the time lag coupling analysis model, calculates the correlation between the two in multiple time windows, and finds the optimal time lag parameter that maximizes the correlation by traversing different time lag parameters. Because the impact of logistics fluctuations on corporate credit strategies involves a significant psychological decision-making delay, this embodiment introduces a normalized cross-correlation operator within a preset time offset search space. The optimal time lag parameter is derived by performing an extremum search within the equation: ; in, The optimal time lag parameter is used to characterize the time delay by which the uncertainty of logistics performance affects the characteristics of credit behavior. This parameter is obtained by searching for the time offset with the highest correlation in the preset time offset search space. The candidate time lag parameters are derived from the preset time offset search space. The search space is a pre-defined set of time intervals, and its value range is determined according to the supply chain business cycle. The time offset search space is defined based on the transaction settlement cycle and credit behavior adjustment cycle in the supply chain. Specifically, it includes the payment period distribution range obtained from historical fund settlement data statistics, and is determined in combination with the preset maximum response period. The maximum response period covers the time range during which changes in logistics performance affect credit behavior. t represents the time series index, which is used to identify the sampling time points of logistics performance data and credit behavior data. This index is generated by the time series processing steps. This represents the total length of the time series, which is determined by the historical sampling period of logistics fulfillment data and transaction data. This represents the characteristic value of logistics fulfillment uncertainty at time index t. This characteristic value is obtained by performing time series processing, frequency domain decomposition, and entropy calculation on the logistics fulfillment data in step 1. This represents the mean of the uncertainty characteristics of logistics fulfillment, which is calculated by analyzing all data within a time window. Calculate the average; Indicates time index The credit behavior feature value at any given time is derived from the payment period change data, prepayment ratio data, and credit ratio data extracted from the transaction data, and is generated through feature extraction processing. This represents the mean of credit behavior characteristics, which is calculated by analyzing all data within the corresponding time window. The average is obtained; the numerator is used to calculate the degree of coordinated change between the uncertainty characteristics of logistics performance and the characteristics of credit behavior, and the denominator is used to normalize the coordinated change to eliminate the influence of different dimensions and scales.

[0041] By calculating the above formula, the optimal time lag parameter with the greatest correlation between the characteristics of logistics performance uncertainty and credit behavior can be obtained, thereby establishing the time lag response relationship between logistics changes and credit behavior. This result is used to construct the credit connection relationship in the supply chain credit network model and to identify whether changes in credit behavior are triggered by logistics instability, thereby improving the causal identification capability and early warning accuracy of abnormal cash flow monitoring.

[0042] After determining the optimal time lag parameter, a mapping relationship between the characteristics of logistics uncertainty and credit behavior is established, and the time lag response relationship is output.

[0043] This step is based on the objective fact that corporate credit behavior has a response lag. By introducing a time lag parameter, it is possible to identify the causal path of logistics fluctuations on credit strategy adjustments, thereby distinguishing between proactive credit management behavior and passive risk response behavior and improving the accuracy of credit behavior analysis.

[0044] Step 3: Construct a supply chain credit network model based on the time-delay response relationship. The supply chain credit network model includes node weights and edge weights. The node weights are determined by the fund settlement data and logistics uncertainty characteristics, while the edge weights are determined by the credit behavior characteristics. Receive the credit behavior characteristics and fund settlement data; generate credit connection relationships between nodes based on the credit behavior characteristics; combine the credit connection relationships and node weights to form a weighted network structure, thus generating the supply chain credit network model.

[0045] Based on time-delay response relationships, credit connection relationships are constructed with enterprises in the supply chain as nodes and credit behavior characteristics as the basis for edge weights. Node weights are calculated based on fund settlement data and logistics uncertainty characteristics. ; This represents the node weight of the i-th node enterprise, used to characterize the importance of that node in the supply chain credit network model; This indicates the scale of fund settlement for node enterprise i within a preset time window. This data is extracted from fund settlement data and obtained by summarizing or averaging the transaction settlement amounts of node enterprises within the statistical period. The preset time window and statistical period are set according to the time granularity of the fund settlement data, including fixed time intervals by day, week, or month. This indicates that a logarithmic transformation is performed on the capital scale to compress the order-of-magnitude differences between enterprises of different sizes, so that the node weights remain numerically stable. This parameter represents the cash collection cycle of node enterprise i. It is obtained by calculating the time difference between the order payment time and the actual arrival time. The data comes from the timestamp information of transaction data and fund settlement data. The logistics uncertainty characteristics associated with node enterprise i are obtained by multi-scale decomposition, low-frequency screening and stability enhancement processing of the logistics performance uncertainty characteristics calculated from the logistics performance data in step 1, and can be aggregated for logistics paths related to this node. This parameter represents the industry benchmark settlement cycle, which is obtained through statistical analysis of transaction settlement cycles in historical industry data. In the absence of historical data, the system initializes the setting based on a preset industry reference cycle, which is configured according to the settlement habits of the industry to which the supply chain belongs. exp(·) represents the exponential function operation, used to characterize the nonlinear suppression effect of settlement cycle and logistics uncertainty on node weights.

[0046] By calculating the node weights mentioned above, the relative importance distribution of each node enterprise in the supply chain credit network model can be obtained. When a node enterprise has a large scale of fund settlement and a short settlement cycle with low logistics uncertainty, its corresponding node weight is high, indicating that the node has a strong stabilizing support role in the network. Conversely, when the settlement cycle is extended or logistics uncertainty increases, the node weight will decrease significantly, thus manifesting as a potential risk node in the network structure change analysis. This result can be directly used for risk assessment calculations in the process of identifying key node enterprises and generating early warning information for abnormal fund flows.

[0047] By integrating credit connections with node weights to form a weighted network structure, a supply chain credit network model is constructed. The network structure can be represented in the form of an adjacency matrix, and the weights are dynamically adjusted by combining time delay parameters to reflect the time evolution characteristics of credit relationships.

[0048] By mapping credit behavior to network topology, this step transforms individual transaction behavior into overall system structure, enabling risk to propagate and be expressed in a structural form within the network, thus providing a foundation for subsequent structural analysis and risk identification.

[0049] Step 4: Extract structural parameters from the supply chain credit network model, perform change detection processing based on the structural parameters to generate network structure change features, and perform phase transition determination processing based on the network structure change features to generate the supply chain credit network state result; periodically sample the structural parameters of the supply chain credit network model; calculate the change trend of the structural parameters; detect abrupt changes in the change trend; and use the structural changes corresponding to the abrupt changes as network structure change features; perform threshold judgment on the network structure change features; extract network structure feature vectors based on the network structure change features, perform spectral feature analysis on the network structure feature vectors to obtain spectral feature data; detect feature value changes in the spectral feature data; when the network structure change features meet the preset abrupt change determination conditions and the spectral feature data meets the feature value change conditions, it is determined to be a credit network phase transition state; output the supply chain credit network state result.

[0050] The supply chain credit network model is periodically sampled to extract structural parameters, including node degree, clustering coefficient, and path length. Time series analysis is performed on the structural parameters to calculate their changing trends. A mutation detection algorithm is used to identify mutation points in the changing trends, and the structural changes corresponding to mutation points are defined as network structural change features. Feature vectors are constructed based on network structural change features, and spectral feature analysis is performed on these feature vectors to obtain spectral feature data. To characterize the structural stability of the supply chain credit network, this embodiment constructs a structural stability index based on the distribution of network eigenvalues: ; in, The network structure stability index reflects the overall structural stability of the supply chain credit network model. The larger the value, the more concentrated the core connection structure in the network and the higher the stability. The smaller the value, the more dispersed or broken the network structure is, and the more likely it is to evolve into an unstable state. This represents the largest eigenvalue of the structure matrix corresponding to the supply chain credit network model. It is used to characterize the concentration of dominant connection strength in the network. Specifically, it is obtained by performing eigenvalue decomposition on the network structure matrix corresponding to the supply chain credit network model constructed in step 3. The network structure matrix is ​​formed by the combination of node weights and edge weights. This represents the second-largest eigenvalue of the network structure matrix, used to characterize the strength of secondary connections besides the dominant connections. It is also obtained by performing eigenvalue decomposition on the network structure matrix. The network structure matrix is ​​constructed as follows: adjacency relationship data is generated based on the credit connections between nodes, and the connection strength is weighted according to node weights, thus forming the matrix input data for calculating the eigenvalues. The above indicators are then calculated. It can obtain the structural stability changes of the supply chain credit network model at different time sampling points, and can be further used to determine whether the network structure change characteristics meet the preset mutation conditions, thus serving as one of the bases for determining the phase transition state of the credit network.

[0051] When the network structure change characteristics meet the preset mutation conditions and the spectral feature data meet the feature value change conditions, the credit network is determined to enter the phase transition state, and the credit network state result is output.

[0052] The preset mutation judgment conditions include: when the change amplitude of the network structure change characteristics exceeds the preset change threshold or the change rate exceeds the preset rate threshold in multiple consecutive sampling periods, it is judged to meet the mutation judgment conditions. The change threshold and rate threshold are determined according to the stable state statistical distribution of the historical supply chain credit network model.

[0053] This step is based on the phase transition mechanism in complex network theory. By monitoring abrupt changes in network structure parameters and spectral characteristics, it can identify the transition of the supply chain system from a stable state to an unstable state, thereby capturing critical signals of systemic risks in advance.

[0054] Step 5: Output early warning information for abnormal cash flow based on the supply chain credit network status results; receive the supply chain credit network status results; calculate risk level parameters based on the supply chain credit network status results; identify key node enterprises in the supply chain credit network model; generate early warning data containing risk level parameters and key node enterprise information; output early warning data to the early warning terminal; construct a benchmark model based on historical data; simulate the stable state of logistics in the benchmark model and obtain simulation results; calculate the deviation data between the actual cash flow data and the simulation results; perform anomaly judgment processing on the deviation data; output the deviation judgment results as a supplementary basis for early warning; construct a potential function model to characterize system stability; perform stability assessment on the supply chain credit network model; calculate the recovery capability change parameter; when the recovery capability change parameter shows a continuous increase and the value increases, generate a critical approach indicator; and use the critical approach indicator to trigger early warning information for abnormal cash flow in advance.

[0055] The system calculates risk level parameters based on the credit network status results and identifies key node enterprises in the network; at the same time, it constructs a benchmark model based on historical data, simulates the stable state of logistics in the benchmark model and generates simulation results; it calculates the deviation data between actual cash flow data and simulation results and performs anomaly detection and processing.

[0056] To quantify the degree of deviation between actual cash flow and the baseline state, this embodiment constructs a deviation index based on distribution differences: ; in, This represents counterfactual bias data, used to characterize the overall difference between the actual distribution of cash flows and the simulated distribution of cash flows. The larger the value, the more significant the deviation of cash flows from the steady state. This represents the k-th interval state in the cash flow state space, used to discretize the cash flow. This state space is obtained by dividing the cash settlement data into intervals according to the cash flow scale or the magnitude of cash flow changes. This indicates that the actual cash flow data is in the status range. The probability distribution is obtained by statistically normalizing the real-time collected fund settlement data, specifically by statistically analyzing the data falling within a preset time window. The frequency of the cash flow data is obtained by performing probability processing. The preset time window is set according to the time granularity of the cash settlement data and the cash flow change cycle, including statistical intervals by day, week or month. This indicates that the distribution of funds simulated under the baseline model falls within the state interval. The probability value is obtained by constructing a benchmark model based on historical stable period data, running the model under stable conditions of logistics performance uncertainty to generate simulated cash flow data, and then performing the same interval statistics and normalization processing on the simulated data; k is the state interval index, and its value range is determined by the preset granularity of cash flow state division. The granularity of cash flow state division is set according to the historical distribution range of cash settlement data, including using equal-width division or equal-frequency division based on quantiles, to ensure that each state interval has statistical representativeness.

[0057] By calculating this counterfactual deviation data, the degree of deviation of the current cash flow status from the ideal stable state can be obtained. This result is used to perform anomaly detection processing. When the counterfactual deviation data exceeds a preset threshold, it is determined that there is a cash flow anomaly driven by logistics performance uncertainty, and this serves as an important basis for generating cash flow anomaly warning information. The preset threshold is determined based on the statistical distribution of counterfactual deviation data under historical normal operating conditions, including statistical thresholds set based on the mean and standard deviation or distribution thresholds set based on quantiles.

[0058] In addition, a potential function model for characterizing system stability is constructed, and the system stability is evaluated, and the parameters for changes in recovery capability are calculated. ; in, The parameter representing the change in resilience is used to characterize the rate of change in the overall credit status of the supply chain credit network model. Its value reflects the strength and trend of the system's resilience. This indicator represents the strength of credit behavior at the current moment. It is obtained by normalizing and weighting the data extracted from transaction data, including data on changes in payment terms, prepayment ratios, and credit ratios. It is used to characterize the overall degree of credit expansion or contraction of enterprises in the current supply chain. The baseline credit level is represented by a parameter obtained through statistical analysis of transaction data from a historically stable operating phase. Specifically, it is calculated and integrated by averaging and combining data on changes in payment terms, prepayment ratios, and credit ratios within a preset time interval. This serves as a reference standard under normal system conditions. The preset time interval is selected based on the period when the supply chain is in a stable operating state in historical transaction data, specifically including periods without significant fluctuations in payment terms or abnormal cash flow. 't' represents a time variable, generated by the system based on a unified timestamp during data collection, used to identify the evolution of credit behavior characteristics over time. This represents the operator for calculating the rate of change with respect to time. In practice, it is calculated using a discrete-time difference method, that is, based on the time difference between adjacent time windows. The ratio of the change in the quantity to the time interval is used to obtain the parameter of change in recovery capability.

[0059] By analyzing parameters related to changes in recovery ability Continuous calculations reveal the trend of resilience changes in the supply chain credit network model over time. The resilience change parameter characterizes the rate of change in credit behavior relative to a baseline level; a higher value indicates a faster decline in system resilience. Specifically, when… When the value remains positive and increases, it indicates a sustained contraction in credit activity and a decline in the system's resilience; when When the value is close to zero or negative, it indicates that the credit status is stabilizing or recovering.

[0060] It can enable dynamic assessment of changes in system stability and be used to generate critical approach indicators. When the resilience parameter changes... When the value continues to rise and increases, a critical approach indicator is generated, and based on this indicator, an early warning of abnormal fund flows is triggered, thereby achieving proactive monitoring and management of potential credit risks. Through... Through continuous monitoring, the system can quantitatively reflect changes in the stability of the credit network, upgrade the early warning mechanism from "outcome-driven" to "trend-driven", and improve the accuracy and robustness of identifying abnormal cash flow in the supply chain.

[0061] This step integrates counterfactual analysis and potential function models to achieve multi-dimensional judgment of system anomalies. Counterfactual analysis is used to quantify the contribution of logistics factors to abnormal cash flow, while potential function models are used to assess whether the system is approaching a critical instability state, thereby realizing a technological leap from "post-event detection" to "pre-event warning".

[0062] Example 2: Specific Application Scenarios

[0063] In this embodiment, a supply chain network comprising vehicle manufacturers, Tier 1 suppliers, and multi-tiered component suppliers is used as the application object. Vehicle manufacturers and suppliers at all levels form a multi-tiered credit association structure through orders, logistics, and settlement relationships.

[0064] The system acquires logistics fulfillment data, transaction data, and fund settlement data of each node within a preset time window by deploying data collection interfaces at each node in the supply chain, and maps the above data into node behavior characteristic data and inter-node interaction data.

[0065] In the event of fluctuations in the logistics environment, step 1 calculates the entropy value of the logistics performance data at each node to obtain the characteristics of the change in the performance entropy value, thereby characterizing the uncertainty changes in logistics performance behavior. Further, in step 2, a time-delay coupling analysis model is constructed to model the time-series relationship between logistics performance data and fund settlement data, identify the time-delay response relationship between different nodes, and extract the characteristics of credit behavior changes.

[0066] Based on this, in step 3, a supply chain credit network model is constructed according to the transaction relationship and credit behavior characteristics between nodes to obtain node weights and credit connection relationships; in step 4, the network stability index is calculated by performing spectral feature analysis on the credit network model, and the structural state of the supply chain credit network is determined based on the changes in the stability index.

[0067] Furthermore, in step 5, based on counterfactual analysis and potential function evaluation, the cash flow deviation of key nodes is quantitatively analyzed, and combined with changes in network structure stability, the critical state of the supply chain credit network is identified, and risk warning information is output.

[0068] Through the above technical solutions, while maintaining consistency with the actual credit relationship mapping between vehicle manufacturers and multi-level suppliers, dynamic identification and early warning of supply chain credit risks can be achieved.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them; although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications can still be made to the specific implementation methods of the present invention or equivalent substitutions can be made to some technical features without departing from the spirit of the technical solutions of the present invention, and all such modifications should be covered within the scope of the technical solutions claimed in the present invention.

Claims

1. A method for monitoring and early warning of abnormal cash flow in supply chain collaboration, characterized in that, Includes the following steps: Step 1: Receive logistics fulfillment data, transaction data, and fund settlement data from multiple nodes in the supply chain. The logistics fulfillment data includes transportation time series, delivery completion records, and node delay records. Perform time series processing on the logistics fulfillment data to calculate the logistics fulfillment uncertainty characteristics. Then, perform multi-scale decomposition processing on the logistics fulfillment uncertainty characteristics to obtain data of different frequency band components. Based on the data of different frequency band components, filter out low-frequency component data with frequencies below a preset threshold, and calculate the fluctuation amplitude and trend of the low-frequency component data to generate logistics uncertainty characteristics that characterize the stability changes of the logistics system. Step 2: Perform feature extraction processing on the transaction data to generate credit behavior features. Input the logistics uncertainty features into the preset time-delay coupling analysis model and perform correlation calculation with the credit behavior features to obtain the time-delay response relationship between logistics changes and credit behavior. The credit behavior features include payment period change data, prepayment ratio data, and credit ratio data. Step 3: Construct a supply chain credit network model based on the time-delay response relationship. The supply chain credit network model includes node weights and edge weights. The node weights are determined by the fund settlement data and logistics uncertainty characteristics, while the edge weights are determined by the credit behavior characteristics. Step 4: Extract structural parameters from the supply chain credit network model, perform change detection processing based on the structural parameters to generate network structure change features, and perform phase transition determination processing based on the network structure change features to generate the supply chain credit network state result. Step 5: Generate early warning information on abnormal cash flow based on the status results of the supply chain credit network.

2. The method for monitoring and early warning of abnormal cash flow in supply chain collaboration according to claim 1, characterized in that, The calculation steps for logistics performance uncertainty characteristics include: receiving time series data from logistics performance data; calculating volatility and dispersion indices on the time series data; combining the volatility and dispersion indices to obtain basic uncertainty characteristics; performing entropy calculation on the basic uncertainty characteristics to obtain performance entropy values; fusing the performance entropy values ​​with the basic uncertainty characteristics to generate logistics performance uncertainty characteristics; and outputting the logistics performance uncertainty characteristics to subsequent processing steps.

3. The method for monitoring and early warning of abnormal cash flow in supply chain collaboration according to claim 1, characterized in that, The multi-scale decomposition process includes: performing frequency domain transformation on the logistics fulfillment data to decompose it into high-frequency and low-frequency components; performing random disturbance filtering on the high-frequency components to obtain the filtered data; performing stability enhancement calculations on the low-frequency components to obtain structural fluctuation data; and inputting the structural fluctuation data as the characteristics of logistics uncertainty into the time-delay coupling analysis model.

4. The method for monitoring and early warning of abnormal cash flow in supply chain collaboration according to claim 1, characterized in that, The time-delay coupling analysis model includes: receiving logistics uncertainty characteristics and credit behavior characteristics; calculating the correlation between logistics uncertainty characteristics and credit behavior characteristics within multiple time windows; calculating the optimal time lag parameter based on the correlation; establishing the mapping relationship between logistics uncertainty characteristics and credit behavior characteristics based on the optimal time lag parameter; and outputting the time-delay response relationship to construct a supply chain credit network model.

5. The method for monitoring and early warning of abnormal cash flow in supply chain collaboration according to claim 1, characterized in that, The steps for constructing a supply chain credit network model include: receiving credit behavior characteristics and fund settlement data; generating credit connection relationships between nodes based on credit behavior characteristics; calculating node weights based on fund settlement data and logistics uncertainty characteristics; and combining credit connection relationships and node weights to form a weighted network structure to generate a supply chain credit network model.

6. The method for monitoring and early warning of abnormal cash flow in supply chain collaboration according to claim 1, characterized in that, The steps for detecting network structure change characteristics include: periodically sampling the structural parameters of the supply chain credit network model; calculating the changing trend of the structural parameters; detecting abrupt changes in the changing trend; and using the structural changes corresponding to the abrupt changes as network structure change characteristics.

7. The method for monitoring and early warning of abnormal cash flow in supply chain collaboration according to claim 1, characterized in that, The phase transition determination process includes: threshold judgment of network structure change features; extraction of network structure feature vectors based on network structure change features, spectral feature analysis of network structure feature vectors to obtain spectral feature data; detection of feature value changes in spectral feature data; determination of credit network phase transition state when network structure change features meet preset mutation judgment conditions and spectral feature data meet feature value change conditions; and output of supply chain credit network state results.

8. The method for monitoring and early warning of abnormal cash flow in supply chain collaboration according to claim 1, characterized in that, The steps for generating early warning information on abnormal cash flow include: receiving the status results of the supply chain credit network; calculating risk level parameters based on the status results of the supply chain credit network; identifying key node enterprises in the supply chain credit network model; generating early warning data containing risk level parameters and information on key node enterprises; and outputting the early warning data to the early warning terminal.

9. The method for monitoring and early warning of abnormal cash flow in supply chain collaboration according to claim 1, characterized in that, Following step 5, a counterfactual analysis step is also included: constructing a baseline model based on historical data; simulating the steady state of logistics in the baseline model and obtaining the simulation results; Calculate the deviation between actual cash flow data and simulation results; Perform anomaly detection and processing on the deviation data; Output the deviation judgment result and use it as a supplementary basis for early warning.

10. The method for monitoring and early warning of abnormal cash flow in supply chain collaboration according to claim 1, characterized in that, Following step 5, a critical state detection step is also included: constructing a potential function model to characterize the system's stability; and conducting a stability assessment of the supply chain credit network model. Calculate the parameters of change in recovery capacity; When the recovery capability parameter shows a continuous increase and the value increases, a critical approach indicator is generated. The critical approach marker will be used to trigger early warning information for abnormal cash flow.

Images