Supply chain intelligent tracking system for process monitoring

Abstract

The present invention discloses a supply chain intelligent tracking system for process monitoring, which relates to the technical field of supply chain intelligent management and is used to solve the problems of flexibility and risk resistance of supply chain breakpoint management; it includes a supply chain data processing module, an energy transmission analysis module, a breakpoint prediction module, and a supply chain adjustment module; the present invention collects original data from supply chain nodes, combines time series, classification features and high-dimensional data processing technology to generate effective features, and constructs a breakpoint prediction model through a gradient boosting decision tree to identify and evaluate breakpoint risks, build a supply chain network topology structure, recursively calculate the propagation path and impact range of breakpoint risks, determine the set of affected nodes, formulate adjustment and recovery strategies based on the prediction results, optimize resource allocation and logistics routes, quickly restore key nodes, ensure supply chain continuity, upgrade the early warning mechanism to dynamically monitor risks, thereby improving the management flexibility and risk resistance of the supply chain.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent supply chain management, and more specifically, to an intelligent supply chain tracking system oriented to process monitoring. Background Art

[0002] Supply chain management refers to the planning, coordination, and control of every link in the supply chain to achieve the efficient flow and delivery of products or services from raw material suppliers to end consumers. Supply chain management encompasses procurement, production, warehousing, transportation, distribution, and other links, involving multiple parties such as suppliers, manufacturers, logistics service providers, retailers, and customers. Its core goals are to optimize resource allocation, shorten delivery times, reduce costs, and enhance the overall efficiency and flexibility of the supply chain. Effective supply chain management can respond to changing market demands, enhance corporate competitiveness, ensure on-time product delivery, reduce inventory and logistics costs, and maintain a high level of customer service.

[0003] Deficiencies in existing technologies: Supply chain management systems often suffer from inaccurate predictions and delayed responses when addressing discontinuity risks, primarily manifesting in the following aspects. First, traditional risk assessment methods rely on simple statistical analysis of historical data and lack in-depth modeling of the interactions between multiple nodes and links in complex supply chain networks, making it difficult to accurately identify high-risk nodes and potential discontinuities. Second, such systems are generally unable to monitor dynamic changes in each link of the supply chain in real time. They lack a keen sense of changes in key factors such as logistics, supplier fulfillment rates, and inventory fluctuations, and are unable to issue timely warnings. Traditional systems are particularly slow to respond to sudden supply chain discontinuities, often only taking action after the problem has spread to multiple nodes and expanded its scope. This lag makes it difficult for supply chain managers to effectively and proactively mitigate risks, prolonging recovery time from supply chain disruptions. This ultimately leads to reduced supply chain efficiency, increased costs, and decreased customer satisfaction. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the following solution is proposed to solve the problems of flexibility and risk resistance of supply chain breakpoint management in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] The supply chain data processing module is used to collect raw data from various nodes and links in the supply chain, perform time series processing, classification feature processing, and high-dimensional data processing on the raw data, and generate effective features for model training;

[0007] The breakpoint prediction module is used to build a supply chain breakpoint prediction model using a gradient boosting decision tree, conduct risk assessments on breakpoints in the supply chain, identify the probability of breakpoints occurring, and provide prediction results of supply chain disruption risks;

[0008] Risk propagation recursive module: Based on the supply chain network topology and combined with the results of the supply chain breakpoint prediction model, it recursively calculates the propagation path and impact range of breakpoint risks in the supply chain, and determines the high-risk nodes and the affected node sets;

[0009] The supply chain adjustment module is used to formulate supply chain adjustment and recovery strategies based on the results of the breakpoint impact range, and implement key node recovery, resource allocation optimization, logistics route adjustment and early warning upgrade mechanism for supply chain management.

[0010] In a preferred embodiment, the system collects raw data from various nodes and links in the supply chain, performs time series analysis, classification feature conversion, and data dimension reduction preprocessing on the raw data to generate effective features for model training. The specific steps are as follows:

[0011] Collect raw data from various nodes and links in the supply chain. The types of raw data collected include logistics data, supplier data, warehousing data, market demand data, and external factors;

[0012] Perform seasonal decomposition on the time series data in the original data, use difference operation to eliminate the trend influence in the time series data, and use moving average to smooth the original data;

[0013] One-hot encoding and label encoding are performed on categorical data to convert them into binary features;

[0014] Use principal component analysis to reduce the dimensionality of the original data;

[0015] Feature selection is performed based on the feature importance of the tree, the importance value of each feature is calculated and output, and feature screening is performed based on the importance value.

[0016] In a preferred embodiment, a gradient boosting decision tree is used to construct a supply chain breakpoint prediction model, perform risk assessment on breakpoints in the supply chain, identify the probability of breakpoints occurring, and provide a prediction result of the supply chain disruption risk. The specific steps include:

[0017] The input data for the supply chain breakpoint prediction model includes transportation time, supplier historical delivery time, inventory turnover rate, logistics cost, product shipment frequency, external environmental data, time series trend of each feature, seasonal fluctuation, and category characteristics;

[0018] Define and build model labels, and classify labels in supply chain breakpoint prediction into binary classification labels;

[0019] The gradient boosting decision tree was selected as the supply chain breakpoint prediction model, and grid search was performed on different hyperparameter combinations to optimize the hyperparameter settings;

[0020] Use 70% of the original data as the training set to train the supply chain breakpoint prediction model, and the remaining 30% of the original data as the validation set;

[0021] After the supply chain breakpoint prediction model is trained, it is evaluated using the validation set to determine the accuracy, F1 score, and area under the ROC curve;

[0022] If the supply chain breakpoint prediction model is evaluated and achieves the expected results, it will be put into use to calculate the probability of supply chain breakpoints occurring and predict the risk of supply chain disruption.

[0023] In a preferred embodiment, based on the supply chain network topology and combined with the results of the supply chain breakpoint prediction model, the propagation path and impact range of the breakpoint risk in the supply chain are recursively calculated to determine the high-risk nodes and the affected node set. The specific steps are as follows:

[0024] According to the output data of the supply chain breakpoint prediction model, the supply chain breakpoints are determined and the network topology of the supply chain is constructed;

[0025] The supply chain network topology is constructed as a graph, which consists of a set of nodes in the supply chain and a set of connecting edges between the nodes;

[0026] The weight of the edge between two nodes is determined by the nonlinear combination of the material flow, transaction volume and transportation frequency of the two nodes;

[0027] Based on the supply chain network topology, a breakpoint risk propagation matrix is ​​established. The risk propagation matrix represents the risk propagation in the network when a breakpoint occurs at a node.

[0028] After the supply chain breakpoint prediction model detects a breakpoint in a supply chain node, the breakpoint propagation matrix is ​​used to recursively calculate the impact range of the breakpoint, evaluate the breadth and depth of the breakpoint propagation from the initial node to each node, and determine the set of nodes affected by the high-risk node.

[0029] In a preferred embodiment, evaluating the breadth and depth of the breakpoint propagation from the initial node to each node and determining the set of nodes affected by the high-risk node includes the following steps:

[0030] The breakpoint nodes detected by the supply chain breakpoint prediction model are used as the high-risk node set for recursive calculation. Each node in the high-risk node set represents a node where a breakpoint occurs in the supply chain. The high-risk node set is used as the starting point for recursive propagation.

[0031] For each initial breakpoint node, the risk propagation intensity of all adjacent nodes is calculated. The propagation intensity is based on the elements in the breakpoint risk propagation matrix. When the propagation intensity exceeds the threshold, the risk spreads to the next node.

[0032] When the recursive calculation is terminated, the impact range will be output. The output results include the affected node set, propagation path and the cumulative propagation intensity of the impact.

[0033] In a preferred embodiment, the system is used to formulate supply chain adjustment and recovery strategies based on the breakpoint impact range results, and implement key node recovery, resource allocation optimization, logistics route adjustment, and early warning upgrade mechanisms for supply chain management, including the following steps:

[0034] Based on the output of the affected node set, nodes in the supply chain are identified and restored first according to the importance and vulnerability calculation results of the nodes;

[0035] Prioritize restoring node P i The ranking is comprehensively evaluated by node dependency and influence strength: Among them, d ij is node v i , v j The dependence between ij is the risk transmission rate, C i Is the recovery node v i the cost of resources required;

[0036] Assign the recovery task of high-priority nodes to the supply chain management team, immediately deploy resources (to restore node functions, monitor the status of each node during the recovery process, and ensure that these nodes are prioritized);

[0037] Carry out resource reallocation and dynamic logistics route adjustment, identify the logistics paths between affected nodes and bottleneck nodes in the supply chain based on propagation path analysis, and optimize logistics routes through resource flow F ij Solve the maximization formula of : Among them, F ij is node v i To node v j Material flow, r ij is the risk transmission rate, L ij It is the physical distance or transportation delay of the logistics route;

[0038] Supply chain managers dynamically select the optimal transportation routes based on resource flows;

[0039] While implementing priority node restoration measures, resource reallocation, and dynamic logistics route adjustment measures, a dynamic risk monitoring and early warning upgrade mechanism is initiated. The supply chain intelligent tracking system continuously monitors changes in node status and transmission paths based on real-time data streams, identifies new potential breakpoints, and upgrades the early warning level.

[0040] The technical effects and advantages of the supply chain intelligent tracking system for process monitoring of the present invention are as follows:

[0041] The present invention collects multi-dimensional raw data from various nodes and links in the supply chain, and combines time series processing, classification feature processing and high-dimensional data processing technologies to generate effective features that can be used for model training. The data features can comprehensively reflect the operating status of the supply chain, especially the dynamic changes in logistics, suppliers, warehousing, market demand and other links. By using the gradient boosting decision tree to construct a supply chain breakpoint prediction model, the system can accurately identify potential breakpoint risks in the supply chain and evaluate the probability of breakpoint occurrence. The prediction results generated by the model help supply chain managers take preventive measures before the breakpoint occurs, avoid supply chain interruptions, and improve the stability and reliability of the overall supply chain.

[0042] By combining the supply chain network topology structure and the breakpoint prediction model, the recursive propagation calculation of breakpoint risks is realized, which can accurately determine the high-risk nodes and the affected node sets, identify the breakpoint propagation path and impact range, and provide early warning for breakpoints. Targeted adjustment and recovery strategies can also be formulated based on the results of the recursive calculation. Combined with resource allocation optimization and logistics route adjustment functions, the system can quickly respond and restore key nodes when a breakpoint occurs, ensuring the continuity of supply chain operations. In addition, the early warning upgrade mechanism can dynamically monitor emerging risks, effectively improving the flexibility and risk resistance of supply chain management. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a schematic diagram of the structure of the supply chain intelligent tracking system for process monitoring of the present invention. DETAILED DESCRIPTION

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0045] In order to achieve the above objectives, Figure 1The present invention provides a schematic diagram of the structure of the supply chain intelligent tracking system for process monitoring, which specifically includes a supply chain data processing module, an energy transmission analysis module, a breakpoint prediction module, and a supply chain adjustment module. The modules are connected by signals.

[0046] The supply chain data processing module is used to collect raw data from various nodes and links in the supply chain, perform time series processing, classification feature processing, and high-dimensional data processing on the raw data, and generate effective features for model training;

[0047] The breakpoint prediction module is used to build a supply chain breakpoint prediction model using a gradient boosting decision tree, conduct risk assessments on breakpoints in the supply chain, identify the probability of breakpoints occurring, and provide prediction results of supply chain disruption risks;

[0048] Risk propagation recursive module: Based on the supply chain network topology and combined with the results of the supply chain breakpoint prediction model, it recursively calculates the propagation path and impact range of breakpoint risks in the supply chain, and determines the high-risk nodes and the affected node sets;

[0049] The supply chain adjustment module is used to formulate supply chain adjustment and recovery strategies based on the results of the breakpoint impact range, and implement key node recovery, resource allocation optimization, logistics route adjustment and early warning upgrade mechanism for supply chain management.

[0050] Collect historical and real-time data from every node and link in the supply chain. This raw data includes logistics data (e.g., shipping time, shipping routes), supplier data (e.g., delivery times, historical fulfillment rates), warehousing data (e.g., inventory changes, inbound / outbound cycles), market demand data (e.g., sales volume, customer order status), and external factors (e.g., weather, economic conditions). This collected data serves as the basic input for building subsequent forecasting models.

[0051] Data collection is the foundation of supply chain breakpoint prediction. By acquiring raw data from multiple sources, the system can establish a comprehensive view of the status of supply chain nodes and the impact of the external environment. Data in different formats needs to be standardized to ensure that the data can be effectively used by subsequent machine learning algorithms. For example, logistics data may need to be obtained through GPS tracking systems, supplier fulfillment rate data can be extracted from historical order records, and weather and economic factors can be obtained through third-party service interfaces. Data collation requires addressing issues such as missing values ​​and noisy data to ensure the quality of input data.

[0052] After collecting and organizing the data, the system needs to convert the raw data into features that are helpful for model training through feature extraction and feature engineering. The system also needs to perform time series analysis, classification feature conversion, and data dimensionality reduction on the raw data. The specific steps are as follows:

[0053] Seasonal decomposition is performed on the time series data X(t) collected from the original data, breaking it down into three components: trend T(t) represents the long-term direction of change; seasonality S(t) represents cyclical fluctuations in the data (such as differences in supply chain order volumes in different quarters); and random noise R(t) represents unexplained random fluctuations. For example, this decomposition can be used to extract the trend and seasonal fluctuation characteristics that affect transportation delays in transportation time X(t) from logistics data. These extracted features are used as input to the model to predict future transportation delays.

[0054] Furthermore, differencing is used to eliminate the influence of trends in time series data and generate a stationary series. For the series X(t), the first-order difference calculation formula is: X'(t) = X(t) - X(t-1), where t represents the time point. After differencing, short-term fluctuations can be captured and the influence of long-term trends can be eliminated. It is particularly suitable for capturing sharp fluctuations in the supply chain in the short term (such as sudden transportation delays).

[0055] Moving average is used to smooth the original data, eliminate short-term fluctuations, and obtain long-term trends. For a moving average with a window size of k, the formula is as follows: The moving average can ignore short-term abnormal fluctuations and focus on long-term trends. For example, the moving average of logistics time can capture the average transportation time over a period of time, helping to assess whether there will be abnormal delays in current logistics.

[0056] In supply chain data, many features are categorical (such as supplier category, transportation method, warehouse location, etc.). Such data cannot be directly used in machine learning models and needs to be converted into numerical features through categorical feature conversion. Common methods include one-hot encoding and label encoding. One-hot encoding converts categorical variables into multiple binary features. For example, if there are three modes of transportation in the supply chain: "air transport", "sea transport", and "land transport", they can be represented by three binary variables: (1, 0, 0) for air transport, (0, 1, 0) for sea transport, and (0, 0, 1) for land transport.

[0057] In the case of high-dimensional features, the data may contain many redundant or irrelevant features. To reduce the complexity of the subsequent model and improve computational efficiency, feature engineering uses data dimensionality reduction techniques, such as principal component analysis (PCA), to reduce the dimensionality of high-dimensional data, projecting it into a low-dimensional space while retaining the main trend of the data.

[0058] Assume there are 100 different features in a supply chain, such as shipping time, inventory levels, and productivity. PCA can compress these high-dimensional features into fewer principal components (e.g., 10), ensuring that the model still captures the main trends while improving computational efficiency.

[0059] After preprocessing the raw data, that is, after feature extraction and conversion, the most predictive features are selected using feature selection methods. Tree-based feature importance is used for feature selection. Tree models such as random forests or gradient boosting trees can directly output the importance value of each feature. The importance value is evaluated by calculating the contribution of each feature to the split point of the model. The formula is expressed as: Where ΔG(t) represents the change in the Gini index or information gain of node t in the decision tree, T is the set of all nodes in the decision tree, and I j is the importance of feature j.

[0060] In summary, through the combination of time series analysis, categorical feature transformation, and data dimensionality reduction, feature extraction and feature engineering steps can effectively transform raw data into structured features that can be used by the model. These features can help the model capture the complex patterns of supply chain breakpoints and improve forecasting accuracy.

[0061] Build a supply chain breakpoint prediction model that can accurately capture the relationship between input data (features) and output data (labels). This model also has good generalization capabilities, can adapt to changes in different supply chain nodes, and effectively predict possible future breakpoints.

[0062] The processed features are used as input data (features), including transportation time, supplier's historical delivery time, inventory turnover rate, logistics cost, product shipment frequency, external environment data (such as weather, market demand fluctuations), etc. The time series trend, seasonal fluctuation, and category characteristics (such as different suppliers and logistics companies) of each feature can be used as input;

[0063] Define and build model labels, and classify labels in supply chain breakpoint prediction into binary labels (0 indicates no breakpoint, 1 indicates a breakpoint has occurred). Label definitions can be based on historical breakpoint event records and business needs, specifically:

[0064] Output data (labels): Breakpoint prediction is typically a binary classification problem, where a label of 1 indicates a breakpoint (e.g., delivery delay, logistics disruption) and a label of 0 indicates no breakpoint. Labels can be generated based on actual breakpoint events in the historical supply chain. For example, a delivery time exceeding the scheduled time by X days can be defined as a breakpoint. Breakpoint labels must be synchronized with the time of the features. For example, if the system is predicting whether a supply chain disruption will occur in the coming week, the breakpoint state corresponding to the label must be aligned with the time point of the input features. For example, to predict a supplier's delivery status within the next month, a label of 1 indicates that the supplier may delay delivery (by more than 3 days), while a label of 0 indicates that the supplier will deliver on time.

[0065] We choose the gradient boosting decision tree (XGBoost) as the model. XGBoost has strong performance in dealing with high-dimensional data, feature complexity, and nonlinear problems. By integrating multiple decision trees, we gradually correct the error, improve the prediction accuracy, and minimize the objective function with regularization terms: Among them, the loss function Used to measure the error between the model prediction value and the actual value, regularization term Used to control the complexity of the model and avoid overfitting; i represents the true label of the i-th sample; represents the predicted value of the i-th sample predicted by the model; f k represents the kth tree; K represents the number of trees;

[0066] Represents the loss function, which is used to measure the error between the model prediction value and the true label. The form of the loss function is different in different tasks;

[0067] Key hyperparameters in the gradient boosting decision tree include: the number of trees, which indicates how many trees are constructed. A larger number of trees can capture more details but may lead to overfitting. The default value is generally 100; the maximum depth of the tree is used to control the depth of each tree. A larger depth will make the model more complex and suitable for scenarios with more complex data; the learning rate is used to control the learning rate of each tree. A smaller learning rate can improve the generalization ability of the model, but requires more trees; the proportion of tree construction samples;

[0068] In the supply chain breakpoint prediction model, the hyperparameters were set as follows: 200 trees were used to improve model accuracy, the maximum depth of each tree was set to 6, suitable for capturing complex feature relationships, and a small learning rate of 0.05 was set to enhance generalization ability. 80% of the samples were used for tree construction to prevent model overfitting.

[0069] Perform model training and hyperparameter optimization. Train the model using historical data and use the gradient boosting algorithm to iteratively optimize the prediction error. During the training process, the model's hyperparameters also need to be tuned. Common methods include grid search or random search.

[0070] The logarithmic loss function is usually used for binary classification problems: Used to measure the deviation between the model prediction value and the actual label;

[0071] Use 70% of the original data as the training set for model training, and the remaining 30% as the validation set;

[0072] Use K-fold cross-validation to prevent overfitting and ensure generalization.

[0073] Perform hyperparameter optimization, i.e., perform grid search on different hyperparameter combinations (such as tree depth, learning rate, and subsample ratio) to find the optimal hyperparameter settings, and use early stopping, i.e., stop training early if the performance of the model on the validation set does not improve after a certain number of epochs;

[0074] When optimizing hyperparameters using grid search, test different combinations of learning rates and tree depths.

[0075] After model training is complete, use an independent validation or test set to evaluate the model. Key metrics include accuracy, F1 score, and area under the receiver operating characteristic (ROC) curve (AUC). Furthermore, generalization depends on the model's performance on unseen data, particularly its ability to predict supply chain disruption scenarios.

[0076] The generalization ability of the model depends on feature selection, hyperparameter settings, and the control of model complexity by regularization. Controlling the tree depth and regularization can improve the model's generalization ability and prevent overfitting. Regarding the requirements for the breadth of input features, time series data processing (such as differencing and moving averages) can capture long-term changes in features and improve the model's accuracy in predicting future supply chain breakpoints.

[0077] After the supply chain breakpoint prediction model has been evaluated and achieved the expected results, it can be deployed to the production environment to predict supply chain breakpoint risks in real time. The system regularly inputs real-time data through the supply chain management platform. The model uses reasoning to derive risk scores and generate early warning signals.

[0078] For example, a global electronics company (Company XYZ) relies on multiple suppliers and logistics partners to transport electronic components from supplier warehouses to various production sites. Because the supply chain involves multiple countries and different modes of transportation (air, sea, and land), logistics delays have become a critical issue for the company's production efficiency. To avoid production line interruptions due to logistics delays, Company XYZ decided to build a supply chain disruption prediction model to predict logistics delay risks in advance and issue early warnings.

[0079] Input data (features) include: historical logistics data: number of days of logistics delay for all shipments over the past year, mode of transport (air, sea, land), distance, time, and route congestion; supplier data: suppliers' historical on-time delivery rates and number of delivery batches; inventory data: inventory levels and inventory turnover days for each production base; market demand data: changes in customer orders over the past six months and market demand fluctuations; external factors: weather conditions along the transportation routes, port congestion, and political risks (such as customs delays and tariff adjustments).

[0080] Output data (labels). The logistics delay label indicates whether logistics delay occurs (label 1 indicates delay, 0 indicates on-time arrival). Delay is defined as the shipping time exceeding the scheduled arrival time by more than 3 days.

[0081] One-hot encoding is performed on transportation modes (air, sea, and land), moving average processing is used on historical transportation time data to extract trend features, and numerical features such as inventory levels and supplier delivery on-time rates are normalized to ensure consistency of model inputs.

[0082] XGBoost was selected as the core model setting, with hyperparameters set to 200 trees, a maximum depth of 6 for each tree, a learning rate of 0.05, 80% of the samples were used to build the tree, and the minimum sample weight on the leaf node was 1;

[0083] We used logistics data from the past year, with 70% used for training and 30% used to validate the model's generalization ability. We used a logarithmic loss function to optimize the prediction performance for binary classification problems, aiming to minimize the cost of misclassification. We tested different hyperparameter combinations through grid search and selected the optimal combination.

[0084] Model evaluation showed an accuracy of 0.88 on the validation set, indicating that the model can effectively distinguish delayed from non-delayed logistics batches. The F1 score was 0.85, indicating that the model has a strong ability to capture delay events and a good balance between positive and negative predictions. The area under the receiver operating characteristic (ROC) curve was 0.92, indicating that the model has excellent overall performance and can effectively distinguish batches with logistics delays from those without.

[0085] The generalization ability of the model was tested by K-fold cross-validation (K=5). The results showed that the AUC in each fold exceeded 0.90, indicating that the model performed well on different datasets and could be generalized to different transportation routes, suppliers, and transportation methods.

[0086] XYZ Company deployed the optimized XGBoost model in its supply chain management platform. The platform receives data daily from suppliers, logistics partners, and market demand, and uses the model to perform real-time reasoning, as follows:

[0087] The platform collects real-time transportation time, route status, weather forecast and port conditions of logistics batches every day. The model calculates the risk score of delay for each transportation batch in real time. When the delay risk score of a batch exceeds 70% (for example, an important transportation route has a 75% probability of delay), the system will issue an early warning and notify logistics management personnel to reschedule the route or increase transportation capacity.

[0088] Based on the output data of the supply chain breakpoint prediction model, the supply chain breakpoints are determined and the network topology of the supply chain is constructed. This structure represents each node in the supply chain (such as suppliers, manufacturers, logistics nodes, warehouses, customers, etc.) in the form of a graph. Each node is connected to its upstream and downstream nodes by edges. The weight of the edge represents the dependency between the nodes. It is calculated based on data such as material flow, transaction volume, and transportation frequency in the supply chain, as follows:

[0089] Suppose the supply chain network topology is a graph G = (V, E), where V represents the set of nodes in the supply chain (such as suppliers, manufacturers, logistics nodes, etc.), and E represents the set of connecting edges between nodes;

[0090] For any two nodes v i , v j ∈V, edge weight w ij The material flow between the two ij , transaction volume t ij and transport frequency q ij The nonlinear combination of Among them, α, β1, β2, γ are parameters to be adjusted, and ∈ is a minimum value used to prevent the denominator from being zero;

[0091] The construction of the supply chain network topology is based on the dependencies between nodes. Dependencies are comprehensively reflected through raw data. Material flow and transportation frequency are important indicators of the intensity of interactions between nodes, while transaction volume reflects the dependencies at the economic level. Through the design of nonlinear formulas, the weights in the network can flexibly capture the changes in dependencies between different supply chain nodes.

[0092] The breakpoint risk propagation matrix is ​​constructed. Based on the supply chain network topology, the breakpoint risk propagation matrix R is established to describe the risk propagation in the network when one or more nodes have breakpoints. Each element r of the risk propagation matrix is ij Defined as: Where N(i) represents the node v i The set of neighbor nodes; η i Represents node v i The vulnerability of a node is determined by multiple factors, such as the node's resource redundancy, response time, and historical interruption rate.

[0093] Node vulnerability η i Reflects the node v i In the supply chain, the vulnerability of nodes is more susceptible to external risks when a breakpoint occurs, and the risk spreads to downstream nodes. Vulnerability calculations are usually based on the following factors:

[0094] Resource redundancy means that nodes with less redundancy are more vulnerable. Response time means that the slower a node responds, the worse its ability to handle external shocks. A node with a higher vulnerability is usually more vulnerable, as indicated by its historical breakpoint rate, which indicates that nodes with more interruptions are more vulnerable.

[0095] It should be noted that the construction of the risk propagation matrix is ​​mainly used to quantify the diffusion pattern of breakpoint risks in the supply chain network. Through the normalized edge weights, it can be ensured that the risk propagation of each node is relative and depends on the connection strength of its adjacent nodes. At the same time, the node vulnerability is used to reflect the self-protection ability of the node. The more vulnerable the node, the more likely it is to be affected in risk transmission.

[0096] After the supply chain breakpoint prediction model detects a breakpoint in one or more supply chain nodes, the breakpoint propagation matrix is ​​used to recursively calculate the impact range of the breakpoint. Specifically, the breakpoint risk starts to propagate from the initial breakpoint node, propagates to its adjacent nodes, and then continues to propagate outward until the propagation intensity is lower than the set threshold.

[0097] After constructing the supply chain network topology and the breakpoint risk propagation matrix to quantify the dependencies and risk propagation relationships between nodes, this step recursively calculates the impact range of the supply chain breakpoint to evaluate the breadth and depth of the breakpoint's propagation from the initial node to other nodes. Starting from the initial breakpoint node, the risk propagation between adjacent nodes is calculated in sequence to determine the scope of the propagation. The specific steps are as follows:

[0098] After the supply chain breakpoint prediction model detects that one or more nodes have breakpoints, these breakpoint nodes are used as the high-risk node set S0 for recursive calculation. High-risk nodes are the source of supply chain breakpoint propagation; that is, let S0 be the initial breakpoint node set, and each node v i ∈S0 represents the node where the breakpoint occurs in the supply chain. The high-risk node set S0 is used as the starting point of recursive propagation. This set can include suppliers, logistics nodes, or warehousing nodes, etc.

[0099] For each initial breakpoint node v i , calculate its risk propagation intensity to all adjacent nodes, the propagation intensity is based on the element r in the breakpoint risk propagation matrix ij , the element represents the node v i For adjacent nodes v i The impact of risk transmission intensity I j From node v i Propagate to adjacent nodes v j Defined as: Among them, r ij is an element in the risk propagation matrix, r ij Represents node v iFor node v j The risk transmission rate; τ is the transmission threshold, 1(r ij >τ) is the indicator function, when the transmission intensity exceeds the threshold τ, the risk will spread to the next node;

[0100] This step describes the local propagation of the breakpoint impact. The risk will only propagate from one node to another when the risk propagation rate between nodes exceeds the preset propagation threshold. In this way, the propagation range can be controlled to avoid excessive alarms.

[0101] Recursively update the impact range. After calculating the risk propagation intensity of the initial breakpoint node, the risk begins to spread to its adjacent nodes. Each affected adjacent node becomes a new breakpoint node and continues to propagate the risk to its adjacent nodes until the propagation intensity of all nodes is below the propagation threshold.

[0102] Assume I(t) represents the influence range at time t, and I(t+1) represents the influence range at the next moment. The recursive update formula is: I(t+1)=I(t)+∑ i∈I(t) ∑ j∈N(i) r ij ·1(r ij >τ), where N(i) is the node v i The neighbor node set of , recursively adds the nodes with strong propagation intensity to the new influence range set until no new nodes meet the propagation conditions;

[0103] After the initial breakpoint node propagates the risk to its adjacent nodes, if the adjacent nodes meet the propagation conditions (propagation intensity is greater than the threshold), they will continue to propagate to their neighboring nodes, recursively updating the gradual diffusion process of the capture breakpoint until all possible propagation paths are evaluated;

[0104] When the propagation intensity of all nodes is lower than the propagation threshold, or the recursive calculation reaches the preset maximum number of iterations, the risk propagation terminates. That is, when the propagation path no longer expands or reaches the preset number of iterations, the recursive process ends, and the impact range of the breakpoint can be determined at this time.

[0105] When the recursive calculation terminates, the system outputs the breakpoint impact range I(t) as the analysis result of the breakpoint's impact on the entire supply chain. The output results include the set of affected nodes, the propagation path, and the cumulative propagation intensity of the impact. This result can be used by supply chain managers to formulate response strategies. The specific steps are as follows:

[0106] Based on the output of the affected node set, the system identifies the most important nodes in the supply chain. These nodes are highly dependent on the entire supply chain, usually including suppliers, logistics hubs, storage centers, etc. Based on the importance and vulnerability calculation of the nodes, the nodes that are critical to the entire supply chain are restored first to prevent further spread. i The ranking can be comprehensively evaluated by node dependency and influence strength: Among them, d ij is node v i , v j The dependence between ij is the risk transmission rate, C i Is the recovery node v i The required resource cost, nodes with higher priority should be restored first;

[0107] Key nodes along the transmission path (such as logistics centers or major suppliers) have a greater impact on the entire supply chain. Therefore, more resources should be allocated to restore normal operations of these nodes as soon as possible. The response strategy focuses on preventing the risk of breakpoints from spreading to a wider area and affecting more nodes.

[0108] Specifically, the supply chain management team is tasked with restoring high-priority nodes. Resources are immediately deployed (e.g., rescheduling logistics, accelerating supplier deliveries) to restore node functionality. The system's real-time monitoring module is invoked to monitor the status of each node during the recovery process to ensure that these nodes receive priority.

[0109] Carry out resource reallocation and dynamic logistics route adjustment. Based on the propagation path analysis, the system identifies the logistics paths between the affected nodes and the bottleneck nodes in the supply chain. Through the intelligent tracking system to monitor the logistics data in real time, the system can dynamically adjust the logistics routes, optimize resource allocation, and ensure that the flow of materials is not affected. Logistics route optimization can be achieved through resource flow F ij Solve the maximization formula of : Among them, F ij is node v i To node v j Material flow, r ij is the risk transmission rate, L ij It is the physical distance or transportation delay of the logistics route;

[0110] In this strategy, the system optimizes logistics routes in the supply chain in real time to ensure that the supply of materials to affected nodes is not interrupted. Through the system's path tracking, supply chain managers can dynamically select the optimal transportation route to avoid logistics delays on routes with high transmission risks, while balancing material distribution and ensuring that resource supply to key nodes is prioritized.

[0111] Specifically, supply chain management platforms can be used to adjust logistics routes and transfer materials from affected areas to alternative routes or hubs. The dynamic route planning function embedded in the system can update routes in real time and optimize material flows through automatic scheduling algorithms. Combined with technologies such as drones and autonomous vehicles, critical materials can be quickly transported to high-priority nodes to ensure supply chain continuity.

[0112] While implementing countermeasures, the system can also activate dynamic risk monitoring and early warning upgrade mechanisms. The supply chain intelligent tracking system continuously monitors changes in node status and transmission paths based on real-time data streams, identifies new potential breakpoints, and promptly upgrades the early warning level to enable more in-depth protective measures.

[0113] Through the above-mentioned specific response strategies, supply chain managers can take quick action after a breakpoint occurs to ensure the stability and continuity of the supply chain. The supply chain intelligent tracking system not only helps identify the impact scope of the breakpoint, but also ensures that the supply chain can recover quickly and resist future risks through real-time monitoring, dynamic scheduling and long-term optimization.

[0114] It should be noted that the relevant threshold information in this embodiment is pre-set by professionals and will not be explained in detail here. For example, the threshold for comparing the transmission intensity with it, etc., some parameters in the embodiments have the same English letters, but are explained with different meanings when used, and will not be explained one by one here.

[0115] The present invention collects multi-dimensional raw data from various nodes and links in the supply chain, and combines time series processing, classification feature processing and high-dimensional data processing technologies to generate effective features that can be used for model training. The data features can comprehensively reflect the operating status of the supply chain, especially the dynamic changes in logistics, suppliers, warehousing, market demand and other links. By using the gradient boosting decision tree to construct a supply chain breakpoint prediction model, the system can accurately identify potential breakpoint risks in the supply chain and evaluate the probability of breakpoint occurrence. The prediction results generated by the model help supply chain managers take preventive measures before the breakpoint occurs, avoid supply chain interruptions, and improve the stability and reliability of the overall supply chain.

[0116] By combining the supply chain network topology structure and the breakpoint prediction model, the recursive propagation calculation of breakpoint risks is realized, which can accurately determine the high-risk nodes and the affected node sets, identify the breakpoint propagation path and impact range, and provide early warning for breakpoints. Targeted adjustment and recovery strategies can also be formulated based on the results of the recursive calculation. Combined with resource allocation optimization and logistics route adjustment functions, the system can quickly respond and restore key nodes when a breakpoint occurs, ensuring the continuity of supply chain operations. In addition, the early warning upgrade mechanism can dynamically monitor emerging risks, effectively improving the flexibility and risk resistance of supply chain management.

[0117] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0118] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0119] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0120] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0121] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0122] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. Supply chain intelligent tracking system for process monitoring, featuring: It includes supply chain data processing module, breakpoint prediction module, risk propagation recursion module and supply chain adjustment module, and each module is connected through signals; The supply chain data processing module is used to collect raw data from various nodes and links in the supply chain, perform time series processing, classification feature processing, and high-dimensional data processing on the raw data, and generate effective features for model training; The breakpoint prediction module is used to build a supply chain breakpoint prediction model using a gradient boosting decision tree, conduct risk assessments on breakpoints in the supply chain, identify the probability of breakpoints occurring, and provide prediction results of supply chain disruption risks; The risk propagation recursive module, based on the supply chain network topology and combined with the results of the supply chain breakpoint prediction model, recursively calculates the propagation path and impact range of breakpoint risks in the supply chain, and determines the high-risk nodes and the affected node sets; The supply chain adjustment module is used to formulate supply chain adjustment and recovery strategies based on the results of the breakpoint impact range, and implement key node recovery, resource allocation optimization, logistics route adjustment and early warning upgrade mechanism for supply chain management; Based on the supply chain network topology and the results of the supply chain breakpoint prediction model, we recursively calculate the propagation path and impact range of breakpoint risks in the supply chain to determine high-risk nodes and the set of affected nodes. The specific steps are as follows: According to the output data of the supply chain breakpoint prediction model, the supply chain breakpoints are determined and the network topology of the supply chain is constructed; The supply chain network topology is constructed as a graph, which consists of a set of nodes in the supply chain and a set of connecting edges between the nodes; The weight of the edge between two nodes is determined by the nonlinear combination of the material flow, transaction volume and transportation frequency of the two nodes; Based on the supply chain network topology, a breakpoint risk propagation matrix is ​​established. The risk propagation matrix represents the risk propagation in the network when a breakpoint occurs at a node. After the supply chain breakpoint prediction model detects a breakpoint at a supply chain node, it uses the breakpoint propagation matrix to recursively calculate the impact range of the breakpoint, evaluate the breadth and depth of the breakpoint's propagation from the initial node to each node, and determine the set of nodes affected by the high-risk node; Evaluate the breadth and depth of the breakpoint's propagation from the initial node to each node and determine the set of nodes affected by the high-risk node. This includes the following steps: The breakpoint nodes detected by the supply chain breakpoint prediction model are used as the high-risk node set for recursive calculation. Each node in the high-risk node set represents a node where a breakpoint occurs in the supply chain. The high-risk node set is used as the starting point for recursive propagation. For each initial breakpoint node, the risk propagation intensity of all adjacent nodes is calculated. The propagation intensity is based on the elements in the breakpoint risk propagation matrix. When the propagation intensity exceeds the threshold, the risk spreads to the next node. When the recursive calculation is terminated, the impact range is output, including the affected node set, propagation path, and the cumulative propagation intensity of the impact; It is used to formulate supply chain adjustment and recovery strategies based on the breakpoint impact range results, and implement key node recovery, resource allocation optimization, logistics route adjustment and early warning upgrade mechanism for supply chain management, including the following steps: Based on the output of the affected node set, nodes in the supply chain are identified and restored first according to the importance and vulnerability calculation results of the nodes; Prioritize recovery nodes The ranking is comprehensively evaluated by node dependency and influence strength: ,in, is a node , The dependencies between is the risk transmission rate, Is the recovery node the cost of resources required; Assign high-priority node recovery tasks to the supply chain management team, immediately deploy resources to restore node functionality, and monitor the status of each node during the recovery process; Carry out resource reallocation and dynamic logistics route adjustment, identify the logistics paths between affected nodes and bottleneck nodes in the supply chain based on propagation path analysis, and optimize logistics routes through resource flow Solve the maximization formula of : ,in, is a node To Node The flow of materials, is the risk transmission rate, It is the physical distance or transportation delay of the logistics path; Supply chain managers dynamically select the optimal transportation routes based on resource flows; While implementing priority node restoration measures, resource reallocation, and dynamic logistics route adjustment measures, a dynamic risk monitoring and early warning upgrade mechanism is initiated. The supply chain intelligent tracking system continuously monitors changes in node status and transmission paths based on real-time data streams, identifies new potential breakpoints, and upgrades the early warning level.

2. The supply chain intelligent tracking system for process monitoring according to claim 1 is characterized by: It is used to collect raw data from various nodes and links in the supply chain, and perform time series analysis, classification feature conversion, and data dimension reduction preprocessing on the raw data to generate effective features for model training. The specific steps are as follows: Collect raw data from various nodes and links in the supply chain. The types of raw data collected include logistics data, supplier data, warehousing data, market demand data, and external factors; Perform seasonal decomposition on the time series data in the original data, use difference operation to eliminate the trend influence in the time series data, and use moving average to smooth the original data; One-hot encoding and label encoding are performed on categorical data to convert them into binary features; Use principal component analysis to reduce the dimensionality of the original data; Feature selection is performed based on the feature importance of the tree, the importance value of each feature is calculated and output, and feature screening is performed based on the importance value.

3. The supply chain intelligent tracking system for process monitoring according to claim 2, characterized in that: It is used to build a supply chain breakpoint prediction model using a gradient boosting decision tree, conduct risk assessment on breakpoints in the supply chain, identify the probability of breakpoints occurring, and provide prediction results of supply chain disruption risks. The specific steps include: The input data for the supply chain breakpoint prediction model includes transportation time, supplier historical delivery time, inventory turnover rate, logistics cost, product shipment frequency, external environmental data, time series trend of each feature, seasonal fluctuation, and category characteristics; Define and build model labels, and classify labels in supply chain breakpoint prediction into binary classification labels; The gradient boosting decision tree was selected as the supply chain breakpoint prediction model, and grid search was performed on different hyperparameter combinations to optimize the hyperparameter settings; Use 70% of the original data as the training set to train the supply chain breakpoint prediction model, and the remaining 30% of the original data as the validation set; After the supply chain breakpoint prediction model is trained, it is evaluated using the validation set to determine the accuracy, F1 score, and area under the ROC curve; If the supply chain breakpoint prediction model is evaluated and achieves the expected results, it will be put into use to calculate the probability of supply chain breakpoints occurring and predict the risk of supply chain disruption.

Images