Data analysis method based on supply chain

Abstract

The invention relates to a data analysis method based on a supply chain, in particular to the technical field of business management, and the method comprises the following steps: S1, collecting supply chain data from an internal system and an external source, the internal data comprising a production plan, a production schedule, an inventory level, a purchase order and a sales record, the external data comprises logistics transportation information, market demand indexes, consumer behavior characteristics, meteorological monitoring data, road traffic flow and policy and regulation change information; according to the invention, by establishing a multi-source data access channel, data of an ERP system, an MES system, a WMS system, an external data platform and an Internet of Things device are uniformly collected, and historical data and real-time data are combined to construct a multi-dimensional time sequence database, so that complete acquisition and structured storage of supply chain global information are realized; and the comprehensiveness and accuracy of data input are ensured.

Description

Technical Field

[0001] This invention relates to the field of business management technology, and in particular to a data analysis method based on the supply chain. Background Technology

[0002] In traditional supply chain management systems, enterprises mainly rely on ERP, MES and WMS systems for internal process management. These systems can record and manage production plans, inventory levels, purchase orders and sales data. Among them, data analysis methods are mainly based on statistical models, usually based on historical averages, trend fitting and other methods to forecast demand and assess risks.

[0003] In practice, some problems still exist:

[0004] With the continuous expansion of the global supply chain system, enterprises face problems such as complex data sources, delayed information exchange, and insufficient risk identification in production, procurement, inventory, and logistics. Existing supply chain management relies heavily on ERP, MES, and WMS systems, but these systems are often limited to internal data processing and lack integration of external information such as market demand indices, consumer behavior characteristics, meteorological monitoring data, traffic flow, and policy and regulatory changes. This results in supply chain decisions not responding promptly to changes in the external environment. At the same time, traditional data analysis methods are mostly based on static statistics and single indicator calculations, making it difficult to dynamically monitor and quantitatively assess the operational status and potential risks of each node in the supply chain, which can easily lead to supply chain disruptions or resource allocation imbalances.

[0005] Current supply chain risk management mainly relies on manual experience and simple threshold warning mechanisms, lacking a systematic data-driven analysis framework. This makes it difficult to achieve real-time warnings and precise adjustments in a multi-source heterogeneous data environment. In terms of supply chain planning optimization, traditional methods often use single linear programming or heuristic algorithms, failing to take into account the multi-dimensional constraints of production capacity, inventory safety levels, and transportation conditions. The optimization results are insufficient in terms of globality and executability. In addition, current cross-departmental collaboration usually relies on manual information transmission and static reports, lacking a unified data sharing platform, which makes it impossible to efficiently synchronize and execute forecast results and risk signals among departments. Summary of the Invention

[0006] (a) Technical problems to be solved

[0007] To address the aforementioned problems in the prior art, this invention provides a supply chain-based data analysis method, resolving the issues raised in the background section.

[0008] (II) Technical Solution

[0009] To achieve the above objectives, the main technical solution adopted by the present invention is as follows:

[0010] Supply chain-based data analytics methods include the following steps:

[0011] S1: Collect supply chain data from internal systems and external sources. The internal data includes production plans, production progress, inventory levels, purchase orders and sales records. The external data includes logistics and transportation information, market demand index, consumer behavior characteristics, meteorological monitoring data, road traffic flow and policy and regulatory changes.

[0012] S2: Preprocess the data, including removing outliers, merging duplicate data, filling in missing values, standardizing the data, and performing feature extraction operations, thereby forming a structured, multi-dimensional dataset that can be used for computation.

[0013] S3: Based on the dataset, time series analysis, cluster analysis and pattern recognition methods are used to predict trends and identify potential risk signals in the supply chain operation status;

[0014] S4: Construct a risk assessment model that includes a risk identification module and a risk quantification module, monitor each node of the supply chain in real time, and generate early warning signals with risk levels;

[0015] S5: By establishing a unified data sharing platform, the forecast results and risk warning signals will be transmitted to relevant departments in real time for each department to obtain and use for business adjustments;

[0016] S6: Based on the aforementioned risk warning signals and feedback information from various departments, cross-departmental collaborative adjustments are made to production, procurement, inventory, and transportation plans;

[0017] S7: Through backtracking analysis and continuous monitoring, correct the parameters of the prediction model and the thresholds for risk assessment, and update the application in the next cycle.

[0018] The data collection step further includes:

[0019] Establish a multi-source data access channel, which includes data interface calls to the enterprise's internal ERP system, MES system, and WMS system, as well as data collection from external third-party data service platforms and IoT devices;

[0020] The historical data collection includes production, sales, inventory, and procurement data for at least one period over the past five years, which are used to train the predictive model;

[0021] The real-time data is acquired through API calls, sensor sampling, and web crawling. The real-time data and historical data are aligned with timestamps and mapped to form a multi-dimensional time-series database.

[0022] The risk assessment model includes:

[0023] The risk identification module is used to identify key risk nodes by statistically analyzing supplier delivery cycles, sampling production equipment operating status in real time by sensors, dynamically calculating inventory turnover rate, monitoring traffic flow for transportation routes, and combining anomaly detection algorithms.

[0024] The risk quantification module is used to calculate risk impact factors for the key nodes. These factors include the node failure probability, the node's topological weight in the supply chain network, and the node's dependence on its neighboring nodes.

[0025] The risk quantification module outputs a supply chain risk map, which is stored in the form of a graph database, recording the operating status, risk level, and correlation strength between each node.

[0026] The data sharing platform includes:

[0027] The data access interface is used to parse JSON, XML, CSV and SQL data formats and write them to a unified database;

[0028] The data processing module is used to perform data cleaning, normalization, feature extraction, trend recognition, and model inference operations in a distributed computing environment.

[0029] The visualization module is used to generate multi-dimensional data views for different departments. These views include demand forecast curves, risk heatmaps, inventory turnover charts, and transportation route visualizations.

[0030] The status synchronization module is used to achieve real-time status synchronization and access control between terminals in various departments and the platform through message queues or publish-subscribe mechanisms.

[0031] The generation of the risk warning signal includes the following steps:

[0032] A method for measuring the difference between predicted and actual values ​​is established, wherein the method for measuring the difference is calculated using weighted absolute deviation or mean square error.

[0033] The difference value is compared with the set multi-level thresholds. When the difference exceeds the first-level threshold, a low-level warning is generated. When the difference exceeds the second-level threshold, a medium-level warning is generated. When the difference exceeds the third-level threshold, a high-level warning is generated.

[0034] The risk warning signals specifically include production schedule deviation signals, supplier delivery delay signals, transportation congestion signals, inventory backlog signals, equipment failure signals, and financial anomaly signals.

[0035] The hierarchical management of the risk warning signals includes:

[0036] The criticality coefficient of a node is calculated based on a graph theory model of a supply chain network. The coefficient is determined by the degree centrality, betweenness centrality, and compact centrality of the node.

[0037] Signal priority is determined based on a combination of node criticality coefficient and risk level.

[0038] The high-priority signals are transmitted to the management monitoring terminal via a high-reliability communication channel, while the low-priority signals are automatically distributed to the corresponding responsible department system via a conventional data link.

[0039] The signal distribution process is implemented through a distributed message queue, ensuring that signals of different levels arrive at the corresponding processing nodes within a specified time.

[0040] The supply chain planning adjustments include:

[0041] Construct a multi-dimensional set of constraints, including production capacity limits, inventory safety levels, procurement cycles, and transportation timeliness.

[0042] The set of constraints is input into the optimization engine, which calculates the adjustment scheme based on both internal and external factors. The internal factors include inventory levels, production capacity, and procurement plans, while the external factors include supplier delivery capabilities, market demand fluctuations, and transportation conditions.

[0043] The adjustment plan is output in a standardized data format, including a production adjustment table, an inventory replenishment list, and transportation scheduling instructions.

[0044] The optimization engine employs a hybrid algorithm, and its optimization model is defined by the following formula:

[0045] The constraints are:

[0046]

[0047]

[0048]

[0049] in, For production scheduling variables, For inventory replenishment variables, Choose variables for transportation routes, either or . , , These are production costs, inventory holding costs, and transportation costs. , , These are weighting coefficients. For production capacity, To maintain a safe inventory level, For transportation capacity, This refers to the volume of goods transported along the transportation route.

[0050] The risk assessment model further includes an external environment monitoring unit, which is used to collect information on global economic indicators, international trade policies, climate conditions and natural disasters, establish a mapping relationship between the external factors and supply chain nodes, calculate the probability of the impact of external events on the node status, and generate a dynamic risk matrix that includes a time dimension.

[0051] The backtracking analysis module includes a historical comparison unit, an error identification unit, a parameter correction unit, and an update unit. The historical comparison unit is used to compare predicted values ​​with actual data. The error identification unit is used to distinguish between data acquisition errors, model calculation errors, and threshold setting errors. The parameter correction unit uses a gradient descent update method to correct model parameters and risk thresholds. The update unit inputs the correction results into the prediction and risk assessment of the next cycle.

[0052] (III) Beneficial Effects

[0053] The beneficial effects of this invention are:

[0054] 1. In this invention, by establishing a multi-source data access channel, data from ERP systems, MES systems, WMS systems, external data platforms, and IoT devices are collected in a unified manner. By combining historical data and real-time data, a multi-dimensional time-series database is constructed, realizing the complete acquisition and structured storage of global supply chain information and ensuring the comprehensiveness and accuracy of data input.

[0055] 2. In this invention, a risk assessment model is constructed that includes a risk identification module and a risk quantification module. Key nodes are identified by means of supplier delivery cycle monitoring, production equipment operation status sampling, inventory turnover rate calculation and transportation route congestion detection. A supply chain risk map is generated based on failure probability, topology weight and node dependency, thereby realizing quantitative description and real-time visualization of supply chain risks.

[0056] 3. In this invention, the prediction results and risk signals are transmitted in real time through a data sharing platform. The platform includes a data access interface, a data processing module, a visualization display module, and a status synchronization module. It not only supports multi-format data parsing and distributed computing, but also generates demand prediction curves, risk heat maps, inventory flow charts, and transportation route visualizations for various departments, ensuring information consistency and timely response during cross-departmental collaboration.

[0057] 4. In this invention, by introducing a hybrid optimization algorithm into the supply chain planning adjustment process, the objective function and constraints are first established using linear programming to solve the initial feasible solution. Then, global iterative optimization is performed based on a genetic algorithm to obtain the optimal solution, which includes the production schedule, inventory allocation scheme, and transportation route selection scheme. The results are then pushed to the production, warehousing, and logistics departments through a data sharing platform, thereby realizing the dynamic optimization and efficient execution of the supply chain plan. Attached Figure Description

[0058] Figure 1 This is a schematic diagram of the workflow of the present invention;

[0059] Figure 2 This is a flowchart illustrating the early warning logic section of the present invention;

[0060] Figure 3 This is a flowchart illustrating the working method of the present invention. Detailed Implementation

[0061] To better explain and facilitate understanding of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0062] Please refer to Figures 1 to 3 As shown, the supply chain-based data analysis method of the present invention includes the following steps:

[0063] S1: Collect supply chain data from internal systems and external sources. Internal data includes production plans, production progress, inventory levels, purchase orders and sales records. External data includes logistics and transportation information, market demand index, consumer behavior characteristics, weather monitoring data, road traffic flow and policy and regulatory changes.

[0064] S2: Preprocess the data, including removing outliers, merging duplicate data, filling missing values, standardizing the data, and performing feature extraction operations, thereby forming a structured, multi-dimensional dataset that can be used for computation.

[0065] S3: Based on the dataset, time series analysis, cluster analysis and pattern recognition methods are used to predict trends and identify potential risk signals in the operation status of the supply chain;

[0066] S4: Construct a risk assessment model that includes a risk identification module and a risk quantification module, monitor each node of the supply chain in real time, and generate early warning signals with risk levels;

[0067] S5: By establishing a unified data sharing platform, the forecast results and risk warning signals will be transmitted to relevant departments in real time for each department to obtain and use for business adjustments;

[0068] S6: Based on risk warning signals and feedback from various departments, coordinate across departments to adjust production, procurement, inventory, and transportation plans.

[0069] S7: Through backtracking analysis and continuous monitoring, correct the parameters of the prediction model and the thresholds for risk assessment, and update the application in the next cycle.

[0070] Optionally, the data collection steps further include:

[0071] Establish multi-source data access channels, including data interface calls to the enterprise's internal ERP system, MES system, and WMS system, as well as data collection from external third-party data service platforms and IoT devices;

[0072] Historical data collection includes production, sales, inventory, and procurement data for at least one cycle over the past five years, which are used to train the predictive model;

[0073] Real-time data is acquired through API calls, sensor sampling, and web crawling. Real-time and historical data are then aligned with timestamps and mapped to form a multi-dimensional time-series database. In practice, the data collection process establishes a unified connection between internal and external systems by creating multi-source data access channels. Data from internal ERP, MES, and WMS systems is accessed through standardized interfaces, while data from external third-party data service platforms and IoT devices is acquired via API interfaces and sensor sampling. Historical data is stored in the time-series database, covering complete cycle information for production, sales, inventory, and procurement over the past five years, serving as training data for predictive models. Real-time data is dynamically acquired through web crawling and sensor sampling, and after timestamp alignment and field mapping, it is fused with historical data to form a multi-dimensional time-series database capable of supporting time-series analysis and feature extraction, ensuring data consistency and availability during transmission, parsing, and storage.

[0074] Optionally, the risk assessment model includes:

[0075] The risk identification module is used to identify key risk nodes by statistically analyzing supplier delivery cycles, sampling production equipment operating status in real time by sensors, dynamically calculating inventory turnover rate, monitoring traffic flow for transportation routes, and combining anomaly detection algorithms.

[0076] The risk quantification module is used to calculate risk impact factors for key nodes. These factors include the node failure probability, the node's topological weight in the supply chain network, and the node's dependence on neighboring nodes.

[0077] The risk quantification module outputs a supply chain risk map, stored in a graph database, recording the operational status, risk level, and inter-node correlation strength of each node. In actual implementation, the risk assessment model operates collaboratively with the risk identification and risk quantification modules. The risk identification module monitors the production process based on statistical analysis of supplier delivery cycles and real-time sensor sampling of production equipment operation status. Simultaneously, it monitors the distribution process by dynamically calculating inventory turnover and collecting traffic flow data along transportation routes. Anomaly detection algorithms are used to identify key risk nodes. The risk quantification module builds a mathematical model by calculating the failure probability of nodes, their topological weight in the network, and their dependence on neighboring nodes, thereby generating the supply chain risk map. This risk map is stored in the graph database, recording node status, risk level, and correlation strength in a graph structure, achieving a networked representation and quantitative analysis of risk.

[0078] Optionally, the data sharing platform includes:

[0079] The data access interface is used to parse JSON, XML, CSV and SQL data formats and write them to a unified database;

[0080] The data processing module is used to perform data cleaning, normalization, feature extraction, trend recognition, and model inference operations in a distributed computing environment.

[0081] The visualization module is used to generate multi-dimensional data views for different departments, including demand forecast curves, risk heat maps, inventory turnover charts, and transportation route visualizations.

[0082] The status synchronization module is used to achieve real-time status synchronization and access control between terminals in various departments and the platform through message queues or publish-subscribe mechanisms. In actual implementation, the data sharing platform uses a data access interface to uniformly parse JSON, XML, CSV, and SQL format data and write it to the database. The data processing module runs in a distributed computing environment, performing cleaning, normalization, feature extraction, and trend identification on the accessed data, and generating prediction and analysis results in conjunction with the model inference process. The visualization module generates charts of different dimensions from the analysis results, including demand forecast curves, risk heatmaps, inventory turnover charts, and transportation route visualizations, to meet the usage needs of different departments. The status synchronization module uses message queues and publish-subscribe mechanisms to achieve real-time data synchronization and access control between terminals in various departments and the platform, thereby ensuring the integrity and consistency of the data processing chain.

[0083] Optionally, the generation of a risk warning signal includes the following steps:

[0084] Establish a method for measuring the difference between predicted and actual values. The method for measuring the difference uses either weighted absolute deviation or mean square error.

[0085] The difference value is compared with the set multi-level thresholds. When the difference exceeds the first-level threshold, a low-level warning is generated. When the difference exceeds the second-level threshold, a medium-level warning is generated. When the difference exceeds the third-level threshold, a high-level warning is generated.

[0086] Risk warning signals specifically include production schedule deviation signals, supplier delivery delay signals, transportation congestion signals, inventory backlog signals, equipment failure signals, and financial anomaly signals. In actual implementation, risk warning signals are generated based on the calculation of the difference between predicted and actual values. The difference measurement method includes a combination of weighted absolute deviation and mean square error. The system compares the predicted data with real-time monitoring data in each period, and then compares the difference value with a set multi-level threshold. When the difference value exceeds the first-level threshold, the system marks it as a low-level warning; when it exceeds the second-level threshold, it is marked as a medium-level warning; and when it exceeds the third-level threshold, it is marked as a high-level warning. The types of warning signals are classified according to the source of the difference, including production schedule deviation, supplier delivery delay, transportation congestion, inventory backlog, equipment failure, and financial anomaly. After generation, the signals enter a unified warning management module for subsequent distribution.

[0087] Optionally, the tiered management of risk warning signals includes:

[0088] The graph theory model based on the supply chain network is used to calculate the criticality coefficient of nodes. The coefficient is jointly determined by the degree centrality, betweenness centrality and compact centrality of the nodes.

[0089] Signal priority is determined based on a combination of node criticality coefficient and risk level.

[0090] High-priority signals are transmitted to the management monitoring terminal via a high-reliability communication channel, while low-priority signals are automatically distributed to the corresponding responsible department system via a regular data link.

[0091] The signal distribution process is implemented through a distributed message queue, ensuring that signals of different levels arrive at their corresponding processing nodes within a specified time. In actual implementation, the hierarchical management of risk warning signals is achieved through a graph theory model of the supply chain network. The node criticality coefficient is calculated using degree centrality, betweenness centrality, and close centrality indices. The system quantifies the role of different nodes in the network based on this coefficient, and then calculates a comprehensive priority based on the risk level. During signal distribution, high-priority signals are transmitted to the management monitoring terminal through a high-reliability communication channel to ensure timely processing, while low-priority signals are automatically allocated to the corresponding responsible department's system for processing via a regular data link. The entire distribution process is implemented through a distributed message queue, ensuring that signals can be accurately delivered to the target terminal within a specified time frame according to the set priority in a multi-node environment.

[0092] Optionally, supply chain planning adjustments may include:

[0093] Construct a multi-dimensional set of constraints, including production capacity limits, inventory safety levels, procurement cycles, and transportation timeliness.

[0094] The set of constraints is input into the optimization engine, which calculates the adjustment scheme based on both internal and external factors. The internal factors include inventory levels, production capacity, and procurement plans, while the external factors include supplier delivery capabilities, market demand fluctuations, and transportation conditions.

[0095] The adjustment plan is output in a standardized data format, including a production adjustment table, an inventory replenishment list, and transportation scheduling instructions. In actual implementation, supply chain planning adjustments first construct a multi-dimensional set of constraints, including production capacity limits, inventory safety levels, procurement cycles, and transportation timeliness. Internal factors such as inventory levels, production capacity, and procurement plans are collected in real time through the enterprise's internal systems, while external factors such as supplier delivery capabilities, market demand fluctuations, and transportation conditions are acquired through external platforms and sensors. After obtaining the complete set of constraints, the optimization engine generates an adjustment plan based on a comprehensive calculation of internal and external factors. The adjustment plan is output in a standardized data format, including a production adjustment table, an inventory replenishment list, and transportation scheduling instructions, for use by the execution systems of various departments, thereby achieving the feasibility and consistency of dynamic supply chain adjustments.

[0096] Optionally, the optimization engine employs a hybrid algorithm, and its optimization model is defined by the following formula:

[0097] The constraints are:

[0098]

[0099]

[0100]

[0101] in, For production scheduling variables, For inventory replenishment variables, Choose a variable for the transportation route, either 0 or 1. , , These are production costs, inventory holding costs, and transportation costs. , , These are weighting coefficients. For production capacity, To maintain a safe inventory level, For transportation capacity, The transportation route capacity is considered. In actual implementation, the optimization engine uses a hybrid algorithm to establish a mathematical model. The objective function is defined as the weighted sum of production cost, inventory holding cost, and transportation cost, with minimization as the optimization objective. Variables include production scheduling variables, inventory replenishment variables, and transportation route selection variables. The transportation route selection variable is in binary form (0 or 1). Constraints include production capacity, inventory safety level, and transportation capacity limits. The algorithm first uses linear programming to initially solve the model and obtain feasible solutions. Then, a genetic algorithm iteratively optimizes the initial solutions. The genetic algorithm uses the objective function as the fitness function and gradually improves candidate solutions through selection, crossover, and mutation operations, ultimately outputting the globally optimal solution.

[0102] Optionally, the risk assessment model further includes an external environment monitoring unit, used to collect information on global economic indicators, international trade policies, climate conditions, and natural disasters. This unit establishes a mapping relationship between external factors and supply chain nodes, calculates the probability of external events affecting node status, and generates a dynamic risk matrix with a time dimension. In practical implementation, the risk assessment model includes an external environment monitoring unit that connects to a global economic indicator database, an international trade policy platform, a climate monitoring system, and a natural disaster monitoring system via a data acquisition interface. An external event correlation module establishes a mapping relationship between the collected data and supply chain nodes, calculates the probability of external events affecting the operational status of each node, and a dynamic risk matrix generation module combines external factors and node status to form a risk matrix with a time dimension. This module also updates the risk levels in the risk map in real time, ensuring that the model maintains the dynamism and completeness of the risk assessment even when the external environment changes rapidly.

[0103] Optionally, the backtracking analysis module includes a historical comparison unit, an error identification unit, a parameter correction unit, and an update unit. The historical comparison unit compares predicted values ​​with actual data; the error identification unit distinguishes between data acquisition errors, model calculation errors, and threshold setting errors; the parameter correction unit uses a gradient descent update method to correct model parameters and risk thresholds; and the update unit inputs the correction results into the prediction and risk assessment for the next cycle. In actual implementation, the backtracking analysis module consists of these units. The historical comparison unit compares historical predicted values ​​with actual data to generate an error vector; the error identification unit classifies the sources of error, identifying data acquisition errors, model calculation errors, and threshold setting errors; the parameter correction unit adjusts model parameters and risk thresholds based on the gradient descent method; and the update unit writes the corrected parameters into the prediction and risk assessment model for application in the next cycle, forming a dynamic iterative optimization process. This ensures that the prediction and risk assessment model can continuously self-correct and improve based on historical performance.

[0104] Working principle: In the data acquisition phase, the system establishes a multi-source data access channel to combine production plans, production progress, inventory levels, purchase orders, and sales records from the enterprise's internal ERP, MES, and WMS systems with external data sources. External data sources include logistics and transportation information, market demand index, consumer behavior characteristics, meteorological monitoring data, road traffic flow, and policy and regulatory changes. Historical data is used to train the prediction model, while real-time data is obtained through API calls, sensor sampling, and web crawling. After timestamp alignment and field mapping, the data is stored in a multi-dimensional time-series database.

[0105] Secondly, in the data preprocessing stage, outlier removal, duplicate data merging, missing value imputation, normalization and feature extraction are performed on the collected data to construct a structured, multi-dimensional and computable dataset. This dataset provides the foundation for subsequent time series analysis, cluster analysis and pattern recognition, and can ensure the consistency and reliability of model input.

[0106] In the prediction and risk identification phase, the system identifies supply chain operation trends based on the dataset through time series modeling and pattern recognition methods, and captures potential risk signals by combining anomaly detection algorithms. The prediction results and risk signals are input into the risk assessment model, which consists of a risk identification module and a risk quantification module. The risk identification module monitors supplier delivery cycles, production equipment operating status, inventory turnover rate, and transportation route congestion to identify key risk nodes. The risk quantification module constructs a supply chain risk map based on factors such as node failure probability, node topology weight, and node dependence. The risk map is stored in the form of a graph database, recording the operating status of each node, risk level, and correlation strength between nodes.

[0107] During the data sharing phase, all forecast results and risk signals are distributed in real time through the data sharing platform. This platform includes a data access interface, a data processing module, a visualization module, and a status synchronization module. The data access interface is responsible for parsing and storing multi-format data. The data processing module performs trend identification and inference operations. The visualization module generates demand forecast curves, risk heat maps, inventory turnover charts, and transportation route visualizations. The status synchronization module ensures the consistency of information between the terminals of each department and the platform.

[0108] During the risk warning and classification stage, the system generates risk warning signals by comparing the difference between predicted and actual values. The difference is calculated using weighted absolute deviation or mean square error. According to the threshold classification mechanism, different deviation levels trigger low, medium, and high-level warning signals respectively. Signal types include production schedule deviation, supplier delivery delay, transportation congestion, inventory backlog, equipment failure, and financial anomalies. Furthermore, based on the graph theory model of the supply chain network, the system calculates the node criticality coefficient and determines the signal priority in combination with the risk level. High-priority signals are transmitted to the management monitoring terminal through a high-reliability channel, while low-priority signals are automatically allocated to the responsible departments via regular links. The distribution process is implemented by a distributed message queue.

[0109] During the supply chain planning adjustment phase, the system performs optimization calculations based on a multi-dimensional set of constraints, including production capacity limits, inventory safety levels, procurement cycles, and transportation timeliness. The optimization engine employs a hybrid algorithm, and the optimization model is defined by the following formula:

[0110] The constraints are:

[0111]

[0112]

[0113]

[0114] in, For production scheduling variables, For inventory replenishment variables, Choose a variable for the transportation route, either 0 or 1. , , These are production costs, inventory holding costs, and transportation costs. , , These are weighting coefficients. For production capacity, To maintain a safe inventory level, For transportation capacity, To determine the transport route capacity, an initial solution is obtained by linear programming, and then iterative optimization is performed based on a genetic algorithm to obtain the global optimal solution. The final output includes a production schedule, an inventory replenishment list, and transport scheduling instructions, which are then pushed to the production, warehousing, and logistics departments for execution through a data sharing platform.

[0115] During the external environment monitoring phase, the system incorporates global economic indicators, international trade policies, climate conditions, and natural disaster information to calculate the potential impact probability of external events on supply chain nodes, forming a dynamic risk matrix that includes a time dimension, thereby achieving real-time quantification of external shocks. Finally, in the retrospective analysis phase, the system compares historical forecasts with actual data, uses an error identification unit to distinguish between data collection errors, model calculation errors, and threshold setting errors, and uses a gradient descent update method through a parameter correction unit to correct the forecast model parameters and risk thresholds. The update unit then inputs the corrected parameters into the forecast and risk assessment model for the next cycle, achieving dynamic iterative optimization of the model.

[0116] The above description shows and illustrates the basic principles, main features, and advantages of the present invention. Standard parts used in the present invention can be purchased from the market, and irregular parts can be customized according to the description and drawings. The specific connection methods of each part adopt conventional methods such as bolts, rivets, and welding that are mature in the prior art. The machinery, parts, and equipment adopt conventional models in the prior art, and the circuit connection adopts conventional connection methods in the prior art, which will not be described in detail here.

[0117] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A data analysis method based on the supply chain, characterized in that, Includes the following steps: S1: Collect supply chain data from internal systems and external sources. The internal data includes production plans, production progress, inventory levels, purchase orders and sales records. The external data includes logistics and transportation information, market demand index, consumer behavior characteristics, meteorological monitoring data, road traffic flow and policy and regulatory changes. S2: Preprocess the data, including removing outliers, merging duplicate data, filling in missing values, standardizing the data, and performing feature extraction operations, thereby forming a structured, multi-dimensional dataset that can be used for computation. S3: Based on the dataset, time series analysis, cluster analysis and pattern recognition methods are used to predict trends and identify potential risk signals in the supply chain operation status; S4: Construct a risk assessment model that includes a risk identification module and a risk quantification module, monitor each node of the supply chain in real time, and generate early warning signals with risk levels; S5: By establishing a unified data sharing platform, the forecast results and risk warning signals will be transmitted to relevant departments in real time for each department to obtain and use for business adjustments; S6: Based on the aforementioned risk warning signals and feedback information from various departments, cross-departmental collaborative adjustments are made to production, procurement, inventory, and transportation plans; S7: Through backtracking analysis and continuous monitoring, correct the parameters of the prediction model and the thresholds for risk assessment, and update the application in the next cycle.

2. The data analysis method based on the supply chain according to claim 1, characterized in that: The data collection step further includes: Establish a multi-source data access channel, which includes data interface calls to the enterprise's internal ERP system, MES system, and WMS system, as well as data collection from external third-party data service platforms and IoT devices; The historical data collection includes production, sales, inventory, and procurement data for at least one period over the past five years, which are used to train the predictive model; The real-time data is acquired through API calls, sensor sampling, and web crawling. The real-time data and historical data are aligned with timestamps and mapped to form a multi-dimensional time-series database.

3. The data analysis method based on the supply chain according to claim 1, characterized in that: The risk assessment model includes: The risk identification module is used to identify key risk nodes by statistically analyzing supplier delivery cycles, sampling production equipment operating status in real time by sensors, dynamically calculating inventory turnover rate, monitoring traffic flow for transportation routes, and combining anomaly detection algorithms. The risk quantification module is used to calculate risk impact factors for the key nodes. These factors include the node failure probability, the node's topological weight in the supply chain network, and the node's dependence on its neighboring nodes. The risk quantification module outputs a supply chain risk map, which is stored in the form of a graph database, recording the operating status, risk level, and correlation strength between each node.

4. The data analysis method based on the supply chain according to claim 1, characterized in that: The data sharing platform includes: The data access interface is used to parse JSON, XML, CSV and SQL data formats and write them to a unified database; The data processing module is used to perform data cleaning, normalization, feature extraction, trend recognition, and model inference operations in a distributed computing environment. The visualization module is used to generate multi-dimensional data views for different departments. These views include demand forecast curves, risk heatmaps, inventory turnover charts, and transportation route visualizations. The status synchronization module is used to achieve real-time status synchronization and access control between terminals in various departments and the platform through message queues or publish-subscribe mechanisms.

5. The data analysis method based on the supply chain according to claim 1, characterized in that: The generation of the risk warning signal includes the following steps: A method for measuring the difference between predicted and actual values ​​is established, wherein the method for measuring the difference is calculated using weighted absolute deviation or mean square error. The difference value is compared with the set multi-level thresholds. When the difference exceeds the first-level threshold, a low-level warning is generated. When the difference exceeds the second-level threshold, a medium-level warning is generated. When the difference exceeds the third-level threshold, a high-level warning is generated. The risk warning signals specifically include production schedule deviation signals, supplier delivery delay signals, transportation congestion signals, inventory backlog signals, equipment failure signals, and financial anomaly signals.

6. The data analysis method based on the supply chain according to claim 5, characterized in that: The hierarchical management of the risk warning signals includes: The criticality coefficient of a node is calculated based on a graph theory model of a supply chain network. The coefficient is determined by the degree centrality, betweenness centrality, and compact centrality of the node. Signal priority is determined based on a combination of node criticality coefficient and risk level. The high-priority signals are transmitted to the management monitoring terminal via a high-reliability communication channel, while the low-priority signals are automatically distributed to the corresponding responsible department system via a conventional data link. The signal distribution process is implemented through a distributed message queue, ensuring that signals of different levels arrive at the corresponding processing nodes within a specified time.

7. The data analysis method based on the supply chain according to claim 1, characterized in that: The supply chain planning adjustments include: Construct a multi-dimensional set of constraints, including production capacity limits, inventory safety levels, procurement cycles, and transportation timeliness. The set of constraints is input into the optimization engine, which calculates the adjustment scheme based on both internal and external factors. The internal factors include inventory levels, production capacity, and procurement plans, while the external factors include supplier delivery capabilities, market demand fluctuations, and transportation conditions. The adjustment plan is output in a standardized data format, including a production adjustment table, an inventory replenishment list, and transportation scheduling instructions.

8. The data analysis method based on the supply chain according to claim 7, characterized in that: The optimization engine employs a hybrid algorithm, and its optimization model is defined by the following formula: The constraints are: in, For production scheduling variables, For inventory replenishment variables, Choose a variable for the transportation route, either 0 or 1. , , These are production costs, inventory holding costs, and transportation costs. , , These are weighting coefficients. For production capacity, To maintain a safe inventory level, For transportation capacity, This refers to the volume of goods transported along the transportation route.

9. The data analysis method based on the supply chain according to claim 1, characterized in that: The risk assessment model further includes an external environment monitoring unit, which is used to collect information on global economic indicators, international trade policies, climate conditions and natural disasters, establish a mapping relationship between the external factors and supply chain nodes, calculate the probability of the impact of external events on the node status, and generate a dynamic risk matrix that includes a time dimension.

10. The data analysis method based on the supply chain according to claim 1, characterized in that: The backtracking analysis module includes a historical comparison unit, an error identification unit, a parameter correction unit, and an update unit. The historical comparison unit is used to compare predicted values ​​with actual data. The error identification unit is used to distinguish between data acquisition errors, model calculation errors, and threshold setting errors. The parameter correction unit uses a gradient descent update method to correct model parameters and risk thresholds. The update unit inputs the correction results into the prediction and risk assessment of the next cycle.

Images