Supply chain order placing management method and system based on big data analysis

Abstract

The invention provides a supply chain order placing management method and system based on big data analysis, and relates to the technical field of supply chain management. The method comprises the following steps: respectively carrying out qualification hierarchical auditing on a purchaser and a supplier to obtain a comprehensive auditing result; the purchaser multi-source data and the supplier multi-source data are integrated, and a dynamic demand-supply cooperation plan is generated; according to a purchaser order, based on a compliance audit result, automatically matching a compliance and adaptive supplier list; a multi-level supply chain bidirectional view is constructed, and cross-level commodity rapid positioning and inventory synchronization are realized through cascade screening; supplier performance data, purchaser demand change data and environment data are monitored in real time, and dynamic risks actually occurring in the performance process are identified; and according to the dynamic risk linkage access risk file, triggering graded early warning and dynamically adjusting the operation authority, thereby effectively solving the problems existing in the traditional supply chain order placing management method, and improving the transparency, flexibility and risk management capability of the supply chain.

Description

Technical Field

[0001] The present application relates to the field of supply chain management technology, and in particular to a supply chain order management method and system based on big data analysis. Background Art

[0002] In today's business environment, where globalization and digitalization are deeply integrated, supply chain management has become a key factor in enhancing corporate competitiveness and achieving sustainable development. With increasingly fierce market competition and evolving consumer demands, the complexity and uncertainty of supply chains have increased significantly, placing higher demands on supply chain order management.

[0003] Traditional supply chain order management methods often rely on manual operations and simple information systems, which present numerous limitations. For one thing, during the qualification review process, traditional methods often employ single, static criteria, making it difficult to comprehensively and accurately assess the qualifications of both buyers and suppliers. For example, superficial reviews based solely on the limited qualification documents submitted by companies lack in-depth analysis and dynamic monitoring of compliance and potential entry risks. This results in some companies with flawed or high-risk qualifications entering the supply chain, creating hidden dangers for subsequent contract fulfillment.

[0004] In traditional supply chains, systems at different levels (such as buyers, suppliers, and distributors) are often independent, making data sharing and circulation difficult. For example, the lack of effective data interfaces between buyers' inventory management systems and suppliers' production management systems prevents buyers from obtaining real-time information on suppliers' raw material inventory and production progress, while suppliers are also unaware of buyers' real-time sales data and inventory dynamics. This information fragmentation makes it difficult to locate products across multiple levels, making it difficult for companies to accurately grasp the location and status of products throughout the supply chain. Due to the lack of a unified information platform and communication mechanism, information transmission between different levels is often subject to delays and errors. For example, when a buyer discovers that a particular product is low in stock, they need to spend a considerable amount of time communicating with the supplier via phone, email, and other means to understand their supply capacity and delivery time. During this process, information can be distorted due to human factors, leading to poor decision-making.

[0005] In traditional supply chains, product location often relies on manual queries and paper records. For example, warehouse managers must manually locate products in the warehouse or review paper ledgers to understand inventory quantities. This approach is not only inefficient but also prone to errors, especially when dealing with a wide variety of products and large inventory levels.

[0006] Inventory synchronization in traditional supply chains often relies on regular manual inventory counts and reporting, lacking a real-time inventory synchronization mechanism. For example, after a supplier ships goods, they must wait for the buyer to complete receipt and warehousing before updating inventory information. This process can cause lags in inventory data, leading to inconsistent inventory information between buyers and suppliers and hindering supply chain collaboration. Summary of the Invention

[0007] In response to the above problems, this application proposes a supply chain order management method and system based on big data analysis. By using big data analysis, a more comprehensive and in-depth hierarchical review of the qualifications of purchasers and suppliers is conducted, multi-source data is integrated to generate dynamic demand-supply collaboration plans, and performance risks are monitored in real time. The access risk files are linked to early warnings and authority adjustments, thereby effectively solving the problems existing in traditional supply chain order management methods and improving the transparency, flexibility and risk management capabilities of the supply chain.

[0008] The purpose of this application is achieved by the following technical solutions: In a first aspect, the present application provides a supply chain order management method based on big data analysis, the method comprising: Conduct a tiered review of the qualifications of both the purchaser and supplier to obtain comprehensive review results; the tiered review includes compliance review and access risk assessment; Integrate multi-source data from buyers and suppliers to generate dynamic demand-supply collaboration plans; Based on the purchaser's order and compliance review results, the system automatically matches a list of compliant and suitable suppliers and generates a recommended supplier list. Build a two-way view of the multi-level supply chain and achieve rapid cross-level product location and inventory synchronization through cascading screening; Real-time monitoring of supplier performance data, purchaser demand change data, and environmental data, and identification of actual dynamic risks occurring during the performance process through performance risk monitoring models; According to the dynamic risk linkage access risk file, a graded warning is triggered and the operation authority is dynamically adjusted.

[0009] In a second aspect, a supply chain order management system based on big data analysis is used to implement the supply chain order management method described in this application, and the system includes: The audit module is used to conduct a hierarchical audit of the qualifications of the purchaser and supplier respectively to obtain a comprehensive audit result; the hierarchical audit includes compliance audit and access risk assessment; Collaborative planning acquisition module, used to integrate multi-source data from buyers and suppliers to generate dynamic demand-supply collaborative plans; The recommendation module is used to automatically match a list of compliant and suitable suppliers based on the purchaser's order and the compliance review results, and generate a list of recommended suppliers; The search and positioning module is used to build a two-way view of the multi-level supply chain, and achieve rapid cross-level product location and inventory synchronization through cascading screening; The dynamic monitoring module is used to monitor supplier performance data, purchaser demand change data and environmental data in real time, and identify dynamic risks that actually occur during the performance process through the performance risk monitoring model; The early warning module is used to trigger graded early warnings and dynamically adjust operation permissions based on the dynamic risk linkage access risk file.

[0010] The beneficial effects of the present invention include: by conducting a hierarchical review of the qualifications of purchasers and suppliers respectively, covering compliance review and access risk assessment, it is possible to comprehensively evaluate the qualifications of enterprises from multiple dimensions. It avoids the loopholes that may exist in traditional single audit methods and reduces the supply chain risks caused by qualification issues. Utilizing the access risk prediction model, combined with the historical access failure case data set and the industry blacklist data set, a graph neural network is used to analyze the supply chain topology relationship, and a multi-agent interactive Monte Carlo simulation is used to quantify the cross-enterprise cascade risk caused by qualification defects, which can accurately output the risk probability and expected loss value in the future cycle; it helps enterprises to timely discover potential risks, take corresponding measures to prevent and respond, and reduce risk losses. By integrating the multi-source data of the purchaser and the multi-source data of the supplier, and performing real-time fusion processing of the multi-source data through the streaming computing engine, it is possible to identify the spatiotemporal conflict between the purchaser's sudden demand peak and the supplier's production capacity bottleneck, and dynamically disassemble the original order into batch delivery sub-plans based on the conflict location results, so that the supply chain can more accurately match demand and supply, and improve the accuracy and flexibility of collaborative planning. By establishing a unified spatiotemporal coordinate mapping model, the conflict area is marked and a supplier-purchaser conflict link map is generated. Based on this information, companies can rationally adjust inventory layout and replenishment strategies to optimize inventory management. Building a bidirectional view of the multi-tier supply chain allows for rapid location of products across tiers. Furthermore, cascading filters facilitate the search for products with specific attributes, improving product management efficiency. Using convolutional neural networks to predict regionalized shelf life decay curves, safety stock levels at each tier are dynamically adjusted. Integrating this with real-time monitoring data ensures accurate synchronization of inventory information across tiers. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 This is a schematic diagram of a supply chain order management method based on big data analysis provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] Below, the present application is further described in conjunction with the accompanying drawings and specific implementation methods. It should be noted that, under the premise of no conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0013] See also Figure 1 , an embodiment of the present application provides a supply chain order management method based on big data analysis, the method comprising: Conduct a tiered review of the qualifications of both the purchaser and supplier to obtain comprehensive review results; the tiered review includes compliance review and access risk assessment; Integrate multi-source data from buyers and suppliers to generate dynamic demand-supply collaboration plans; Based on the purchaser's order and compliance review results, the system automatically matches a list of compliant and suitable suppliers and generates a recommended supplier list. Build a two-way view of the multi-level supply chain and achieve rapid cross-level product location and inventory synchronization through cascading screening; Real-time monitoring of supplier performance data, purchaser demand change data, and environmental data, and identification of actual dynamic risks occurring during the performance process through performance risk monitoring models; According to the dynamic risk linkage access risk file, a graded warning is triggered and the operation authority is dynamically adjusted.

[0014] The working principle and effects of the above technical solution are as follows: We conduct a tiered review of the qualifications of both buyers and suppliers, encompassing compliance reviews and entry-level risk assessments. Compliance reviews primarily examine compliance with relevant laws, regulations, and industry standards; entry-level risk assessments assess the potential risks both parties face when entering into supply chain collaboration, such as the supplier's financial stability and the buyer's creditworthiness. Ultimately, we arrive at a comprehensive review outcome, providing a foundational guarantee for future collaboration.

[0015] By integrating multi-source data from buyers (such as purchasing history, demand forecasts, and sales data) and suppliers (such as production capacity, inventory levels, and lead times), and leveraging big data analytics, a dynamic demand-supply collaboration plan is generated. This plan can be adjusted in real time based on market changes, fluctuations in buyer demand, and other factors, ensuring buyers receive the necessary materials in a timely manner and suppliers can rationally arrange production and supply.

[0016] When a purchaser places an order, the system automatically matches a list of compliant and compatible suppliers from a wide range of suppliers based on compliance audit results. Compatibility considerations include product quality, price, delivery time, and service level. A list of recommended suppliers is then generated for the purchaser to select.

[0017] Build a multi-level, two-way view of the supply chain, integrating and displaying information from all links in the supply chain (such as buyers, suppliers, and logistics providers). Through cascading filtering, users can quickly locate the desired product in the supply chain, while also synchronizing inventory data in real time, ensuring accurate inventory status for all parties.

[0018] Real-time monitoring of supplier performance data (e.g., delivery progress, product quality), buyer demand change data (e.g., order quantity, delivery time), and environmental data (e.g., policy and regulatory changes, natural disasters, etc.). This data is fed into the performance risk monitoring model, which analyzes and identifies dynamic risks that actually occur during the performance process, such as supplier delivery delays and sudden increases in buyer demand.

[0019] Based on the dynamic risks identified, the system links access risk profiles and triggers graded warnings. Different levels of risk correspond to different warning methods and response measures. At the same time, operational permissions are dynamically adjusted. For example, in high-risk situations, certain operational permissions may be restricted to prevent further escalation of risks.

[0020] In a possible implementation, the qualification review of the purchaser and the supplier is conducted in a hierarchical manner to obtain a comprehensive review result, including: Receive qualification documents submitted by purchasers and suppliers respectively, and conduct basic compliance review of the purchaser or supplier's qualifications based on the qualification documents using a pre-set compliance rule library; the qualification documents include business information, credit records, industry certification certificates, and quality inspection reports; Input the compliance data into a pre-trained access risk prediction model; output an access risk determination result; the access risk determination result includes a risk probability score and a network-level loss expectation value; Combining the basic compliance review results and the access risk assessment results, a comprehensive qualification review conclusion is generated; Implement hierarchical authority control based on the audit conclusions.

[0021] In one possible implementation, qualification documents submitted by the purchaser and supplier are received respectively, and a basic compliance review of the purchaser's or supplier's qualifications is performed based on the qualification documents using a preset compliance rule library; including: For scanned industry certification certificates, OCR is used to identify the seal area and compare it with the preset certification agency seal library to verify authenticity; For product quality inspection reports, extract key inspection item data and match them with the industry safety threshold knowledge graph; When unstructured data verification is abnormal, the manual review process is automatically triggered and risk tags are added.

[0022] In one possible implementation, inputting the compliance data into a pre-trained access risk prediction model and outputting the access risk determination result include: The model is trained based on a historical dataset of failed access cases and an industry blacklist dataset; the historical dataset includes qualification defect types, chain risk events, and loss results; Use graph neural networks to analyze supply chain topology relationships, including OEM relationships, distribution links, and resource sharing relationships; Through multi-agent interactive Monte Carlo simulation, the cross-enterprise cascade risk caused by qualification defects is quantified, and the risk probability Pr and expected loss value in the future cycle are output.

[0023]

[0024] in, is the topological subgraph risk weight, which represents the risk transmission capability of the topological subgraph where the i-th associated entity (such as OEM manufacturer, distributor) in the supply chain network is located; The degree of qualification deficiency, , obtained by normalizing the basic compliance audit scores; is the industry correction factor (e.g., the frozen food industry has stricter requirements for temperature control qualifications and shelf life management); n is the number of entities in the supply chain network that have an associated relationship with the target entity (purchaser or supplier); It is a risk mapping function that describes the nonlinear coupling between the risk weight of the topological subgraph (Gᵢ) and the degree of qualification defects (Dᵢ), and outputs a risk probability value in the range of 0-1. Its specific form can be understood as an activation function in a graph neural network (such as Sigmoid, ReLU, etc.), or a probability conversion function in a Monte Carlo simulation. Its function is to convert the composite features of Gᵢ and Dᵢ into risk probabilities.

[0025] The working principle of the above technical solution is: The application and process of the access risk prediction model are detailed as follows: The pre-trained access risk prediction model is trained and constructed based on a historical access failure case dataset and an industry access blacklist dataset. The historical access failure case dataset covers the entire process of past cooperation failures due to qualification issues, including the type of qualification defects, the chain risk events triggered by the defects, and the final loss results. The industry access blacklist dataset integrates information on illegal enterprises, reasons for violations, and penalty records published by industry regulatory authorities. The access risk prediction model integrates a graph neural network to analyze the OEM relationship between suppliers and brands in the supply chain network, regional agent distribution links, and shared resource relationships of purchasers or suppliers, and outputs a risk probability score that represents hidden associated risks. When compliance data is input into the access risk prediction model, a multi-agent interactive Monte Carlo simulation is used to quantitatively simulate the cross-enterprise cascading risks caused by qualification defects in the supply chain network; the risk probability, network-level loss expectation value, and comprehensive risk score caused by qualification defects in the future cycle are output.

[0026] When verified qualification data is input into the access risk prediction model, the Monte Carlo simulation algorithm is integrated to quantitatively simulate the chain risks caused by potential qualification defects in the qualification documents. For example, if an expired supplier quality inspection report is identified as a qualification defect, the model simulates the complete risk transmission path of this defect leading to product sampling failure, which in turn leads to the purchasing store being punished by the regulatory authorities. The above simulation outputs two core results: First, the probability of risk caused by the qualification defect within a specified period in the future (e.g., three months), presented in numerical form, intuitively reflecting the likelihood of the risk occurring. For example, if a supplier has incomplete environmental protection qualifications, the output is the probability value of being penalized by the environmental protection department within three months; Second, the expected value of risk loss, which comprehensively considers the direct losses (such as the amount of fines) and indirect losses (such as loss of business volume due to damaged brand reputation) that may arise from each link in the chain risk, and is quantitatively calculated through an algorithm. When the comprehensive qualification review concludes with passing, the corresponding complete operation permissions will be unlocked to the purchaser or supplier account, covering product listing, full order processing (creation, modification, confirmation, etc.) and logistics tracking (logistics node query, abnormal feedback, etc.); if the conclusion is failing, the purchaser will be restricted to browsing the basic product catalog (only basic product information is displayed, without interactive functions such as placing orders and bargaining), and the supplier will be marked as a risk list and synchronized to the risk information sharing module of the supply chain collaboration platform for reference by other purchasers on the platform.

[0027] The method of combining the basic compliance verification results and the access risk determination results to generate a comprehensive qualification review conclusion includes: If a certain item in the qualification document is a mandatory compliance item in the industry, then the item will be regarded as a core item. If it is qualified, the item will be given a corresponding score according to the preset rules. If it does not meet the rules, it will directly result in failure of the review; If an item in the qualification document does not belong to the industry mandatory compliance items, then the item is an auxiliary item; the score for the item is obtained according to the auxiliary item scoring rules; Obtain basic compliance scores based on core and supplementary items; For example, the basic compliance score for scoring food suppliers is shown in Table 1: Table 1:

[0028] Take a weighted average of the risk probability score and network loss expectation score to obtain a comprehensive risk score; A weighted average of the basic compliance score and the comprehensive risk score is taken to obtain a final comprehensive score. Based on the basic compliance score and the final comprehensive score, a comprehensive qualification review conclusion is generated. The comprehensive qualification review conclusion includes pass, fail, risk probability, and expected loss. If any core item fails, the application is directly failed. If the final comprehensive score is less than the score threshold, the application is failed. The risk probability score is Rp=(1-Pr)×100; Pr is the risk probability predicted by the admission risk prediction model; The expected loss value is divided into: Rl=max(minimum preset score, (industry average loss - expected loss value)×100 / industry average loss)); the expected loss value is the expected loss value predicted by the access risk prediction model.

[0029] For example, if a supplier's environmental qualifications pass the basic verification but have a historical environmental penalty record (credit record verification item), the risk prediction model will extract data such as "number of environmental penalties in the past year" and "penalty amount" to calculate the risk probability (P): If the number of historical penalties is ≥ 2 times, the predicted probability of being penalized again within 3 months is P = 20%, then (Rp = (1-20%) × 100 = 80 points.

[0030] Calculate the expected value of loss: the industry average environmental penalty loss is 100,000 yuan, and the supplier's predicted loss = 100,000 yuan × (1 + brand reputation loss coefficient 30%) = 130,000 yuan; and the minimum value is the preset score of 0, so RL = 0 points.

[0031] In one possible implementation, the compliance review further includes: accessing the supplier's ESG (sustainable development, including environmental, social, and governance) data source to conduct a quantitative assessment of carbon emission intensity, labor compliance, and governance structure deficiencies; and incorporating ESG risk scores into the comprehensive qualification review conclusion.

[0032] The effects of the above technical solution are: By receiving a wealth of qualification documents submitted by buyers and suppliers, including business information, credit records, industry certifications, and quality inspection reports, and utilizing a pre-set compliance rule library to conduct basic compliance reviews, we ensure the comprehensiveness and accuracy of qualification reviews across multiple key dimensions. For example, the review of business information confirms a company's legal registration and business scope, while a review of its credit record provides insight into the company's credit standing and mitigates partnership risks.

[0033] For unstructured data such as scanned industry certification certificates and product quality inspection reports, we use OCR technology to identify the seal area and compare it with the pre-set certification agency seal library to verify authenticity. We then extract key inspection item data and match it with the industry safety threshold knowledge graph. This processing method effectively solves the challenge of unstructured data verification, improves the accuracy and efficiency of audits, and avoids cooperation risks caused by false or substandard qualification documents.

[0034] When unstructured data verification is abnormal, the manual review process is automatically triggered and risk tags are added, giving full play to the flexibility and accuracy of manual review and ensuring the reliability of the review results.

[0035] The access risk prediction model is trained on a dataset of historical access failure cases and industry blacklists. These datasets cover the entire process of failed collaborations due to qualification issues, including the types of qualification defects, cascading risk events, and resulting losses. This allows the model to learn a rich set of risk characteristics and patterns. Furthermore, a graph neural network is used to analyze supply chain topology relationships, and a multi-agent interactive Monte Carlo simulation is used to quantify the cross-enterprise cascading risks caused by qualification defects, further enhancing the scientific nature and accuracy of risk assessment.

[0036] Output the risk probability score and network-level loss expectation in the future cycle, present the risk assessment results in a quantitative form, and intuitively reflect the possibility of risk occurrence and the potential degree of loss, which helps enterprises to more clearly understand the cooperation risks and formulate corresponding risk response strategies.

[0037] When generating the comprehensive qualification review conclusion, the various items corresponding to the qualification documents are divided into core items and supplementary items. Core items are mandatory industry compliance items; failure to meet them will result in a direct audit failure, ensuring that the company possesses basic compliance qualifications. Supplementary items are scored according to scoring rules, taking into account the company's other strengths and weaknesses. The risk probability score and loss expectation score are weighted averaged to obtain the comprehensive risk score, and the basic compliance score and comprehensive risk score are weighted averaged to obtain the final comprehensive score. This comprehensive scoring system fully considers basic compliance and risk factors, making the review conclusion more scientific and objective.

[0038] Hierarchical permission control is implemented based on the audit conclusion. If the comprehensive qualification review is deemed passed, full operational permissions are unlocked for the purchaser or supplier account. If the review is deemed failed, the purchaser is restricted to browsing only the basic product catalog, while the supplier is marked as a risk and synchronized with the risk information sharing module of the supply chain collaboration platform. This hierarchical permission control effectively reduces collaboration risks and ensures the safe and stable operation of the supply chain.

[0039] In a possible implementation, the method further includes: Establish a real-time or regular data interface with an external authoritative data source; continuously monitor the change information of the purchaser or supplier, including changes in industrial and commercial status, new administrative penalty information, serious illegal and dishonest information, and major negative public opinion information; when key information changes that affect qualification compliance or risk level are monitored, automatically trigger the re-evaluation of the qualification status, and dynamically adjust the operating authority or risk mark of the purchaser or supplier based on the re-evaluation results.

[0040] This proactive risk warning mechanism enables enterprises to detect risks in the early stages of their occurrence, buying time to take appropriate risk prevention and control measures; it dynamically adjusts the purchaser's or supplier's operating permissions or risk tags based on the re-evaluation results, thus achieving dynamic management of risks.

[0041] In one possible implementation, integrating multi-source data from purchasers and multi-source data from suppliers to generate a dynamic demand-supply collaboration plan includes: Access the buyer's real-time inventory level, historical order fluctuation trends, in-transit logistics data, and market forecast demand; Synchronize suppliers' real-time capacity utilization, raw material inventory dynamics, production line scheduling data, and alternative supply source status; The streaming computing engine performs real-time fusion processing on multi-source data to identify the temporal and spatial conflicts between sudden demand peaks of buyers and production bottlenecks of suppliers. Based on the conflict location results, the original order is dynamically split into batch delivery sub-plans.

[0042] In one possible implementation, the real-time fusion processing of multi-source data by a streaming computing engine to identify the spatiotemporal conflict between the sudden demand peak of the purchaser and the production capacity bottleneck of the supplier includes: Establish a unified spatiotemporal coordinate mapping model to map the purchaser's demand location, supplier's production capacity location, and logistics hub nodes to a geographic grid coordinate system; Calculate in real time the ratio of the peak demand increment ΔD to the remaining capacity margin ΔPD in each grid: When ΔD / ΔPD>threshold K1, it is marked as a red conflict area; When ΔD / ΔPD∈[threshold K2, threshold K1], it is marked as a yellow warning area; Based on the topological relationship of the conflict area, a supplier-purchaser conflict link map is generated.

[0043] The dynamic disassembly of original orders includes: for orders in the red conflict area: splitting orders according to the supplier's capacity release cycle to generate a stepped delivery sequence (for example: T1 delivery 30%, T2 delivery 50%, T3 delivery 20%); For orders in the yellow warning area: split the orders according to the response priority of alternative supply sources and generate parallel supply sub-plans (for example: the main supplier takes 70% and the backup supplier takes 30%).

[0044] In one possible implementation, a multi-objective optimization model is constructed with the goal of minimizing total logistics costs and maximizing supply stability; Real-time congestion data of transportation routes, supplier geographic topology relationships, and storage node capacity constraints are embedded to generate a multi-level supply routing solution with time-sensitive labels.

[0045] In one possible implementation, when real-time monitoring detects abnormalities in supplier capacity or changes in purchaser demand, the collaborative plan incremental update mechanism is triggered; Based on the reinforcement learning algorithm, the optimal adjustment strategy for similar scenarios is matched in the historical optimization decision library, and the plan update weight coefficient is output to achieve dynamic rolling optimization of the plan.

[0046] The above technical solution achieves the following: By integrating multi-source data in real time through a streaming computing engine, it can promptly identify spatiotemporal conflicts between sudden peaks in buyer demand and bottlenecks in supplier capacity, enabling enterprises to quickly respond, adjust production plans and logistics arrangements, and improve supply chain responsiveness. When supplier capacity anomalies or buyer demand changes occur, a collaborative plan incremental update mechanism is triggered, and dynamic rolling optimization of the plan based on a reinforcement learning algorithm is implemented, enabling the supply chain to quickly adapt to changes and maintain efficient operation. Based on the different situations in the conflicting areas, orders are rationally broken down, employing a stepped delivery sequence or parallel supply sub-plans to fully utilize supplier capacity and resources from alternative sources, avoiding resource waste and improving resource utilization. A multi-objective optimization model is constructed and a multi-level supply routing solution with time-sensitive labels is generated. This minimizes total logistics costs while ensuring supply stability, reducing the enterprise's operating costs. By establishing a unified spatiotemporal coordinate mapping model and calculating the ratio of peak demand increments to remaining capacity margins in real time, conflicting areas can be accurately identified and early warnings issued, enabling enterprises to take proactive measures to avoid the risk of supply disruptions.

[0047] Integrating multi-source data from buyers and suppliers provides comprehensive and accurate information support for decision-making, enabling enterprises to use reinforcement learning algorithms to match the optimal adjustment strategies for similar scenarios in the historical optimization decision library, providing a scientific basis for plan updates and improving the accuracy and effectiveness of decision-making.

[0048] In one possible implementation, the multi-level supply chain bidirectional view is constructed to achieve rapid cross-level commodity location and inventory synchronization through cascading screening; including: Build a two-way view of the multi-level supply chain, including the buyer level and the supplier level; Build a two-way view of the multi-level supply chain, including the buyer level and the supplier level; Purchasing party level: headquarters (global inventory control) - regional center (provincial forward warehouse) - store (terminal inventory); Supplier level: manufacturer (raw material production capacity) - brand (finished product inventory) - regional agent (localized distribution inventory); Supports cross-level penetration query (e.g. headquarters can view the inventory of specific packages in a store); Filter products by temperature control requirements, shelf life thresholds, and cold chain certification status through cascading filters; The cascade filtering function supports the following food industry-specific conditions: Product attribute screening: filter by temperature control requirements (refrigerated / frozen / room temperature), remaining shelf life (≤30 days / ≤60 days), and cold chain certification (HACCP / ISO22000); Supply chain collaborative screening: sorting by supplier fulfillment time (48 hours / 72 hours), regional distribution coverage (e.g., Guangdong Province / South China), and historical order on-time rate (≥95% / ≥90%); Generate the topological path with the best shelf life loss based on real-time temperature and humidity data; The inventory synchronization mechanism includes: Real-time data linkage: When the freezer temperature is abnormal (e.g. > -15°C), the system automatically marks the corresponding inventory as "risk inventory" and freezes orders, while also triggering the "distribution vehicle management system" to adjust transportation routes; Intelligent replenishment model: Automatically generates replenishment recommendations based on historical store sales data and regional center inventory turnover rates (e.g., a store replenishes twice a week, with each replenishment quantity being 1.2 times the previous week's sales).

[0049] Predict regional shelf life decay curves based on convolutional neural networks and dynamically adjust safety stock levels at all levels .

[0050]

[0051] in, is a decay rate function that characterizes the shelf life loss rate of a product over time (e.g., the quality decay rate of frozen food under abnormal temperature conditions); The average daily demand for a specific product in a certain area or store, reflecting the basic demand level; The time required from order placement to product delivery (e.g., supplier production cycle + logistics delivery time) is used to calculate the basic inventory requirements during the lead time. z is the safety factor, a coefficient set based on the risk of demand fluctuations (e.g., the quantile under a normal distribution) to ensure sufficient inventory in response to sudden demand or supply delays. The standard deviation of daily demand is a quantitative indicator of demand fluctuation, reflecting the degree of demand uncertainty. The larger the standard deviation, the more severe the demand fluctuation, and the more safety stock needs to be reserved; is the cumulative decay, which represents the reduction in available inventory due to shelf life decay from the initial time t0 to the current time t.

[0052] In one possible implementation, the weight of the circulation path is modified based on real-time temperature and humidity data to generate the optimal topological path for shelf life loss; this includes: Based on the temperature and humidity data of the logistics vehicle, the remaining shelf life is dynamically calculated; specifically, the temperature and humidity data of the logistics vehicle are input into the shelf life attenuation acceleration model to dynamically calculate the remaining shelf life :

[0053] is the nominal shelf life, k is the temperature and humidity sensitivity coefficient, is the temperature and humidity offset; When the predicted remaining shelf life is less than the safe delivery threshold: This batch of goods will be automatically marked as "urgent delivery" and allocated to the nearest store first.

[0054] The above technical solution offers the following benefits: a multi-level supply chain bidirectional view and cross-level penetrating query capabilities enable companies to quickly locate product locations and inventory information, shortening information query time and improving management efficiency. The cascading filtering function supports filtering and sorting based on a variety of food industry-specific criteria, helping companies quickly identify qualified products and suppliers and optimize supply chain decisions. By adjusting distribution path weights using real-time temperature and humidity data, optimal topological routes are generated based on shelf life loss, minimizing product losses during transportation and storage, and lowering costs. An intelligent replenishment model and dynamic safety stock level adjustment mechanism enable precise inventory management, preventing overstocking and stockouts, and reducing inventory costs. Real-time data linkage promptly identifies risks such as abnormal freezer temperatures and initiates appropriate measures, such as freezing orders and adjusting transportation routes, to ensure product quality and safety, enhancing supply chain stability. When a product's remaining shelf life is insufficient, emergency distribution and delivery route replanning are automatically triggered, and a bidding process for alternative suppliers is initiated, ensuring timely product supply and improving supply chain reliability. The multi-level supply chain bidirectional view and cascading filtering provide rich data support, improving the accuracy and effectiveness of decision-making. The accelerated shelf life decay model, optimal routing objective function, and convolutional neural network prediction model provide scientific optimization and prediction methods for enterprises' supply chain management. By incorporating temperature and humidity data from logistics vehicles, the remaining shelf life of goods during transportation and storage can be dynamically calculated. Compared to traditional fixed shelf life calculation methods, this more accurately reflects the quality changes of goods in real-world environments. The objective function aims to minimize total loss and generates an optimal supply routing solution for enterprises. By optimizing routing, losses during transportation can be reduced, lowering logistics costs. The dynamic safety stock level adjustment formula comprehensively considers factors such as average daily demand, lead time, demand fluctuation risk, and shelf life decay, and can dynamically adjust safety stock levels at all levels. Compared to traditional fixed safety stock settings, this more accurately reflects the actual needs and risk profile of enterprises, avoiding inventory overstocking or stockouts.

[0055] In one possible implementation, the fulfillment risk monitoring model inputs include supplier-side data, purchaser-side data, and environmental data. The supplier-side data includes production schedule deviation rate, quality inspection abnormality frequency, and logistics vehicle temperature and humidity; the purchaser-side data includes order change intensity index, store sales acceleration, and promotion plan penetration rate; and the environmental data includes ambient temperature and humidity. Dynamic risks are identified through the following steps: Construct a risk transmission topology network and map the fulfillment nodes to production nodes, logistics nodes, warehousing nodes, and delivery nodes; Calculate the risk contagion intensity coefficient β between nodes based on graph neural network: β = f (historical correlation failure probability, current data deviation amplitude, path dependence) f() is a functional relationship. In one possible implementation, the risk contagion intensity coefficient β between nodes is obtained by weighted averaging the historical correlation failure probability, the current data deviation amplitude, and the path dependency. That is, f() is the weighted average. When the risk value of any node is greater than the risk threshold Q and β·number of downstream nodes is greater than the critical value Y, it is marked as a cascading risk event: Cascading risk events include: risk node type (production / logistics / warehousing / delivery node), risk contagion intensity coefficient β (quantifying the risk diffusion ability between nodes), risk value (risk severity of the current node), number of affected downstream nodes (cascade impact range) and risk type identification (derived from input data, such as production delays, temperature control failure, etc.).

[0056] The impact of this technical solution is that the model inputs include supplier data (production schedule deviation rate, quality inspection anomaly frequency, logistics vehicle temperature and humidity), buyer data (order change intensity index, store sales acceleration, promotion plan penetration rate), and environmental data (ambient temperature and humidity), comprehensively integrating key information from upstream and downstream supply chains and the external environment. This multi-source data integration enables the model to perceive potential risks in the supply chain from multiple dimensions, avoiding the limitations of a single data source and improving the accuracy and comprehensiveness of risk identification.

[0057] By collecting and analyzing various data in real time, the model can promptly detect abnormal changes in the supply chain. For example, real-time monitoring of the temperature and humidity of logistics vehicles can ensure the quality and safety of goods during transportation. Real-time tracking of the order change intensity index can reflect changes in purchasing needs, thereby providing early warning of potential fulfillment risks.

[0058] The fulfillment nodes are mapped to production nodes, logistics nodes, warehousing nodes, and delivery nodes, and a risk transmission topology network is constructed. This visual network structure can intuitively display the relationships between nodes in the supply chain and the risk transmission paths, helping enterprise managers quickly understand the source and transmission direction of risks, and providing strong support for the development of targeted risk response measures.

[0059] When a risk event occurs, the risk node can be quickly located and the source of the risk can be traced through the risk transmission topology network. This helps companies take timely measures to cut off the risk transmission chain and reduce the impact of the risk on the entire supply chain.

[0060] Graph neural networks are used to calculate the inter-node risk contagion intensity coefficient β, which takes into account factors such as the historical probability of association failure, the magnitude of current data deviation, and path dependency. Graph neural networks can fully utilize the topological structure and data characteristics between nodes to more accurately calculate risk contagion intensity, providing a scientific basis for risk identification.

[0061] By setting a risk threshold Q and a critical value Y, a cascading risk event is flagged when the risk value of any node exceeds the risk threshold Q and β·number of downstream nodes exceeds the critical value Y. This precise identification method can promptly detect cascading events that could trigger large-scale risks, enabling enterprises to take effective preventive measures before risks erupt, thereby preventing the spread and escalation of risks.

[0062] The fulfillment risk monitoring model monitors supply chain risks in real time and issues alerts when risks reach warning thresholds. This early warning mechanism allows companies ample time to take countermeasures, such as adjusting production plans, optimizing logistics and distribution, and increasing inventory reserves, thereby mitigating the impact of risks on supply chain operations.

[0063] In one possible implementation, triggering a graded warning and dynamically adjusting operation permissions based on the dynamic risk linkage access risk profile includes: When the performance risk monitoring model identifies a cascading risk event, it extracts the node type, risk value, risk contagion intensity coefficient β, and number of downstream nodes where the event occurs; Determine the actual risk intensity value based on the risk value of the cascading risk event, the risk contagion intensity coefficient β, and the number of downstream nodes; Actual risk intensity value = risk value × log (1 + number of downstream nodes) × β Use association rules to match similar risk scenarios in the access risk file and extract the corresponding historical treatment plans, responsible parties, and comprehensive risk scores; Determine the dynamic risk level based on the actual risk intensity value and the comprehensive risk score; Automatically trigger graded warnings based on dynamic risk levels and push warning information containing risk details and emergency response recommendations to relevant parties; Determine the dynamic risk level based on the actual risk intensity value and the comprehensive risk score, including: Normalize the actual risk intensity value and the comprehensive risk score respectively, and map them to the same numerical range in proportion, such as [0,1]; The normalized actual risk intensity value and the comprehensive risk score are weighted averaged to obtain a dynamic risk score; the actual risk intensity value is given a higher weight; The dynamic risk level is determined based on the dynamic risk score, and multi-level warnings are performed based on the dynamic risk level. For example, the multi-level warnings may include: Level 1 warning (low risk): Send risk alerts to suppliers / purchasers; Level 2 warning (medium risk): freeze some operating permissions and initiate reassessment; Level 3 Alert (High Risk): Account suspended and an alternative supply source assigned.

[0064] Synchronously update the risk credit scores of suppliers and buyers: Set the initial credit score, monthly credit decay ratio, and the weights of dynamic risk and access risk in the deduction calculation; decay the initial credit score at a fixed ratio each month; calculate deductions by multiplying the current cascade risk score and access review comprehensive score by their corresponding weights; subtract the two deductions from the decayed credit score to obtain the final credit score, ensuring that the score is within the preset range; Good credit maintains permissions; average credit limits high-risk operation permissions; poor credit freezes the account and forces a qualification review.

[0065] The effect of the above technical solution is: establishing a mapping relationship library between dynamic risks and access risk files, associating qualification defect types, performance risk characteristics and historical loss data, so that when the performance risk monitoring model identifies dynamic risks, it can quickly match similar risk scenarios in the access risk files through association rules, and accurately extract the corresponding risk levels, historical processing plans and responsible parties, thereby improving the efficiency and accuracy of risk response.

[0066] A multi-level early warning mechanism automatically triggers based on risk levels, taking appropriate measures for each risk level. Level 1 alerts push risk warnings, keeping suppliers and buyers informed of potential risks. Level 2 freezes some operational permissions and initiates a reassessment for further risk review and control. Level 3 suspends accounts and assigns alternative sources of supply. When risks are extremely high, swift action is taken to ensure supply chain stability. This ensures effective risk control while minimizing unnecessary disruption to normal operations.

[0067] By linking historical data from access risk profiles, including historical treatment plans and responsible parties, the system provides rich data support for enterprise decision-making. Based on this historical data and current risk situations, enterprises can develop more scientific and rational risk response strategies, achieving data-driven decision-making. Early warning information is pushed through the system to suppliers, buyers, and platform operators, enabling information sharing and collaborative operations among all parties. This improves collaboration efficiency across the supply chain, mitigates issues caused by information asymmetry, and enhances operational efficiency across the entire supply chain.

[0068] Synchronously updating the risk credit scores of suppliers and buyers provides timely information on their risk status. Changes in risk credit scores can serve as an important basis for companies to assess the qualifications of suppliers and buyers, encouraging them to pay more attention to risk management and improve their risk prevention and control capabilities.

[0069] This triggers a qualification status reassessment process, dynamically adjusting operational permissions and resource allocation priorities on the self-operated platform, forming a closed-loop management mechanism of "risk identification - early warning - disposal - credit linkage." This closed-loop management mechanism ensures that risks are addressed promptly and effectively. By dynamically adjusting risk credit scores and qualification status, it incentivizes suppliers and buyers to continuously improve their risk management capabilities, forming a virtuous cycle.

[0070] The present embodiment provides a supply chain order management system based on big data analysis, which is used to implement the supply chain order management method described in the embodiment of the present application. The system includes: The audit module is used to conduct a hierarchical audit of the qualifications of the purchaser and supplier respectively to obtain a comprehensive audit result; the hierarchical audit includes compliance audit and access risk assessment; Collaborative planning acquisition module, used to integrate multi-source data from buyers and suppliers to generate dynamic demand-supply collaborative plans; The recommendation module is used to automatically match a list of compliant and suitable suppliers based on the purchaser's order and the compliance review results, and generate a list of recommended suppliers; The search and positioning module is used to build a two-way view of the multi-level supply chain, and achieve rapid cross-level product location and inventory synchronization through cascading screening; The dynamic monitoring module is used to monitor supplier performance data, purchaser demand change data and environmental data in real time, and identify dynamic risks that actually occur during the performance process through the performance risk monitoring model; The early warning module is used to trigger graded early warnings and dynamically adjust operation permissions based on the dynamic risk linkage access risk file.

[0071] The working principle and effect of the above technical solution are the same as those in the embodiment of the method of this application, and will not be repeated here.

[0072] This application is explained from the perspectives of purpose of use, effectiveness, progress and novelty, and has complied with the functional enhancement and use requirements emphasized by the Patent Law. The above description and drawings of this application are only preferred embodiments of this application and are not intended to limit this application. Therefore, all structures, devices, features, etc. that are similar or identical to those of this application, that is, all equivalent replacements or modifications made in accordance with the scope of the patent application of this application, should fall within the scope of protection of the patent application of this application.

Claims

1. A supply chain order management method based on big data analysis, characterized in that: The method comprises: Conduct a tiered review of the qualifications of both the purchaser and supplier to obtain comprehensive review results; the tiered review includes compliance review and access risk assessment; Integrate multi-source data from buyers and suppliers to generate dynamic demand-supply collaboration plans; Based on the purchaser's order and compliance review results, the system automatically matches a list of compliant and suitable suppliers and generates a recommended supplier list. Build a two-way view of the multi-level supply chain and achieve rapid cross-level product location and inventory synchronization through cascading screening; Real-time monitoring of supplier performance data, purchaser demand change data, and environmental data, and identification of actual dynamic risks occurring during the performance process through performance risk monitoring models; According to the dynamic risk linkage access risk file, a graded warning is triggered and the operation authority is dynamically adjusted.

2. The supply chain order management method according to claim 1, characterized in that: Conduct qualification review of purchasers and suppliers separately and obtain comprehensive review results, including: Receive qualification documents submitted by purchasers and suppliers respectively, and conduct basic compliance review of the purchaser or supplier's qualifications based on the qualification documents using a pre-set compliance rule library; the qualification documents include business information, credit records, industry certification certificates, and quality inspection reports; Input the compliance data into a pre-trained access risk prediction model; output an access risk determination result; the access risk determination result includes a risk probability score and a network-level loss expectation value; Combine the basic compliance review results and access risk assessment results to generate a comprehensive qualification review conclusion; Implement hierarchical authority control based on the audit conclusions.

3. The supply chain order management method according to claim 2, characterized in that: The compliance data is input into a pre-trained admission risk prediction model; Output access risk assessment results, including: The model is trained based on a historical dataset of failed access cases and an industry blacklist dataset; the historical dataset includes qualification defect types, chain risk events, and loss results; Use graph neural networks to analyze supply chain topology relationships, including OEM relationships, distribution links, and resource sharing relationships; Through multi-agent interactive Monte Carlo simulation, the cross-enterprise cascading risks caused by qualification defects are quantified, and the risk probability and expected loss value in the future cycle are output.

4. The supply chain order management method according to claim 2, characterized in that: The combination of the basic compliance review results and the access risk assessment results; Generate comprehensive qualification review conclusions; including: If any of the core items in the qualification document is not in compliance, the review will fail; if qualified, points will be awarded according to the preset rules; Obtain the auxiliary project score according to the auxiliary project scoring rules; Obtain basic compliance scores based on core and supplementary items; Take a weighted average of the risk probability score and network loss expectation score to obtain a comprehensive risk score; Take a weighted average of the basic compliance score and comprehensive risk score to obtain the final comprehensive score; A comprehensive qualification review conclusion is generated based on the basic compliance score and the final comprehensive score. If the final comprehensive score is less than the scoring threshold, the application fails. If all core items pass and the final comprehensive score is greater than or equal to the scoring threshold, the application passes. The output includes a comprehensive risk score.

5. The supply chain order management method according to claim 1, characterized in that: The integration of multi-source data from purchasers and multi-source data from suppliers to generate a dynamic demand-supply collaboration plan includes: Access the buyer's real-time inventory level, historical order fluctuation trends, in-transit logistics data, and market forecast demand; Synchronize suppliers' real-time capacity utilization, raw material inventory dynamics, production line scheduling data, and alternative supply source status; The streaming computing engine performs real-time fusion processing on multi-source data to identify the temporal and spatial conflicts between sudden demand peaks of buyers and production bottlenecks of suppliers. Based on the conflict location results, the original order is dynamically split into batch delivery sub-plans.

6. The supply chain order management method according to claim 5, characterized in that: The method uses a streaming computing engine to perform real-time fusion processing on multi-source data to identify the temporal and spatial conflicts between the purchaser's sudden demand peak and the supplier's production capacity bottleneck; including: Establish a unified spatiotemporal coordinate mapping model to map the purchaser's demand location, supplier's production capacity location, and logistics hub nodes to a geographic grid coordinate system; Calculate in real time the ratio of the peak demand increment ΔD to the remaining capacity margin ΔPD in each grid: When ΔD / ΔPD>threshold K1, it is marked as a red conflict area; When ΔD / ΔPD∈[threshold K2, threshold K1], it is marked as a yellow warning area; Based on the topological relationship of the conflict area, a supplier-purchaser conflict link map is generated.

7. The supply chain order management method according to claim 1, characterized in that: The multi-level supply chain bidirectional view is constructed to achieve rapid cross-level commodity location and inventory synchronization through cascading screening; including: Build a two-way view of the multi-level supply chain including the buyer level and the supplier level; Filter products by temperature control requirements, shelf life thresholds, and cold chain certification status through cascading filters; Predict regional shelf life decay curves based on convolutional neural networks and dynamically adjust safety stock levels at all levels; The circulation path weight is corrected according to the real-time temperature and humidity data to generate the optimal topological path for shelf life loss.

8. The supply chain order management method according to claim 1, characterized in that: The performance risk monitoring model inputs include supplier-side data, purchaser data, and environmental data. The supplier-side data includes production schedule deviation rate, quality inspection abnormality frequency, and logistics vehicle temperature and humidity; the purchaser data includes order change intensity index, store sales acceleration, and promotion plan penetration rate; and the environmental data includes ambient temperature and humidity. Dynamic risks are identified through the following steps: Map fulfillment nodes to production, logistics, warehousing, and delivery nodes; Calculate the risk contagion intensity coefficient β between nodes based on graph neural network: β = f (historical correlation failure probability, current data deviation amplitude, path dependence) When the risk value of any node is greater than the risk threshold and β·number of downstream nodes is greater than the critical value Y, it is marked as a cascade risk event.

9. The supply chain order management method according to claim 1, characterized in that: The triggering of graded warnings and dynamic adjustment of operation permissions based on the dynamic risk linkage access risk file includes: When the performance risk monitoring model identifies a cascading risk event, it extracts the node type, risk value, risk contagion intensity coefficient, and number of downstream nodes where the event occurred; Determine the actual risk intensity value based on the risk value of the cascading risk event, the risk contagion intensity coefficient, and the number of downstream nodes; Use association rules to match similar risk scenarios in the access risk file and extract the corresponding historical treatment plans, responsible parties, and comprehensive risk scores; Determine the dynamic risk level based on the actual risk intensity value and the comprehensive risk score; Automatically trigger graded warnings based on dynamic risk levels, and push warning information containing risk details and emergency suggestions to relevant parties.

10. A supply chain order management system based on big data analysis, used to implement the supply chain order management method according to claim 1, characterized in that: The system comprises: The audit module is used to conduct a hierarchical audit of the qualifications of the purchaser and supplier respectively to obtain a comprehensive audit result; the hierarchical audit includes compliance audit and access risk assessment; Collaborative planning acquisition module, used to integrate multi-source data from buyers and suppliers to generate dynamic demand-supply collaborative plans; The recommendation module is used to automatically match a list of compliant and suitable suppliers based on the purchaser's order and the compliance review results, and generate a list of recommended suppliers; The search and positioning module is used to build a two-way view of the multi-level supply chain, and achieve rapid cross-level product location and inventory synchronization through cascading screening; The dynamic monitoring module is used to monitor supplier performance data, purchaser demand change data and environmental data in real time, and identify dynamic risks that actually occur during the performance process through the performance risk monitoring model; The early warning module is used to trigger graded early warnings and dynamically adjust operation permissions based on the dynamic risk linkage access risk file.

Images