Enterprise supply chain collaborative management method and system

Abstract

The invention relates to the technical field of supply chain management, in particular to an enterprise supply chain collaborative management method and system, and the method comprises the steps: collecting supply chain cross-subject multi-source data, generating a standardized data set through lightweight preprocessing, and building a data intercommunication channel through multi-system interface adaptation; quantizing contribution values and credit rating of participants by combining federal learning with a differential privacy technology to form decision support data; constructing a dynamic benefit and risk adaptation decision model to generate a collaborative execution instruction, and outputting a cross-subject closed-loop execution result; a whole-process responsibility tracing system is constructed through combination of block chain evidence storage and data watermark evidence storage, and a data credibility and responsibility definition result is output; and according to full-link feedback data, forming an iterative optimization closed loop through index diagnosis, scheme generation and landing verification, inputting optimization data into a collaborative management link, and outputting a collaborative management strategy adaptive to a supply chain scene. According to the scheme, the collaborative efficiency and safety of the supply chain are improved through integrated design.

Description

Technical Field

[0001] This invention relates to the field of supply chain management technology, and in particular to a method and system for collaborative management of enterprise supply chains. Background Technology

[0002] In the process of digital transformation of the supply chain, cross-entity collaboration has become the key to enhancing supply chain resilience. However, the existing supply chain management model still has many technical bottlenecks, making it difficult to meet the modern supply chain's demand for real-time, secure, and precise collaboration.

[0003] First, there are shortcomings in cross-entity data interoperability and standardization. Supplier capacity data, factory order data, logistics provider tracking data, and inventory data in the supply chain are often scattered across independent systems of various entities, and the data formats lack a unified standard: date records use "MM / DD / YYYY" and "YYYY-MM-DD" interchangeably, and capacity units are expressed in different ways, such as "pieces / day" and "boxes / day"; moreover, the interface protocols of multiple systems are incompatible, and it is difficult for non-standard systems of small and medium-sized entities to access the core collaboration platform. Data needs to be manually exported and re-entered, which not only leads to low data integration efficiency (processing time of more than 2 hours for a single step) but also makes it easy for data deviations to occur due to human error, failing to provide reliable support for subsequent decision-making.

[0004] Secondly, it is difficult to balance the security and integrity of decision support. Existing decision-making relies heavily on centralized platforms to aggregate data, requiring each entity to upload complete business data (such as supplier costs and factory production plans), which poses a risk of commercial data leakage. Some entities conceal key data (such as backup capacity and emergency logistics resources) to protect their privacy, resulting in decisions based on only partial information. Order allocation often ignores the real-time fulfillment rate of suppliers, and logistics scheduling does not consider route redundancy, which can easily lead to resource mismatch and response delays.

[0005] Finally, the traceability and emergency response mechanisms are inadequate. Instruction records and performance documents during supply chain execution are mostly stored locally. When order delays or logistics damage occur, manual cross-entity data verification is required, with responsibility determination taking 3-5 days. Emergency resources (backup suppliers, temporary routes) rely on manual screening, resulting in response times exceeding 24 hours, making it difficult to cope with sudden risks and impacting supply chain stability. Summary of the Invention

[0006] This invention improves the collaborative efficiency and security of the supply chain through integrated design.

[0007] The technical solution proposed in this invention is: a method for collaborative management of enterprise supply chains, the method comprising: Collect multi-source data across the supply chain, generate standardized datasets through lightweight preprocessing, and build data interoperability channels through multi-system interface adaptation. Based on standardized datasets transmitted through data interoperability channels, federated learning combined with differential privacy technology is used to quantify the contribution value and credit rating of participants. Through rule engine and unsupervised learning, needs and risks are perceived to form decision support data. Based on decision support data, a dynamic benefit and risk-adaptive decision model is constructed to generate collaborative execution instructions. Lightweight smart contracts are used to synchronize instructions and match emergency resources, and output cross-entity closed-loop execution results. Based on the key node data in the closed-loop execution results, a full-process responsibility traceability system is constructed by combining blockchain notarization with data watermark notarization, and outputs data credibility and responsibility definition results. By forming full-chain feedback data through standardized datasets, data exchange channels, decision support data, closed-loop execution results, and data credibility and responsibility definition results, and through indicator diagnosis, solution generation and implementation verification, an iterative optimization closed loop is formed. This optimizes the data input collaborative management process and outputs collaborative management strategies adapted to supply chain scenarios.

[0008] Preferably, the specific process for obtaining the standardized dataset is as follows: Collect capacity data from supplier systems, order data from factory systems, logistics tracking data from logistics providers, and inventory data from inventory management systems; use a lightweight ETL engine to process the collected raw data, removing duplicate data records; fill missing data fields, using the mean of data in the same dimension to fill numeric fields, and leaving text fields unfilled; unify data format and numerical units to generate a standardized dataset.

[0009] Preferably, the specific process for obtaining the data interoperability channel is as follows: Conduct system link planning, identify the cross-entity system clusters of the supply chain that need to be interconnected, mark the basic information and data flow direction of each system, identify the open ports and supported communication protocols of each system, and form a system interface link list; configure interface adapters based on the interface link list, using pre-made standard templates for mainstream systems and custom drag-and-drop components for non-standard systems; match the configured adapters in the no-code configuration center, enter the system connection parameters and verify the connection validity, configure real-time data flow rules and scheduled synchronization strategies; stress test the established channels, confirm the stability of data transmission, solidify the operation and maintenance process, and complete the construction of the data interoperability channel.

[0010] Preferably, the specific process for obtaining the decision support data is as follows: From the standardized dataset transmitted through the data exchange channel, capacity fulfillment data, order response data, and logistics and distribution data of the participants are extracted. Federated learning technology is used, where each participant trains a model locally on the extracted data, uploading only the trained model parameters to the collaborative node. Differential privacy technology is used to add noise to the uploaded parameters. Based on the processed model parameters, the contribution value and credit rating of each participant are quantified. A rule engine matches preset demand identification rules, and unsupervised learning is used to cluster abnormal data to perceive supply chain demand and risks. The quantified contribution value, credit rating, and perceived demand and risk information are integrated to form decision support data.

[0011] Preferably, the specific process for obtaining the cross-entity closed-loop execution result is as follows: Based on the contribution value and credit rating in the decision support data, a dynamic benefit and risk-adaptive decision model is constructed by configuring contribution value weight and credit rating weight. The perceived supply chain demand and risk input model is used to generate collaborative execution instructions. With the help of lightweight smart contracts, the execution instructions are synchronized to the systems of each participating party and the execution verification rules in the contract are triggered. Emergency resources are automatically matched by combining real-time supply chain resource status data. The execution progress of each participating party's instructions is tracked, and the execution results are recorded after the completion of the execution, and integrated to form a cross-entity closed-loop execution result.

[0012] Preferably, the specific process for obtaining the data credibility and responsibility delineation results is as follows: From the cross-entity closed-loop execution results, key node data is extracted; robust data watermarking technology is used to embed corresponding participant identifiers and timestamps into the key node data; the watermarked key node data is uploaded to the consortium blockchain, and a data hash value is generated through the blockchain's hash algorithm to complete the data on-chain notarization; when data disputes or anomalies occur, the watermark information in the notarized data is extracted to verify the consistency of participant identifiers and timestamps, and the integrity of the notarized data is compared with the original data through blockchain hash verification; based on the verification results, the chain of responsibility is restored, the responsible parties and types of responsibility are determined, and a result of data credibility and responsibility definition is formed.

[0013] Preferably, the specific process for obtaining the collaborative management strategy is as follows: Collect end-to-end data to form end-to-end feedback data; identify high-priority optimization indicators through indicator diagnosis; generate targeted optimization plans for optimization indicators and implement them according to the pilot-verification-promotion process; input the system optimization data corresponding to the optimization plan into the collaborative management link, combined with supply chain scenario tags; build a collaborative management model, configure strategy generation rules for different scenarios, and output collaborative management strategies adapted to supply chain scenarios.

[0014] The present invention also provides an enterprise supply chain collaborative management system, the system being used to execute the aforementioned enterprise supply chain collaborative management method.

[0015] The present invention also provides a computer-readable storage medium storing a computer program that is executed by a processor to implement the aforementioned enterprise supply chain collaborative management method.

[0016] The beneficial effects of this invention are: 1. This solution uses a lightweight ETL engine to automatically preprocess multi-source data across entities (supplier capacity, factory orders, logistics trajectories, etc.), unifies the data format (date specification is "yyyy-MM-ddHH:mm:ss", capacity unit is unified as "pieces / day"), and builds a data interoperability channel by combining multi-system interface adaptation (prefabricated templates for mainstream systems and custom components for non-standard systems). No manual data entry is required, and the data processing time for a single step is reduced from more than 2 hours to less than 10 minutes. This not only solves the problems of chaotic data format and incompatible interfaces, but also avoids human error.

[0017] 2. This solution employs federated learning and differential privacy technology. Each entity only needs to train its data model locally and upload parameters (without uploading complete business data). Noise is added to the parameters to ensure that commercial privacy (such as supplier costs and factory production plans) is not leaked. At the same time, it integrates quantified participant contribution values ​​(capacity, response efficiency) and credit ratings (fulfillment rate, data credibility), and combines demand and risk perception to generate decision-making basis. This solves the contradiction between "privacy leakage" and "data bias" in existing centralized decision-making, improves order allocation accuracy by more than 30%, accelerates logistics scheduling response speed by 50%, and avoids resource misallocation and delays.

[0018] 3. This solution uses blockchain notarization (data from key nodes is uploaded to the blockchain to generate hash values) + data watermarking (embedded with participant identifiers and timestamps) to achieve full traceability of execution instructions, performance certificates, and progress feedback. In case of disputes, the responsible party can be identified within 10 minutes through watermark verification and hash verification, shortening the responsibility determination cycle from 3-5 days to within 1 hour. At the same time, by using lightweight smart contracts to automatically match emergency resources (backup suppliers, logistics routes), the emergency response time is reduced from more than 24 hours to within 2 hours, effectively responding to sudden risks and significantly improving supply chain stability. Attached Figure Description

[0019] Figure 1 A flowchart of a method for collaborative management of enterprise supply chains; Figure 2 This is a flowchart illustrating the management process of a collaborative supply chain management method for enterprises. Detailed Implementation

[0020] The following description is intended to disclose the present invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art. The basic principles of the invention defined in the following description can be applied to other embodiments, modifications, improvements, equivalents, and other technical solutions that do not depart from the spirit and scope of the invention.

[0021] It is understood that the term "a" should be understood as "at least one" or "one or more," that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple, and the term "a" should not be understood as a limitation on the number.

[0022] like Figure 1 and Figure 2 As shown, lightweight interconnection of multiple systems is achieved through interface adapters and a no-code configuration center, solving the problems of long development cycles, high adaptation costs, and poor compatibility in traditional system interconnection. This quickly establishes data channels, laying a low-cost and highly compatible foundation for subsequent data sharing. Federated learning combined with differential privacy technology enables cross-entity trusted data processing, addressing issues of trade secret leakage and difficulty in quantifying contributions during cross-entity data sharing. A lightweight ETL engine ensures data quality, strengthening the foundation of collaborative trust. A combination of a rule engine and unsupervised learning enables demand and risk perception, addressing the problems of delayed dynamic supply chain information capture and the inability to identify new risks. Real-time acquisition of demand and risk information provides accurate basis for decision-making. Dynamic benefit-risk matching decisions incorporating contribution values ​​and credit rating weights address unfair distribution of benefits and risks and conflicting execution, generating fair and implementable collaborative instructions. Lightweight smart contracts automatically match instructions and emergency resources to achieve closed-loop execution across stakeholders. Execution instructions are automatically synchronized to all participating systems and progress is tracked in real time. When a sudden risk is detected, backup resources are selected based on credit rating and historical performance records, solving the problems of difficult implementation of collaborative instructions and slow response to sudden risks, ensuring traceable execution and guaranteed emergency response. Blockchain-based evidence storage combined with data watermarking is used for key node evidence storage and accountability. Key data such as profit distribution results, risk-bearing commitments, raw material quality inspection reports, and logistics receipt records are uploaded to the blockchain to ensure immutability. At the same time, unique watermarks are added to the data to identify the source and ownership, solving the problems of easily tampered data and difficulty in defining responsibility, strengthening execution constraints, and efficiently handling disputes. Through iterative optimization driven by end-to-end feedback data, the solution's adaptability and inability to dynamically adjust are addressed, continuously optimizing rules at each stage to build a sustainable collaborative ecosystem.

[0023] Furthermore, the specific process of achieving lightweight interconnection of multiple systems through interface adapters and no-code configuration centers is as follows: Obtaining the multi-system data interoperability channel and interface adaptation configuration list, with data synchronization delay Seconds and interface adaptation success rate Using this as the core benchmark, the entire process of "system link planning - adapter configuration - parameter quantification and verification - operation and maintenance solidification" is implemented. The specific process is as follows: First, system link planning is carried out. A visual questionnaire is used to identify the system clusters that the enterprise needs to interconnect (such as the entire link of "CRM→ERP→MES→WMS"). The brand of each system (SAPS / 4HANA, Yonyou U9Cloud, etc.), deployment environment (local IP / cloud address) and core data flow requirements (such as "CRM orders need to be synchronized to ERP in real time to generate production plans") are marked. At the same time, the intranet port scanning tool is launched to automatically identify the open ports of each system (HTTP / 80, MQTT / 1883, JDBC / 3306, etc.) and supported protocols. Combined with the built-in library of more than a thousand mainstream system interface documents (including interface paths, request methods, and historical response times), a "System Interface Link List" is generated to plan the complete data flow link of "CRM order interface → ERP production plan interface → MES work order interface → WMS inventory interface", and to determine the physical and logical paths for channel construction.

[0024] The interface adapter cluster is configured based on the link list: For mainstream systems, pre-built standard templates are used (such as the SAPODataAPI template to adapt to ERP order data and the Siemens Opcenter template to adapt to MES production data). The templates have built-in protocol parsing and field mapping logic. For non-standard systems (such as self-developed logistics tracking systems), the request header (such as the Authorization field), parameter mapping ("logistics number" → "Logistics_No"), and return value parsing rules (JSON field extraction) are configured by dragging and dropping components in the "custom adapter". The system verifies the compliance of the configuration in real time and generates an "Adapter Configuration Report" to ensure that the basic capabilities of interface adaptation of each system meet the standards.

[0025] Enter the no-code configuration center to establish the connection: Select each system in sequence and match the corresponding adapter, fill in the connection parameters (enter the API key for the cloud system, enter the database account and password for the local system, and store sensitive information with AES-256 encryption), click "Test Connection" to initiate 3 verification requests for each link, record the "connection successful / failed" status and response time, and immediately output troubleshooting suggestions (such as "check firewall port open" and "confirm API key validity") if "IP unreachable" or "insufficient interface permissions" occur, until all links have 3 successful requests, and initially establish multi-system interface connections; then configure cross-system data flow rules (such as "automatically synchronize to ERP when CRM order status is 'confirmed'"), set trigger conditions, field mapping ("Customer Name" → "Customer_Name"), synchronization method (real-time trigger), initiate a full-link flow test of 100 simulated data entries, and count the number of successful flows. Total number of tests Through formula Calculate (where To improve the initial interface adaptation success rate, For a successful transfer, (Total number of tests) needs to reach (like In this test Success on the first attempt. If the requirements are not met, the field mapping or synchronization strategy is adjusted until the requirements are met. At this point, the multi-system data interoperability channel is initially formed.

[0026] Configure data synchronization strategies to ensure channel performance: For core data such as orders and inventory, an "event-triggered + real-time push" mechanism is adopted. When data changes in the source system, a synchronization command is triggered via WebHook, transmitted via HTTP / 2 protocol, and a timestamp is used to record the "data transmission time" during the synchronization process. "and the target system's confirmation time" ", through formula Calculate single delay (in For the time the data was sent, (The time it takes for the target system to receive confirmation) is used to repeatedly test 50 times (covering peak and off-peak periods) and take the average value to ensure... Seconds; set a timed synchronization cycle (accurate to minutes) for non-core data, enable incremental synchronization (synchronize only changed data) and data filtering (synchronize only valid status data); configure synchronization failure retry (3 times, 5-second interval), and store failed data in the "abnormal data pool" for retransmission to ensure the integrity of channel data.

[0027] Finally, the channel was solidified through stress testing and operations and maintenance: a 24-hour full-link stress test was conducted before launch, simulating the flow of 1000 data entries per minute in real business data, and the number of successfully processed interfaces was counted every hour. Total number of interfaces Through formula Statistics (of which) To improve the success rate of interface adaptation, To ensure successful flow of interfaces, (Total number of interfaces), the results of 24 iterations must satisfy 100% of the " " After going live, the monitoring panel displays the channel status, interface connection status, and data throughput in real time, automatically records synchronization logs (including timestamps, data volume, and status), and pushes alarms and provides repair suggestions in real time if parameters exceed the limits, ensuring the continuous stability of the multi-system data communication channel and providing a transmission link for subsequent data processing.

[0028] Throughout the entire process of building the multi-system data interoperability channel, the interface adaptation configuration list is automatically generated and continuously improved with each step of the operation: During the system access phase, the configuration center automatically records the connection parameters of each system (such as API key storage path, database connection string), adapter template selection (such as "SAPODataAPIV3.2"), and field mapping rules (such as "CRM order number → ERPOrder_No"), forming the initial version of the "Interface Basic Configuration Table"; In the data synchronization strategy configuration phase, synchronization trigger conditions (such as "order status = confirmed"), synchronization cycle (such as "once every 30 minutes"), retry mechanism ("3 times, 5-second interval") and other rules are automatically added to the table; Finally, during the permission and security configuration, information such as the interface access whitelist (such as "192.168.1.100") and role permissions ("configuration personnel can only modify field mappings") are also included. Based on the configuration operations throughout the entire process, the configuration center automatically integrates the information from all stages and generates a standardized "Interface Adaptation Configuration List" through the "List Export" function. This list covers all aspects, including system name, interface path, protocol type, connection parameters, field mapping, synchronization strategy, and security configuration. All configuration items are related to the data synchronization delay of the multi-system data interoperability channel. Interface adaptation success rate "The parameters are strongly correlated and can be directly used for continuous operation and iterative optimization of the channel, ensuring that the channel always meets the real-time and integrity requirements of subsequent cross-subject trusted data processing."

[0029] Furthermore, the specific process of cross-entity trusted data processing through federated learning combined with differential privacy technology is as follows: Cross-entity trusted data processing is carried out by combining federated learning with differential privacy technology, and the bias is calculated based on the contribution value. Credit rating accuracy "Using this as the core standard, it undertakes the real-time and complete data transmission through multi-system data interoperability channels (data sources include raw data such as capacity, production progress, and logistics tracking from suppliers, factories, logistics providers, and other cross-entity systems). The specific process and parameter acquisition methods are as follows:" Based on the established multi-system data interoperability channel, the original data from the systems of all participating parties is accessed. (The data source is the cross-entity data link already specified in the interface adaptation configuration list.) First, data preprocessing is performed through a lightweight ETL engine: duplicate records are removed, missing numerical fields are filled with the mean of data in the same dimension, missing text fields are filled with "none", and the data format (date field is "yyyy-MM-ddHH:mm:ss") and numerical units (capacity is unified as "pieces / day", transportation capacity is unified as "tons / trip", and duration is unified as "hours"), outputting standardized data. Synchronously calculate data integrity metrics (Non-collected parameters), through the formula Derivation (where for The number of valid data entries for (Total number of original data entries), requirements This provides high-quality input for federated learning.

[0030] A horizontal federated learning architecture is adopted to ensure data privacy and security, with each participating party based on its local... Train the initial model to obtain local model parameters ( Model weight matrix With bias vector The set, specifically the process of obtaining it: initializing model parameters. (Randomly generate a matrix that conforms to a normal distribution) (Initial value set to 0); Input by batch Data (batch size) The model predicts values ​​through forward propagation. ,in The activation function is (the Sigmoid function is selected). for Feature data; loss value calculated using cross-entropy loss function. ( for (The true labels in the image); parameters are updated via backpropagation (gradient descent): , ,in The learning rate is used; after 20 iterations, the model weights for the current iteration are output. With bias ,Right now ), raw data The entire process is not transmitted to external parties.

[0031] Each participant uses the Paillier homomorphic encryption algorithm to... Encryption (encryption process: ...) and Each element is converted to a large integer based on the Paillier algorithm public key. Calculate ciphertext ,in For plaintext parameter values, After generating a random number, it is uploaded to the federated learning coordinator; the coordinator is based on... - Differential privacy mechanism adds Laplace noise (Non-collected parameters), through the formula Generate (where for The parameter sensitivity, through Value range derivation, preset ; For privacy budget, preset ; (Using standard Laplace distributed random numbers, generated by a random number generator), resulting in noisy encryption parameters. .

[0032] Coordinator Decryption (based on Paillier algorithm private key) Calculate plaintext After that, the global model parameters are generated by weighting and aggregating the data according to their proportion of data volume. (Non-collected parameters), through the formula Calculate (where The number of participants is derived from statistics obtained when multiple entities access the system. For the first Square weight, , For the i-th party Data volume For all participants The total data volume was obtained through statistical aggregation of data from all participating parties. Each participant updates its local model, repeating the above process of "local training - encrypted upload - noise addition - global aggregation - parameter update". Repeat until the model loss value Output the trained federated learning model .

[0033] Contribution value quantification is carried out concurrently during federated learning training. An indicator system is constructed from three dimensions: "data quality, resource input, and response efficiency." The scores for each dimension are non-collected parameters, and are derived and calculated based on raw data: Data Quality Score ( For data integrity; To ensure data real-time performance, a formula is used. calculate, for The number of records where the difference between the data generation time and the transmission completion time is ≤10 seconds (obtained by timestamp difference statistics); resource investment score. ( The data represents the actual production capacity / transport capacity of the participating parties, sourced from... ; The average resource investment of all participants, through (Calculated); Response efficiency score ( The order processing time for participating parties, sourced from ; The minimum response time for all participants. The maximum value is obtained by passing through. (Originated from sorting and statistical analysis of medium-duration data).

[0034] Calculate the three-dimensional weights using the entropy weight method. , , (Non-collected parameters), first use the formula Calculate information entropy ( , For the first Fang Di The original dimensional scores (sourced from the statistical analysis of dimensional scores from all participating parties) are then processed using a formula. Derivation of weights (satisfying) ), ultimately through the formula Calculate the overall contribution value of each participant (Value range: 0-100 points). Reference scores are obtained through an automated method that uses publicly available industry benchmark libraries and historical performance verification. (Access to authoritative public databases in the supply chain industry, extracting benchmark data on the contribution value of similar companies over the past 3 years) Match corresponding benchmark intervals based on enterprise size and business type; link historical performance data of each participant over the past 12 months. Calculate the equivalent value of historical contributions Among them, the on-time delivery rate and order completion quality score both improved from [previous data]. Obtained from; via formula (Generate final reference score) using the formula Calculate contribution value and calculate deviation ,Require If the target is not met, the indicator weights or score calculation logic will be readjusted.

[0035] Credit ratings are generated by combining historical performance data, and historical performance data from multiple entities over the past 12 months are collected. (Sources include archived data such as order delivery records, quality inspection reports, and dispute resolution results from cross-entity systems), key indicator extracted: number of defaults. (Source: (Statistical results of breaches of contract such as failure to deliver on time and substandard quality) and deviation in performance time (Non-collected parameters, obtained through formulas) calculate, For the first The actual fulfillment time of this order The agreed-upon fulfillment period for orders is all sourced from... (order records), historical complaint rate (Non-collected parameters, obtained through formulas) calculate, for The number of complaints in for The total number of orders is obtained through historical statistics.

[0036] Using logistic regression model Training, Input , , and current contribution value Output A / B / C three-tier credit rating (Category A: No record of default) Hour, Category B: Second-rate, Hour, Category C: Those that do not conform to categories A and B; Use the test set. (Source: 20% of fulfillment sample data that did not participate in model training) The accuracy of the model was verified by randomly selecting samples from the data, using the formula... Calculating the accuracy of credit rating (in To test the consistency between the centralized rating results and actual performance. (The total number of samples in the test set is obtained through test set validation statistics). Requirements: If the target is not met, new indicators such as timely payment collection rate will be added and the model will be retrained.

[0037] The final output consists of three core achievements: a trusted data resource pool. (integration) and The analysis results support real-time querying and a list of contributor values. (Records of all participating parties) value, (Details of three-dimensional scores and weights), credit rating table (Note the participating parties) class, , , (and other key indicators), all results data are synchronized with the data delay. Interface adaptation success rate "Strong correlation ensures data timeliness and integrity."

[0038] Furthermore, the specific process of achieving demand and risk perception by combining rule engines with unsupervised learning is shown below: Trusted data resource pool based on preceding output (Including standardized data) Federated learning model Analysis results), participant contribution value (Source: Quantitative calculation results of prior contribution value), Credit Rating (Source: Preceding credit rating model) Output results and historical performance data (Source: archived data from the past 12 months across multiple systems) First, construct a dual-dimensional data input set encompassing both demand and risk. Integration Real-time business data (order volume) Capacity utilization rate Logistics turnover efficiency (Source: Real-time data transmission from multiple system interconnection channels) Quantitative data of contribution value in , Credit rating related indicators (number of defaults) Deviation in contract performance time Historical complaint rate (Source: Extraction results of previous credit rating indicators) and Historical demand fluctuation patterns Risk event records Normalization was performed using Min-Max (formula). ,in These are the original indicator values. , This represents the historical value range of the indicator, sourced from... (Statistical results) Map all indicators to the [0,1] interval and output standardized input data. This provides high-quality data support with a unified dimension for demand identification and risk perception.

[0039] Launching a rules engine to build a system for identifying explicit requirements and known risks: based on supply chain industry business specifications (sourced from national supply chain management standards and industry association guidance documents), compliance requirements, and Based on historical experience data, two types of rule bases are constructed: a demand identification rule base. With known risk rule base All rule parameters are derived from Compared with industry standards: Demand Identification Rules (Example): 1. When the real-time order volume (Source: (Real-time business data) and the average of the past 7 days ratio At that time, it was determined to be a "surge in demand", among which ( This represents the daily order volume for the past 7 days, sourced from... (Statistics on historical data of the past 7 days); 2. When the core product capacity utilization rate (Source: (Real-time business data) And order backlog (Source: (Statistical data on order status) At that time, it was determined to be "demand gap between production capacity and demand"; Known risk rules (Example): 1. Credit rating Order percentage of different types of participants When this is triggered, a "performance risk warning" is issued, in which... (Molecular origin is) middle Statistics on related orders from participating parties, the denominator source is 1. Total order volume statistics); 2. Logistics turnaround time (Source: (China's real-time logistics data) compared to historical averages deviation At that time, a "logistics delay risk warning" is triggered, in which... This is the average logistics turnover time over the past 90 days, sourced from... (China Logistics Data Statistics).

[0040] Real-time reading of rules engine Data is processed according to the logic of "matching line by line - condition judgment - result output": [The following is a list of parameters and their corresponding values] Any rule in the above is marked as an explicit requirement. ,satisfy Any of the rules will trigger a warning of known risks. Synchronously record rule matching details.

[0041] Simultaneous deployment of unsupervised learning models to uncover latent needs and potential risks: using the Isolation Forest algorithm to construct the model. ,by For training data, the core training process and parameter sources are as follows: 1. Initialize model parameters: number of decision trees (Based on best practices in similar industry scenarios), number of samples per tree (according to (Data volume is dynamically adjusted); 2. For (Source: Standardized input data) Perform random sampling to generate multiple subsets; 3. Based on random features (source: 4. Construct multiple isolated trees using all dimensions of metrics and random thresholds (randomly generated based on the feature value range), and calculate the path length L for each sample (the number of isolated nodes from the root of the tree, derived from node traversal statistics during model training); 5. Normalize the path length to obtain the anomaly score. Through formula Calculation, where This is the expected value of the path length for all samples (derived from the mean path length of all samples collected during model training). The value range is [0,1]. The closer to 1, the higher the degree of sample abnormality.

[0042] Model After training is completed, input in real time Data calculation of outlier scores :Will And not by The matched samples were determined to be implicit requirements. (Such as potential regional demand growth, changes in product mix demand, sourced from model pairings) (Results of mining abnormal data patterns in the data); And not by The matched samples were identified as potential risks. (Such as hidden bottlenecks in the supply chain and decreased stability of cooperation among participants, the source of which is the model's...) Risk correlation analysis of abnormal data in China.

[0043] To verify the accuracy of perception, a two-dimensional evaluation system was constructed. The core parameters and their sources are as follows: Demand identification accuracy : Through formula Calculation, where (Total number of recognition requests, sourced from statistics of rule engine and model output results). The data should be based on real demand verified through business scenarios (sourced from order fulfillment feedback and market demand survey data statistics over the next 30 days). ; Risk warning recall rate : Through formula Calculation, where The total number of risk events that actually occurred within a 30-day period (source: (Updated data and real-time business risk record statistics) The number of risk events that actually occurred and are jointly warned by the model and rule engine (source: warning records and...) (match statistics), requirements .

[0044] If the requirements are not met, optimizations and adjustments will be made: Regarding requirement identification, the following additions will be made: User behavior preferences (sourced from end-customer order details data) and market trend correlation data (sourced from publicly available industry market analysis reports) are iterated upon. Rule threshold and Abnormal score thresholds; regarding risk perception, supplementary data includes the financial health of participating parties (sourced from publicly available data from third-party corporate credit platforms) and industry policy change data (sourced from documents issued by government regulatory departments), and updates. Risk correlation indicators and The training sample set (after including new data) (until the core compliance requirements are met).

[0045] The final output is an integrated perception of "demands and risks": a demand list. (integration) , Label the demand type, triggering metric, and confidence level. Risk Warning List (integration) , (Including risk level, impact range, and early warning basis), all results are synchronized and delayed with previous data. Interface adaptation success rate "Credit rating accuracy" "Strong correlation ensures the real-time nature and reliability of the perception results."

[0046] Furthermore, the specific process of dynamic benefit-risk matching decision-making by incorporating contribution value and credit rating weights is as follows: Based on the requirements list output in the preceding steps (Including explicit requirements) Latent needs and confidence level , source is the output result of the requirement and risk perception stage), risk warning list (including known risks , potential risks and scope of influence, source is the same as above), contribution value C of participants (source is the quantification result of the cross-subject trusted data processing stage), credit rating G (source is the output of the previous credit rating model output), first construct a decision input data set : Integrate the requirement type, urgency (mapped according to the confidence level , is "extremely high", is "high", is "medium", source is confidence level statistics), requirement scale (such as order volume ratio, production capacity gap, source is real-time business data), the risk level in (classified as "major", "relatively large", "general" according to the scope of influence, source is the evaluation result of the risk perception stage), risk occurrence probability (calculated based on the historical occurrence frequency of similar risks in , ), the contribution value in , the credit rating correlation indicators in , , ), after all indicators are standardized, output a decision-specific data set .

[0047] Construct a two-dimensional weight system for contribution value and credit rating, and all weight parameters are obtained through quantitative calculation: Contribution value weight : Calculate based on the distribution characteristics of the contribution value of participants , and deduce through the formula (where is the contribution value of a single participant, source is ; , are the minimum and maximum values of the contribution values of all participants, source is statistical results), the value range is [0.4, 1.0], and the higher the contribution value, the greater the weight; Credit rating weight : According to the credit rating Mapping quantization weights, the rules are as follows: Class Time , Class Time , Class Time (The mapping rules are derived from industry credit risk assessment standards and...) Correlation analysis of performance success rates among different rating participants in China); Comprehensive decision weight : Using the weighted summation formula Calculation (0.5 is the two-dimensional equilibrium coefficient, derived from the optimal ratio in industry decision-making practices). The value range is [0.35, 0.95], which is used to measure the priority of the participants in the distribution of benefits and risk-bearing.

[0048] Build a dynamic benefit-risk adaptation decision model The core decision-making logic is derived from quantitative indicators: Benefit distribution decision: Prioritizing the satisfaction of needs Based on the core evidence, (in For comprehensive decision-making weighting, The urgency level of the demand is quantified as "extremely high = 3, high = 2, medium = 1". (Standardized value of demand size), according to Sort by high to low, prioritizing... High-performing participants receive benefits such as allocation of core orders and resource allocation; a benefit distribution coefficient is also calculated. ( (Probability of the risk faced by the participant) The higher the level, the larger the share of benefits allocated (such as profit sharing ratio, cooperation priority); Risk-taking decision: based on risk-taking fit coefficient calculate, (in The risk level is quantified as "Serious = 3, Significant = 2, Moderate = 1". The higher the threshold, the heavier the responsibility for risk prevention and control (such as the proportion of risk reserve contributions and the leading role in emergency response plans); at the same time, a risk circuit breaker threshold is set. (Source: (Statistics on the critical value of medium-risk out-of-control), when At the same time, the participant is restricted from undertaking high-risk demands.

[0049] Decision model validation and dynamic adjustment: Decision-making effectiveness evaluation: Verify the model's effectiveness through core achievement parameters and assess decision execution satisfaction. ( (Data source: participant decision-making feedback survey) Risk control efficiency ( (Data source: Business risk records statistics within 30 days after the decision was implemented). Dynamic optimization: If or Then adjust the overall decision-making weight coefficient (e.g., when the distribution of benefits is biased). Weighting increased to 0.6), risk circuit breaker threshold (according to The mean is dynamically adjusted, and newly added real-time demand change data is incorporated (source: Updated results), risk evolution data (source: (Update results) Retrain the model Continue until the requirements are met.

[0050] The final output is a dynamic benefit-risk adaptation decision-making scheme. Includes: a list of the distribution of benefits among the participating parties (marked) , (and specific allocation details), a list of risk-bearing responsibilities (marked) (This includes) prevention and control responsibilities and circuit breaker mechanisms, and decision adjustment logs (recording the parameter optimization process and basis). All decision results are consistent with the accuracy of previous requirement identification. Risk warning recall rate "Contribution value calculation deviation" "Strong correlation."

[0051] Furthermore, the specific process of achieving cross-entity closed-loop execution through lightweight smart contract synchronization instructions and automatic matching of emergency resources is shown below: Output-based decision-making scheme (Including a list of profit distribution and a list of risk-bearing responsibilities), and a list of needs. (Including the type of need and its urgency) Demand Scale Risk Warning List (including risk level) (Scope of impact), combined with resource status data transmitted in real time through multi-system interconnection channels. (For example, real-time production capacity, logistics capacity, and inventory levels of participating parties, sourced from real-time data reported by cross-entity systems), first construct the smart contract input dataset. Integration Interest distribution coefficient Risk tolerance fit coefficient Circuit breaker threshold , In , , Probability of risk occurrence ,according to Resource types in Resource availability Resource response radius (Logistics resource reachability range, sourced from logistics provider system GPS positioning and regional division data), after standardization processing (unified data exchange format is JSON), outputting contract-recognizable structured data. .

[0052] Deploying lightweight smart contracts Developed using the Solidity language (adapted to cross-entity consortium blockchain architecture, reducing node computing power requirements), the core modules and parameter sources are clearly defined as follows: Command synchronization module: Presets command generation rules, based on and Automatically generate execution instructions (Instr) such as order allocation instructions and risk control instructions. Instruction parameters include the execution subject. (Source is a unique identifier across systems), Execution Content (e.g., "allocate orders for 100 core products" or "initiate emergency logistics allocation"), Completion Deadline (according to Mapping: Extremely High = 4 hours, High = 8 hours, Medium = 24 hours, source: (Emergency level definition); the contract broadcasts instructions across the consortium blockchain, employing an encrypted signature mechanism (based on private key signing by participants, public key verification, and the key source being an asymmetric key pair generated during cross-entity system access) to ensure the security of instruction transmission and synchronously record the instruction reception status. (“Received” or “Not Received”), the party that did not receive the message will trigger a second push (1 minute interval).

[0053] Resource matching rules module: Constructs a matching algorithm based on "demand-resource" adaptation logic. The core parameters all come from 1. Requirement-Resource Type Matching: When the requirement type (e.g., "capacity replenishment", "express logistics") and resource type 1. If they match, an initial match is triggered; 2. Priority sorting: scores are given based on matching priority. calculate( The weighting for comprehensive decision-making is derived from... ; Standardized value for resource availability; (For resource response radius, the smaller the value, the stronger the adaptability); 3. Risk constraint verification: if the participating parties This excludes high-risk resource matching, ensuring the security of resource allocation.

[0054] Initiating the automatic matching and execution process for emergency resources: Key parameters and execution logic: Resource pool dynamically updated: Contracts synchronized in real time The data is updated every 30 seconds to show the resource availability status. ("Available", "Occupied", "Pending Release"), through formulas Calculate the amount of resources available in real time ( The allocated resources (sourced from contract instruction execution records) ensure the real-time nature of resource data.

[0055] Matching Execution and Feedback: Contract by Sort by high to low, automatically send resource call instructions to the participant with the best matching resources, and synchronously generate execution credentials. (Including instruction ID, resource information, and timestamp, sourced from contract execution logs); After execution, the participating parties upload the execution results via the system interface. After verifying the validity of the contract results ("success", "failure", "partial completion") and supporting data (such as logistics receipt records, production capacity achievement certificates, sourced from the participating party's business system), the instruction execution status is updated. .

[0056] Execution effect verification and closed-loop optimization: Key performance indicator calculation: Instruction execution response time (Non-collected parameters) = Command sending time -Execution result feedback time ( , (Source: Contract timestamp record) Required Minutes; Resource matching accuracy (Non-collected parameters) are obtained through formulas Calculation (numerator from demand fulfillment verification records, denominator from contract resource matching log statistics), requirements: .

[0057] Emergency adjustment mechanism: If Within minutes, the contract automatically triggers a "resource expedited allocation" command, prioritizing resources within the response radius. Local resources of kilometers; if Then optimize the matching algorithm. Weighting coefficients (such as boosting) The weight was increased to 0.6), and a new resource type tag library was added (resources corresponding to special needs were added, sourced from industry resource classification standards), and the matching model was retrained.

[0058] The final result is a cross-entity closed-loop execution report: Smart Contract Execution Report. (Including instruction synchronization records, resource matching details, and Indicators), Emergency Resource Utilization Ledger (Records data throughout the entire process of resource allocation, usage, and feedback), cross-entity collaborative execution logs. All results were correlated with the satisfaction level of previous decision implementation. "Effective risk control" The strong correlation between "data synchronization delay t" enables a closed loop throughout the entire process of "decision-instruction-execution-feedback," ensuring the efficiency and stability of cross-entity collaboration in the supply chain.

[0059] Furthermore, the specific process of using blockchain-based evidence storage combined with data watermarking to conduct evidence storage and accountability for key nodes is as follows: Clearly define the key nodes and data scope requiring evidence preservation, and trace the source of all evidence-preserving data to each stage of the entire process: 1. Data transmission nodes: Interface adaptation configuration list for multi-system interoperability channels. Data synchronization logs (including) , Indicators); 2. Data processing nodes: Trusted data resource pool List of Contribution Values Credit rating table 3. Decision-making and execution nodes: Requirement list Risk Warning List Dynamic decision-making schemes 4. Closed-loop execution nodes: smart contract execution instruction Instr, resource matching record Execution results Integrate to form a complete set of evidence storage data After format standardization (unified as hash value + original data digest), the output is... .

[0060] The core parameters and operation process for deploying a dual-security mechanism of "data watermark embedding + blockchain notarization" are as follows: Data watermark generation and embedding: A robust text watermarking algorithm is adopted, based on cross-subject unique identifiers. (Source: Unique code assigned during cross-entity system access) and key node timestamps (Source: Data generation time records for each stage) Generate a unique watermark, with the watermark format being " - - Random checksum (random checksum is obtained through a formula) calculate, (Using the SHA-256 hash function). Watermarking is embedded for different data types: semantic watermarking is used for text-based data (such as decision plans and implementation reports) (without affecting data readability); redundant bits in fields are embedded for structured data (such as lists and ledgers). The embedding process is implemented using a lightweight algorithm to ensure that watermark embedding is time-efficient. Second( (This refers to the time difference between watermark generation and embedding completion).

[0061] Blockchain-based evidence storage: A network of evidence storage nodes is built based on a consortium blockchain architecture, with various cross-entity entities participating in consensus as consortium blockchain nodes (the consensus mechanism adopts the PBFT algorithm, and the number of nodes is consistent with the number of participants, sourced from cross-entity access statistics). The evidence storage data is then embedded with a watermark. Double encryption is performed: first, encryption is performed using the AES-256 algorithm (key). The source is a key pool jointly maintained by the consortium blockchain nodes, and then the hash value of the encrypted data is calculated. (Generated using the SHA-256 algorithm). Watermark information, node signature (Source: Node private key signature result), timestamp (Source: System time at the time of on-chain) Packaged into a notarization block, verified through consortium blockchain consensus, and written into the distributed ledger, generating a unique notarization ID. (Source: Block hash value).

[0062] A full-process accountability traceability system is constructed, with the traceability logic and core parameters as follows: Traceability Triggering and Watermark Extraction: The traceability process is triggered when data disputes arise (such as denial of execution results or shirking of risk responsibility) or data anomalies occur (such as data tampering or missing data). This is based on the stored evidence ID. Retrieve the corresponding evidence data from the blockchain ledger, restore the watermark using a watermark extraction algorithm, and analyze the data. , and The verification formula is: ,like This confirms that the data source is authentic.

[0063] Chain of responsibility restoration: combining blockchain ledger data , , Information such as these, expressed through formulas Trace the entire path of data flow ( (For the chain of responsibility). Synchronously verify the hash value of the evidence storage data: recalculate the hash value of the current data. , and on-chain storage In comparison, if If the data has not been tampered with, then it is determined that the data has not been altered (i.e., ...). If they are inconsistent, the tampering node can be located through the watermark information. Combined with timestamps Identify the responsible party and the time of the violation.

[0064] Traceability effect verification and optimization: Calculation of core indicator: Accuracy of accountability (Non-collected parameters) are obtained through formulas Calculate (numerator from the statistical matching of traceability results with the actual responsible parties, denominator from the statistical analysis of traceability process trigger records), requiring... ; Optimization mechanism: If Then, the watermarking algorithm will be upgraded (by adding data feature factor embedding) and the blockchain evidence storage fields will be expanded (by adding operator information). (Source: Operation logs of each node) to strengthen the granularity of the responsibility chain; if a tampering situation of mismatched hash value occurs, the consortium blockchain alarm will be automatically triggered, and the operation permissions of the violating node will be frozen simultaneously to ensure the authority of evidence storage and traceability.

[0065] The final output is the key node evidence storage and traceability results: a blockchain evidence storage master ledger. (Record all) Watermark (and corresponding data relationships), accountability report (Including tracing the triggering cause, responsible party, and handling result), full-process data flow diagram All results are related to the response time of the preceding instructions. Resource matching accuracy "Strong correlation enables a trustworthy closed loop throughout the entire lifecycle of 'data generation - decision execution - accountability traceability,' providing technical assurance for the compliance and traceability of cross-entity collaboration."

[0066] Furthermore, the specific process of iterative optimization driven by end-to-end feedback data is as follows: First, construct a full-link feedback data pool. This dataset integrates all raw data, historical trends, abnormal fluctuation records, and feedback suggestion labels for all indicators. Through data standardization (mapping indicator values ​​to the [0,1] interval) and correlation analysis, it outputs an optimization analysis dataset. This provides data support for determining the direction of iteration.

[0067] A three-tiered iterative optimization system of "indicator diagnosis - solution generation - implementation verification" is established, with the core parameters and process as follows: End-to-end indicator diagnosis: based on Calculate the weight of indicators in each stage (Non-collected parameters), derived using the entropy weight method: ( For the first Information entropy of each indicator The total number of indicators, sourced from (Statistical results of key indicators); the higher the weight, the greater the impact of the indicator on the entire process. The difference between the current value and the historical best value of the indicator is calculated simultaneously. ( For the first The historical best value of each indicator The current value is from [source]. (Historical records of indicators), by Sort and filter out the top 3 high-priority optimization metrics (such as...) , , ), and clarify the core direction of iteration.

[0068] Targeted optimization solution generation: Differentiated optimization strategies are developed for high-priority indicators, and all solution parameters are derived from... In practice: like Low (Insufficient resource matching accuracy): Optimize the smart contract matching algorithm Adjust the weighting coefficient to ; like Low accuracy (insufficient traceability): Upgrade the data watermarking algorithm and embed operator identification. (Source: Operation logs of each node) Expand the blockchain evidence storage fields to "hash value + watermark + operation trajectory" to strengthen the granularity of the responsibility chain; like Too high (contribution value calculation deviation is too large): Add "historical cooperation satisfaction" indicator. (Source: (Evaluation record of cross-entity cooperation), optimized contribution value calculation formula is as follows: Recalculate the entropy weight method weights - .

[0069] Implementation and Verification of Results: The optimized solution will be implemented according to the principle of "pilot first, then rollout"—selecting 30% of cross-entity entities as pilot projects (source: contribution value of participating parties). With credit rating (Uniformly distributed samples), deploy the optimized algorithm model, smart contract, or rule engine, and record the implementation cycle of the solution. (The time difference between the finalization of the plan and the completion of the pilot deployment is sourced from the project management log.) Optimized indicator values ​​were collected in real-time during the pilot phase. Calculate the improvement rate of a single indicator Through formula Calculate the overall improvement rate of indicators throughout the entire process ( (The sum of the weighted improvement rates of all indicators).

[0070] Iterative closed loop and continuous optimization: Solution Promotion and Consolidation: If and The optimization plan will be extended to all cross-entity entities, and the full-process operation manual and system configuration parameters (such as smart contract code and algorithm model weights) will be updated and synchronously written into the blockchain for evidence storage (the source is the optimized configuration file and execution rules). Dynamic adjustment mechanism: If the target is not met, adjustments will be made based on pilot feedback data. (Source: Pilot participant execution logs and optimization suggestions) Readjust the scheme parameters (such as weight coefficients and threshold settings), shorten the implementation cycle (optimize the deployment process and adopt modular updates), and re-enter the "pilot-verification" cycle until the core compliance requirements are met.

[0071] Final output: Summary of end-to-end iterative optimization results (Including problem diagnosis report, specific optimization measures, parameter adjustment details), indicator improvement comparison table (Record the values ​​of each indicator before and after optimization, and the improvement rate), full-process standardized update manual All results are strongly correlated with the core indicators of the preceding entire process, forming an iterative closed loop of "data feedback - solution optimization - implementation verification - evidence preservation", continuously improving the efficiency, accuracy and stability of cross-entity collaboration.

[0072] The processes described above with reference to the flowcharts in the embodiments disclosed in this invention can be implemented as computer software programs. The embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), it performs the functions defined in the methods of this application. It should be noted that the computer-readable medium described above in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wire segments, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless segments, wire segments, optical fibers, RF, etc., or any suitable combination thereof.

[0073] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0074] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been shown and explained in the embodiments. Without departing from the stated principles, the implementation of the present invention may have any variations or modifications.

Claims

1. A method for collaborative management of enterprise supply chains, characterized in that, The method includes: Collect multi-source data across the supply chain, generate standardized datasets through lightweight preprocessing, and build data interoperability channels through multi-system interface adaptation. Based on standardized datasets transmitted through data interoperability channels, federated learning combined with differential privacy technology is used to quantify the contribution value and credit rating of participants. Through rule engine and unsupervised learning, needs and risks are perceived to form decision support data. Based on decision support data, a dynamic benefit and risk-adaptive decision model is constructed to generate collaborative execution instructions. Lightweight smart contracts are used to synchronize instructions and match emergency resources, and output cross-entity closed-loop execution results. Based on the key node data in the closed-loop execution results, a full-process responsibility traceability system is constructed by combining blockchain notarization with data watermark notarization, and outputs data credibility and responsibility definition results. By forming full-chain feedback data through standardized datasets, data exchange channels, decision support data, closed-loop execution results, and data credibility and responsibility definition results, and through indicator diagnosis, solution generation and implementation verification, an iterative optimization closed loop is formed. This optimizes the data input collaborative management process and outputs collaborative management strategies adapted to supply chain scenarios.

2. The enterprise supply chain collaborative management method according to claim 1, characterized in that, The specific process for obtaining the standardized dataset is as follows: Collect capacity data from supplier systems, order data from factory systems, logistics tracking data from logistics providers, and inventory data from inventory management systems; use a lightweight ETL engine to process the collected raw data, removing duplicate data records; fill missing data fields, using the mean of data in the same dimension to fill numeric fields, and leaving text fields unfilled; unify data format and numerical units to generate a standardized dataset.

3. The enterprise supply chain collaborative management method according to claim 2, characterized in that, The specific process of obtaining the data interoperability channel is as follows: Conduct system link planning, identify the supply chain cross-entity system clusters that need to be interconnected, mark the basic information and data flow direction of each system, identify the open ports and supported communication protocols of each system, and form a system interface link list; configure interface adapters based on the interface link list, using pre-made standard templates for mainstream systems and custom drag-and-drop components for non-standard systems. In the no-code configuration center, match the configured adapter, enter the system connection parameters and verify the connection validity, configure the real-time data flow rules and timed synchronization strategies; perform stress testing on the built channel, confirm the stability of data transmission, solidify the operation and maintenance process, and complete the construction of the data interoperability channel.

4. The enterprise supply chain collaborative management method according to claim 3, characterized in that, The specific process for obtaining the decision support data is as follows: From the standardized dataset transmitted through the data exchange channel, the capacity fulfillment data, order response data, and logistics and distribution data of the participants are extracted; federated learning technology is adopted, and each participant trains the model on the extracted data locally, and only uploads the trained model parameters to the collaborative node, and adds noise processing to the uploaded parameters in combination with differential privacy technology; Based on the processed model parameters, the contribution value and credit rating of the participants are quantified. By matching preset demand identification rules with a rules engine and combining unsupervised learning to perform cluster analysis on abnormal data, supply chain demand and risks are perceived; quantitative contribution values, credit ratings and perceived demand and risk information are integrated to form decision support data.

5. The enterprise supply chain collaborative management method according to claim 4, characterized in that, The specific process for obtaining the cross-entity closed-loop execution result is as follows: Based on the contribution value and credit rating in the decision support data, a dynamic benefit and risk-adaptive decision model is constructed by configuring contribution value weight and credit rating weight. The perceived supply chain demand and risk input model is used to generate collaborative execution instructions. With the help of lightweight smart contracts, the execution instructions are synchronized to the systems of each participating party and the execution verification rules in the contract are triggered. Emergency resources are automatically matched by combining real-time supply chain resource status data. The execution progress of each participating party's instructions is tracked, and the execution results are recorded after the completion of the execution, and integrated to form a cross-entity closed-loop execution result.

6. The enterprise supply chain collaborative management method according to claim 5, characterized in that, The specific process for obtaining the data credibility and responsibility delineation results is as follows: From the cross-entity closed-loop execution results, key node data is extracted; robust data watermarking technology is used to embed corresponding participant identifiers and timestamps into the key node data; the watermarked key node data is uploaded to the consortium blockchain, and a data hash value is generated through the blockchain's hash algorithm to complete the data on-chain notarization; when data disputes or anomalies occur, the watermark information in the notarized data is extracted to verify the consistency of participant identifiers and timestamps, and the integrity of the notarized data is compared with the original data through blockchain hash verification; based on the verification results, the chain of responsibility is restored, the responsible parties and types of responsibility are determined, and a result of data credibility and responsibility definition is formed.

7. The enterprise supply chain collaborative management method according to claim 6, characterized in that, The specific process for obtaining the collaborative management strategy is as follows: Collect end-to-end data to form end-to-end feedback data; identify high-priority optimization indicators through indicator diagnosis; generate targeted optimization plans for optimization indicators and implement them according to the pilot-verification-promotion process; input the system optimization data corresponding to the optimization plan into the collaborative management link, combined with supply chain scenario tags; build a collaborative management model, configure strategy generation rules for different scenarios, and output collaborative management strategies adapted to supply chain scenarios.

8. A collaborative management system for enterprise supply chain, characterized in that, The system is used to execute an enterprise supply chain collaborative management method as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that is executed by a processor to implement an enterprise supply chain collaborative management method as described in any one of claims 1-7.

Images