A supply chain contract core index penetrating risk management method and device

Abstract

This application provides a method and apparatus for penetrating risk management of core indicators in supply chain contracts, belonging to the field of supply chain contract risk management technology. The method provided includes: collecting internal and external data related to the supply chain; processing the internal and external data to establish multi-dimensional data associations, forming a unified risk data lake; parsing key clauses of the contract text, mapping the parsed contract clauses to monitored data indicators, calculating the data indicators based on the associated data in the risk data lake, and obtaining a standardized risk indicator dataset; based on the risk indicator dataset, conducting risk assessment through a hybrid intelligent mode of rule engine and prediction model, identifying risk levels and locating risk sources, and generating structured early warning events; initiating a preset collaborative handling process based on the structured early warning events, recording the risk handling results, and feeding back the risk handling results to the data collection, indicator calculation, and risk assessment stages.

Description

Technical Field

[0001] This application relates to the field of supply chain contract risk management technology, and in particular to a method and apparatus for penetrating risk management of core indicators of supply chain contracts. Background Technology

[0002] Against the backdrop of deepening globalization, supply chain networks are constantly expanding and becoming increasingly complex, with frequent unforeseen events highlighting their vulnerability. Simultaneously, supply chain contract risks exhibit significant dynamism and concealment. Suppliers' financial situations, performance capabilities, and the political and economic environment of their regions are constantly changing, with numerous risks hidden at the second, third, and even Nth tier supplier levels, making them difficult to access using traditional management methods. Furthermore, internal contract texts, performance data, financial data, and supplier information are scattered across different departments and systems, creating information silos and hindering the construction of a unified risk view. These real-world challenges underscore the increasing necessity and urgency of supply chain contract risk management.

[0003] Currently, the management solutions for supply chain contract risks in the market and within enterprises are relatively fragmented and have not yet formed a systematic framework: some enterprises rely on the contract and performance management functions of traditional ERP / SRM modules, recording transaction data and storing contract texts through the ERP system, and using the SRM system to conduct supplier onboarding and performance evaluation, and calculate key indicators such as on-time delivery rate; other enterprises use independent third-party risk intelligence tools to obtain external data such as the financial health and regional risks of suppliers and conduct risk scoring; still other enterprises only use the electronic contract management system of the legal department to realize the digital processes of contract drafting, approval and archiving, or use tools such as Tableau and Power BI to create business intelligence dashboards to display data such as supplier delivery performance.

[0004] However, existing management solutions have many significant shortcomings: traditional ERP / SRM modules suffer from static, superficial, outdated, and isolated issues; contract terms cannot be linked to business data in real time, making it difficult to penetrate to N-tier suppliers, and lacking external data integration to achieve real-time early warning; third-party risk intelligence tools are disconnected from enterprise business processes, making it difficult to accurately match their own supply chain networks, leaving enterprises to passively receive information and lacking customized analysis capabilities; electronic contract management systems have fallen into an "archive room" mode, neglecting the contract execution stage, and key terms cannot be quantified into monitoring indicators; business intelligence dashboards can only describe what has happened, lacking predictive and traceability analysis capabilities, and reliance on manual judgment easily leads to inefficiency and risk omissions. These problems result in traditional risk management being mostly a passive, post-event accountability model, making it difficult to achieve early risk identification and precise control.

[0005] Therefore, there is an urgent need for a method to break down information barriers, achieve dynamic monitoring, penetrating traceability, intelligent early warning and decision support, so as to promote the transformation of risk management from post-event remediation to pre-event early warning and in-event intervention, and enhance the resilience and core competitiveness of the supply chain. Summary of the Invention

[0006] In view of this, this application provides a method and apparatus for penetrating risk management of core indicators of supply chain contracts, which can break down information barriers, realize dynamic monitoring, penetrating traceability, intelligent early warning and decision support, so as to promote the transformation of risk management from post-event remediation to pre-event early warning and in-event intervention, and enhance the resilience and core competitiveness of the supply chain.

[0007] Specifically, this application is implemented through the following technical solution:

[0008] The first aspect of this application provides a method for penetrating risk management of core indicators of supply chain contracts, the method comprising:

[0009] Collect internal and external data related to the supply chain, clean and standardize the internal and external data, establish multi-dimensional data associations, and form a unified risk data lake.

[0010] The contract text is parsed for key clauses, and the parsed contract clauses are mapped to monitored data indicators. Based on the associated data in the risk data lake, the data indicators are calculated according to preset calculation rules to obtain a standardized risk indicator dataset.

[0011] Based on the aforementioned risk indicator dataset, risk assessment is conducted through a hybrid intelligent mode combining rule engines and prediction models to identify risk levels, locate risk sources, and generate structured early warning events.

[0012] Based on the structured early warning event, a preset collaborative handling process is initiated, the risk handling results are recorded, and the risk handling results are fed back to the data collection, indicator calculation, and risk assessment stages.

[0013] The second aspect of this application provides a supply chain contract core indicator penetration risk management device, the device including a data acquisition module, a calculation module, an identification module and a processing module;

[0014] The acquisition module is used to collect internal and external data related to the supply chain, clean and standardize the internal and external data, establish multi-dimensional data associations, and form a unified risk data lake.

[0015] The calculation module is used to parse key clauses of the contract text, map the parsed contract clauses into monitored data indicators, and calculate the data indicators based on the associated data in the risk data lake through preset calculation rules to obtain a standardized risk indicator dataset.

[0016] The identification module is used to assess risks based on the risk indicator dataset through a hybrid intelligent mode of rule engine and prediction model, identify risk levels and locate risk sources, and generate structured early warning events.

[0017] The processing module is used to initiate a preset collaborative handling process based on the structured early warning event, record the risk handling results, and feed the risk handling results back to the data collection, indicator calculation and risk assessment stages.

[0018] The supply chain contract core indicator penetration-based risk management method and device provided in this application firstly collects data from multiple systems within the supply chain and diverse external data. After cleaning and standardization, it establishes a multi-dimensional data association centered on "contracts / orders," penetrating upstream supplier networks and horizontally expanding to the external environment, forming a unified risk data lake. This breaks down the information silos caused by data fragmentation in traditional management, providing a complete and interconnected data foundation for subsequent analysis. Simultaneously, it uses natural language processing technology to parse key contract clauses and map them into monitorable data indicators. Through a stream processing engine, it calculates indicator values ​​in real time based on preset calculation rules, constructing a standardized risk indicator dataset. This achieves a dynamic mapping of "contract-indicator-data," transforming static contract clauses into "living" indicators active on the front lines of business. This completely changes the traditional contract "archive room" model, enabling real-time perception of contract performance status.

[0019] Building upon this foundation, a hybrid intelligent model combining a "rule engine" and a "predictive model" is employed for risk assessment. The rule engine handles explicit, hard-line rules and regulations, while the predictive model learns from historical data to predict the probability of potential risks. Simultaneously, it integrates a supply chain knowledge graph to achieve penetrating risk tracing across multiple levels of suppliers, accurately locating risk sources and transmission paths. This achieves a precise diagnosis from "symptoms" to "root causes," addressing the limitations of traditional management, which can only cover first-tier suppliers and lacks risk tracing capabilities. Furthermore, the generated structured early warning events not only contain core information such as risk level and scope of impact but also simulate the cost and delivery impact of different response strategies based on data such as inventory and alternative supplier capacity. This implements a dual-engine approach of "risk warning + decision support," providing managers with quantitative decision-making support rather than simply issuing risk alerts.

[0020] Finally, by initiating a pre-set collaborative handling process and feeding back the handling results to each front-end stage, a closed loop of continuous optimization is formed. This enables the system to continuously improve the accuracy of data correlation, the rationality of indicator calculation, and the precision of risk assessment. In the long run, this method can effectively reduce the risk of supply chain disruption, optimize inventory levels, and improve negotiation efficiency. It transforms risk management from passive post-event remediation to proactive pre-event warning and in-event intervention, realizing the transformation of risk management from a "cost center" to a "value creation center." The resulting benefits, such as loss avoidance and efficiency improvement, can be directly or indirectly converted into financial returns, becoming an important part of the company's core competitiveness. Attached Figure Description

[0021] Figure 1 A flowchart of the supply chain contract core indicator penetration risk management method provided in Embodiment 1 of this application;

[0022] Figure 2 This is a schematic diagram of the structure of the supply chain contract core indicator penetration risk management device provided in Embodiment 2 of this application. Detailed Implementation

[0023] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0024] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0025] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0026] The following specific embodiments are given to illustrate the technical solution of this application in detail.

[0027] Figure 1The flowchart illustrates the supply chain contract core indicator penetration risk management method provided in Embodiment 1 of this application. Please refer to... Figure 1 The method provided in this embodiment may include:

[0028] S101. Collect internal and external data related to the supply chain, clean and standardize the internal and external data, establish multi-dimensional data associations, and form a unified risk data lake.

[0029] Optionally, the internal data includes purchase orders, receipts, and payment data from the ERP system; supplier master data and performance scores from the SRM system; customer order forecast data from the CRM system; transportation status data from the TMS system; and original contract data from the legal CLM system. The external data includes financial credit and public opinion monitoring data from third-party data providers; weather, port congestion index, and commodity price data from public data sources; and external environmental data collected by IoT sensors.

[0030] Specifically, internal data refers to the collection of raw data generated and stored by various business systems within an enterprise throughout the entire supply chain operation and contract management process, directly related to contract performance and risk control. This is core business data that the enterprise can independently control. External data refers to various supplementary data from outside the enterprise related to supply chain contract performance risks. It is used to perceive the potential impact of changes in the external environment on the supply chain and is an important extension and supplement to internal data. Multi-dimensional data association refers to establishing a three-dimensional relationship between internal data and between internal and external data, using "contracts / orders" as the core link and combining supply chain business logic and risk transmission paths, thus breaking down data fragmentation. A risk data lake refers to a unified, centralized, structured data storage and management platform formed after data collection, cleaning, standardization, and multi-dimensional association; it is the data foundation of the entire risk management system.

[0031] In practice, multi-dimensional data associations are established, including: using contracts or orders as the core association key, assigning a unique identifier to each contract and its corresponding purchase order, and using the identifier to associate basic information and execution data to form a panoramic view of contract performance; constructing a supply chain network map, and using the supply chain network map to associate the capacity, delivery performance, and geographical location risk data of upstream multi-level suppliers; and using the supplier's registration location, production location, and logistics route as key dimensions to horizontally associate contracts or orders with external environmental data.

[0032] Optionally, the basic information includes supplier master data, procurement categories, product specifications, and pricing terms; the execution data includes order fulfillment status, delivery records, logistics trajectory, warehousing information, quality inspection reports, invoices, and payment status; and the external environment data includes macroeconomic data, weather data, maritime news, public opinion data, and geopolitical risk index.

[0033] Specifically, a unique identifier is assigned to each contract and its corresponding purchase order, and this identifier is embedded in the contract text, purchase order, and related business data records as a core linking credential. Based on this unique identifier, basic information such as supplier master data, purchase categories, product specifications, and pricing terms are linked, along with execution data such as order fulfillment status, delivery records, logistics tracks, warehousing information, quality inspection reports, invoices, and payment status, integrating them to form a complete panoramic view of contract fulfillment. Information such as the hierarchical relationships, cooperation agreements, and supply material associations of upstream multi-level suppliers is collected to construct a supply chain database that includes entities and relationships such as multi-level suppliers, materials, geographical locations, and logistics routes. Supply chain network map; Based on the supply chain network map, through entity association matching, it retrieves relevant data such as capacity data, historical delivery performance records, production base geographical locations and corresponding risk indices of upstream second-tier, third-tier and even N-tier suppliers to establish risk transmission path associations; It extracts supplier registration location, production location information and logistics route information corresponding to contracts / orders, and uses these geographical dimensions as association indexes; Through association indexes, it horizontally matches contracts / orders with external environmental data such as macroeconomic data, weather data, maritime news, public opinion data, geopolitical risk indices, etc., to achieve the association and binding of internal business data and external environmental data.

[0034] In practice, data is extracted in real-time or periodically (e.g., every 15 minutes) from the ERP system (purchase orders, receipts, and payments), the SRM system (supplier master data and performance scores), the CRM system (customer order forecasts), the TMS system (transportation status data), and the legal CLM system (original contract data) via API interfaces such as RESTful APIs. For legacy systems, the aforementioned internal data is directly extracted from the database using ETL tools such as Kettle. Financial credit and public opinion monitoring data are obtained through APIs from third-party data providers; weather, port congestion index, and commodity price data from public data sources are crawled via web scraping; and data such as temperature and humidity in supplier warehouses collected by IoT sensors are obtained through data subscription services or sensor interfaces. Furthermore, redundant, erroneous, and missing data in both internal and external data are removed, and inconsistent data formats are corrected to ensure data accuracy. The cleaned internal and external data undergo a unified format conversion, standardizing field names, data units, and coding rules to ensure consistency and comparability across different sources. Establish multi-dimensional data associations, centrally store all data after cleaning, standardization and multi-dimensional association, and build a unified risk data lake to provide a clean and related data source for subsequent indicator calculations.

[0035] S102. Perform key clause parsing on the contract text, map the parsed contract clauses to monitored data indicators, and calculate the data indicators based on the associated data in the risk data lake through preset calculation rules to obtain a standardized risk indicator dataset.

[0036] Optionally, the data indicators include delivery indicators, quality risk indicators, financial risk indicators, compliance and ESG indicators, and supply chain resilience indicators; the delivery indicators include on-time delivery rate and delivery delay days; the quality risk indicators include batch pass rate, customer complaint rate, and quality cost ratio; the financial risk indicators include price volatility, payment term deviation, and supplier financial health; the compliance and ESG indicators include contract fulfillment rate, environmental compliance inspection pass rate, and labor standard audit score; and the supply chain resilience indicators include dependence on a single supplier, order concentration, and average recovery time.

[0037] Specifically, key clauses refer to the core content in the contract text that is directly related to supply chain performance, risk management, and the definition of rights and obligations, and can be quantified into monitorable indicators. These include delivery-related clauses (clearly specifying delivery time, delivery cycle, delivery location, delivery quantity, and delivery method, such as "a penalty will be paid for a delay exceeding 5 working days"), quality-related clauses (agreeing on product quality standards, acceptance standards, quality pass rate requirements, handling of quality issues, and compensation), price and financial-related clauses (including pricing rules, price adjustment mechanisms, payment cycles, payment conditions, and methods for calculating penalties), compliance and responsibility-related clauses (involving contractual performance obligations, environmental compliance requirements, labor standards agreements, compliance inspection obligations, and the definition of liability for breach of contract), and supply chain resilience-related clauses (such as supplier alternatives, restrictions on reliance on a single supplier, order allocation ratios, and force majeure responses, all related to supply chain stability).

[0038] Furthermore, data metrics refer to quantifiable and real-time monitorable structured parameters generated after analyzing key contract terms using natural language processing technology. These metrics serve as a concrete measure of contract performance and supply chain risk levels. Each data metric clearly corresponds to a key contract term, possessing a clear definition, calculation dimension, and threshold standards. It can be dynamically calculated and updated in real-time based on related data in a risk data lake, and can be mapped to standardized scores or risk levels. Specifically, data metrics can be categorized as follows: delivery metrics (used to monitor delivery performance, including on-time delivery rate and delivery delay days); quality risk metrics (used to assess product quality levels and quality risks, including batch pass rate, customer complaint rate, and quality cost ratio); financial risk metrics (used to assess financial risks, including price volatility, payment term deviation, and supplier financial health); compliance and ESG metrics (used to manage compliance and social responsibility risks, including contract term fulfillment rate, environmental compliance inspection pass rate, and labor standard audit score); and supply chain resilience metrics (used to measure the supply chain's ability to withstand risks, including dependence on a single supplier, order concentration, and average recovery time).

[0039] In practice, key clauses of the contract text are parsed, and the parsed contract clauses are mapped to monitored data indicators. This includes: extracting key clauses from the contract text using natural language processing technology, including clauses related to delivery time, quality standards, price agreements, liability for damages, service level agreements, payment cycles, and penalty rules; automatically creating corresponding monitored data indicators for each extracted key clause, determining the indicator definition, calculation dimensions, and threshold standards; specifically, creating on-time delivery rate and delivery delay days indicators based on delivery time clauses, batch pass rate indicators based on quality standard clauses, and frequency of breaches and percentage of penalty amounts based on liability for damages and penalty rules clauses; and establishing mapping relationships between the created data indicators and related data in the risk data lake to determine the data sources required for indicator calculation.

[0040] Specifically, the system collects contract texts (such as structured contracts in Word and PDF formats) and imports them into the designated text processing module. Natural language processing (NLP) technology is then used to perform semantic analysis and extract key clauses from the contract texts, identifying core clauses related to delivery time, quality standards, price agreements, liability for damages, service level agreements (SLAs), payment cycles, and penalty rules. For each extracted key clause, corresponding monitored data indicators are automatically created: based on the delivery time clause, indicators such as "on-time delivery rate" and "delivery delay days" are created; based on the quality standards clause, an indicator such as "batch pass rate" is created; based on the liability for damages and penalty rules clauses, indicators such as "frequency of breach" and "percentage of penalty amount" are created; and for other key clauses such as price agreements, payment cycles, and service level agreements, corresponding data indicators are created separately. According to the indicators; furthermore, for each created data indicator, clarify the indicator definition (e.g., "on-time delivery rate refers to the proportion of order lines received on time to the total number of received order lines"), calculation dimensions (e.g., dividing calculation dimensions by supplier, contract, and time period) and threshold standards (e.g., setting the on-time delivery rate threshold to 95%); determine that the data source for "on-time delivery rate" and "delivery delay days" is the planned delivery date and actual receipt date data from the ERP system; determine that the data source for "batch qualification rate" is the quality inspection report data; determine that the data source for "frequency of default" and "percentage of liquidated damages" is the contract default records, payment data, and liquidated damages calculation-related clause data; match internal data (e.g., ERP, SRM, TMS system data) or external data (e.g., supplier financial data, compliance audit data) in the risk data lake as the calculation data source for other data indicators.

[0041] Optionally, based on the associated data in the risk data lake, the data indicators are calculated using preset calculation rules to obtain a standardized risk indicator dataset. This includes: building an indicator calculation engine; the indicator calculation engine is built based on a stream processing engine and extracts associated datasets from the risk data lake as needed. The associated datasets include order delivery records, quality inspection data, and payment data from internal data, and supplier financial data, compliance audit data, and regional risk index from external data; configuring preset calculation rules; the preset calculation rules include SQL query statements or target formulas, and set corresponding aggregation or statistical operation logic for different types of data indicators; calling the corresponding preset calculation rules through the indicator calculation engine to perform calculations on the extracted associated datasets to obtain the original values ​​of each data indicator; mapping the calculated original indicator values ​​to a standardized scoring system of 0-100 points; and associating and integrating the normalized indicator values ​​with metadata to form a structured dataset.

[0042] Specifically, an indicator calculation engine is built, based on the Apache Flink or Spark stream processing engine. A connection channel with the risk data lake is configured to ensure the ability to extract related data on demand. Related datasets are extracted from the risk data lake, including order delivery records, quality inspection data, and payment data from internal data, and supplier financial data, compliance audit data, and regional risk indices from external data. Pre-defined calculation rules are configured, including SQL queries or dedicated formulas. For different types of data indicators such as delivery, quality, financial, compliance / ESG, and supply chain resilience, corresponding aggregation or statistical calculation logic is set. For example, on-time delivery rate = (number of on-time delivered orders / total number of delivered orders) × 100%, and supplier financial risk score is calculated using Z-factor mapping. The system calculates using preset formulas such as the Score model; it then starts the indicator calculation engine, calls the preset calculation rules corresponding to each data indicator, and processes the extracted associated dataset to obtain the original values ​​of each data indicator; it normalizes the original values ​​of all data indicators, mapping them uniformly to a standardized scoring system of 0-100 points to achieve comparability between different types of indicators; furthermore, it collects metadata such as indicator ID, indicator name, calculation object (e.g., supplier, contract, material), data validity period, and calculation timestamp, and integrates the normalized indicator values ​​with the above metadata to form a structured risk indicator dataset. Each indicator value in this dataset is associated with the original detailed data in the risk data lake through the calculation object, ensuring the traceability of the calculation results.

[0043] S103. Based on the aforementioned risk indicator dataset, risk assessment is performed using a hybrid intelligent mode combining rule engine and prediction model to identify risk levels and locate risk sources, generating structured early warning events.

[0044] Specifically, the rule engine is a core component that uses pre-defined "if-then" hard logic rules to perform real-time verification of risk indicator datasets and trigger early warnings. It uses structured rule configuration based on "conditions + results" to clearly define the trigger thresholds, durations, and corresponding warning levels (such as yellow and red warnings). These rules are formulated based on key contract terms (such as delivery time and quality standards) and business management requirements; for example, "If the supplier's on-time delivery rate is <95% and the duration is >3 days, then a yellow warning is triggered."

[0045] Furthermore, a predictive model is an intelligent predictive component built based on machine learning algorithms. It learns from historical risk data patterns, analyzes real-time input of risk indicator datasets, and predicts the probability of potential risk events occurring within a preset timeframe. Structured early warning events are standardized, directly applicable early warning data sets that integrate the judgment results of rule engines and predictive models. They include core risk information, analytical conclusions, and decision support content, representing the final output of risk assessment results. Specifically, structured early warning events include basic information, scope of impact, analytical results, decision support, and traceability information. The basic information includes risk level (low / medium / high risk, corresponding to green / yellow / red) and risk type (such as delivery risk, quality risk, and financial risk); the scope of impact includes specific business-related objects such as contract numbers, material information, purchase orders, and production plans; the analysis results include conclusions on risk source location (such as insufficient capacity of second-tier suppliers), risk transmission paths, and predicted probabilities (such as an 85% probability of delay in the next 30 days); decision support includes preliminary response suggestions and simulation results of different response strategies (such as the impact of activating backup suppliers on costs and delivery time); and traceability information includes detailed risk indicators and data sources (such as order delivery records and supplier financial data in the risk data lake) to ensure traceability.

[0046] In specific implementation, a rule engine is configured; hard early warning rules in the form of preset targets are established, and these hard early warning rules are formulated based on contract terms and business needs, including triggering conditions and corresponding early warning levels; a prediction model is constructed and trained; the rule engine performs real-time verification of the risk indicator dataset, and when the data indicators meet the preset triggering conditions, it directly generates the corresponding level of early warning; the prediction model receives real-time input of the risk indicator dataset, calculates using built-in mathematical functions, and outputs the probability of occurrence of various risk events within a preset future time period; a scoring card model is used to comprehensively evaluate the results of the rule engine and the output of the prediction model, first mapping individual indicator values ​​to sub-scores, then calculating a comprehensive risk score by weighting according to preset weights, and finally mapping the comprehensive score to... The system categorizes risks into different risk levels; it constructs a supply chain knowledge graph containing entities and relationships; when anomalies in risk indicators are detected, it uses graph computing technology to traverse upstream along the knowledge graph, analyzing the capacity, delivery performance, geographical location risk, and financial status of multi-level suppliers; it combines risk transmission path algorithms to infer the risk diffusion logic, locate the root cause of the risk, and determine the transmission path; it integrates risk level, risk type, scope of impact, risk source analysis results, predicted probability, risk transmission path, and preliminary response suggestions to form an early warning event and push it to a message queue or generate an early warning work order; for the highest risk level early warning event, it simulates the impact of different response strategies on costs and delivery time based on associated data, and supplements the structured early warning event with the simulation results.

[0047] Specifically, configure the rule engine and preset hard warning rules in the form of "if-then": based on contract terms and business needs, clarify the triggering conditions (including indicator thresholds, duration, etc.) and corresponding warning levels for each rule. For example, "If the supplier's on-time delivery rate is <95% and the duration is >3 days, then a yellow warning will be triggered." Build and train the prediction model. The rule engine performs real-time verification on the risk indicator dataset. When the data indicators meet the preset triggering conditions, it directly generates the corresponding level of warning. The prediction model receives real-time input from the risk indicator dataset and calculates the probability (in the range of 0-1) of various risk events occurring within a preset future time period using built-in mathematical functions such as the Sigmoid function and multi-level activation functions. Furthermore, a comprehensive assessment is conducted using a scoring card model: individual indicator values ​​are mapped to corresponding sub-scores (e.g., 10 points for on-time delivery rate ≥ 98%); each sub-score is weighted according to preset weights (e.g., on-time delivery rate accounts for 40% of delivery risk) to obtain a comprehensive risk score; the comprehensive risk score is then mapped to three levels: low risk (green, ≥ 90 points), medium risk (yellow, 70-89 points), and high risk (red, < 70 points). A supply chain knowledge graph is constructed, containing entities and relationships such as multi-level suppliers, materials, geographical locations, and logistics routes; when anomalies in risk indicators are detected, graph computing technology is used to traverse upstream along the knowledge graph, analyzing data such as the capacity, delivery performance, geographical location risk, and financial status of multi-level suppliers, and combining this with risk transmission path algorithms to infer the risk diffusion logic and accurately locate the root cause and transmission path of the risk. The system integrates risk level, risk type (e.g., delivery risk, financial risk), scope of impact (involving contracts, materials, orders, production plans), risk source analysis results, predicted probabilities, risk transmission paths, and preliminary response suggestions to form standardized early warning events, which are then pushed to a message queue or generated as early warning work orders. For high-risk early warning events, based on relevant data such as existing inventory, in-transit orders, alternative supplier capacity, and logistics costs, the system quickly simulates the impact of different response strategies (e.g., activating backup suppliers, adjusting logistics methods) on costs and delivery times, and supplements the structured early warning events with the simulation results. For the specific implementation process of graph computing technology and risk transmission path algorithms, please refer to the descriptions in related technologies; they will not be elaborated here.

[0048] Optionally, a prediction model is constructed and trained, including: extracting snapshots of risk indicator datasets from historical time points as feature variables; labeling each historical data snapshot corresponding to a feature variable; the label is determined based on whether the target risk event actually occurs within a subsequent preset time period, and the label value is 0 or 1; selecting a machine learning classification algorithm as the base model, the machine learning classification algorithm including one or more combinations of gradient boosting decision trees, logistic regression, and neural networks; randomly configuring the initial parameters of the base model, setting the number of iterations, learning rate, and regularization coefficient for model training; dividing the prepared training data into a training set and a validation set according to a preset ratio, and inputting the training set into the initialized base model; the base model predicts the training set data based on the current parameters and outputs the predicted probability of the risk event occurring; calculating the difference between the predicted probability and the true label through a loss function, backpropagating the loss value using a gradient descent algorithm, adjusting the internal parameters of the model, minimizing the loss function, and repeating the process of prediction, loss calculation, and parameter adjustment until the model's performance index on the validation set converges to a preset threshold; using historical data that did not participate in training as a test set, inputting it into the trained model, and verifying the model's generalization ability.

[0049] Specifically, a snapshot of the risk indicator dataset at historical time points is extracted and used as the feature variables for model training. The feature variables include multi-dimensional indicators such as historical on-time delivery rate, number of quality complaints, supplier financial indicators, price volatility, compliance audit score, and external regional risk index. For each historical data snapshot corresponding to a feature variable, a fact label is assigned based on whether the target risk event (including supplier bankruptcy, major delivery delay, batch quality non-conformity, compliance violation, etc.) actually occurs within the subsequent preset time period. The label value is "occurred (1)" or "not occurred (0)". Machine learning classification algorithms are selected as the basic model, including Gradient Boosting Decision Tree (GBDT), Logistic Regression Decision Tree (LRRD), and Logistic Regression Decision Tree (LRRD). The model employs one or more combinations of logical regression and neural networks; it randomly configures the initial parameters of the base model, while setting hyperparameters such as the number of training iterations, learning rate, and regularization coefficient; it divides the prepared training data into a training set and a validation set according to a preset ratio (e.g., 7:3), and inputs the training set data into the initialized base model; the base model predicts the training set data based on the current parameters, outputting the predicted probabilities of various risk events; it calculates the difference between the predicted probabilities and the true labels using a preset loss function (e.g., cross-entropy loss function, mean squared error loss function); and it uses the gradient descent algorithm to backpropagate the loss value and adjust the model's internal parameters to minimize the loss function.

[0050] Repeat the process of prediction, loss calculation, and parameter tuning until the model's performance metrics (such as accuracy, recall, and AUC) on the validation set converge to a preset threshold; select historical data that was not used in training as the test set and input it into the trained model to verify the model's generalization ability.

[0051] S104. Based on the structured early warning event, initiate a preset collaborative handling process, record the risk handling results, and feed the risk handling results back to the data collection, indicator calculation, and risk assessment stages.

[0052] Specifically, after receiving structured early warning events, the system automatically initiates corresponding preset collaborative handling processes based on the risk type (e.g., delivery risk, quality risk, financial risk) and risk level (low / medium / high). Through a visual early warning center, the system displays the overall risk situation in the form of heatmaps, topology maps, etc., and simultaneously pushes early warning work orders to the work portals of relevant personnel in procurement, legal, and supply chain departments. Relevant personnel then carry out their work according to the collaborative handling process: the procurement manager confirms the authenticity and scope of the risk event; the legal department assesses the breach of contract liability and response basis in the contract terms; the supply chain team develops alternative solutions (e.g., activating backup suppliers, adjusting logistics methods); and management makes the final decision. All communication and operations are recorded within the system. After risk handling is completed, relevant personnel record the handling results in the system, including handling measures, execution process, actual results (e.g., whether supply disruptions were avoided, cost changes, delivery time adjustments), and actual loss amount. The recorded risk handling results are fed back to the risk data lake to supplement internal data resources. Simultaneously, the handling results are used as new training data to adjust the weights of indicator calculations, optimize the thresholds of early warning rules, and improve the prediction accuracy of the prediction model, achieving a continuous optimization loop for the system.

[0053] The method provided in this embodiment firstly collects data from multiple systems within the supply chain and diverse external data. After cleaning and standardization, it establishes a multi-dimensional data association centered on "contracts / orders," penetrating upstream supplier networks and horizontally expanding to the external environment. This forms a unified risk data lake, breaking down the information silos caused by data fragmentation in traditional management and providing a complete and interconnected data foundation for subsequent analysis. Simultaneously, it uses natural language processing technology to parse key contract terms and map them into monitorable data indicators. Through a stream processing engine, it calculates indicator values ​​in real time based on preset calculation rules, constructing a standardized risk indicator dataset. This achieves a dynamic mapping of "contract-indicator-data," transforming static contract terms into "living" indicators active on the front lines of business. This completely changes the traditional contract "archive" model, enabling real-time perception of contract performance status.

[0054] Building upon this foundation, a hybrid intelligent model combining a "rule engine" and a "predictive model" is employed for risk assessment. The rule engine handles explicit, hard-line rules and regulations, while the predictive model learns from historical data to predict the probability of potential risks. Simultaneously, it integrates a supply chain knowledge graph to achieve penetrating risk tracing across multiple levels of suppliers, accurately locating risk sources and transmission paths. This achieves a precise diagnosis from "symptoms" to "root causes," addressing the limitations of traditional management, which can only cover first-tier suppliers and lacks risk tracing capabilities. Furthermore, the generated structured early warning events not only contain core information such as risk level and scope of impact but also simulate the cost and delivery impact of different response strategies based on data such as inventory and alternative supplier capacity. This implements a dual-engine approach of "risk warning + decision support," providing managers with quantitative decision-making support rather than simply issuing risk alerts.

[0055] Finally, by initiating a pre-set collaborative handling process and feeding back the handling results to each front-end stage, a closed loop of continuous optimization is formed. This enables the system to continuously improve the accuracy of data correlation, the rationality of indicator calculation, and the precision of risk assessment. In the long run, this method can effectively reduce the risk of supply chain disruption, optimize inventory levels, and improve negotiation efficiency. It transforms risk management from passive post-event remediation to proactive pre-event warning and in-event intervention, realizing the transformation of risk management from a "cost center" to a "value creation center." The resulting benefits, such as loss avoidance and efficiency improvement, can be directly or indirectly converted into financial returns, becoming an important part of the company's core competitiveness.

[0056] Overall, this application systematically addresses the pain points of traditional supply chain contract risk management, such as static nature, superficiality, lag, and lack of decision support, through the coordinated action of each step. It achieves multiple benefits, including dynamic real-time monitoring, penetrating traceability, intelligent early warning, and value creation.

[0057] Corresponding to the aforementioned embodiment of a supply chain contract core indicator penetration-based risk management method, this application also provides an embodiment of a supply chain contract core indicator penetration-based risk management device.

[0058] Figure 2 This is a schematic diagram of the supply chain contract core indicator penetration risk management device provided in Embodiment 2 of this application. Please refer to... Figure 2 The device provided in this embodiment includes a data acquisition module 210, a calculation module 220, an identification module 230, and a processing module 240.

[0059] The acquisition module 210 is used to collect internal and external data related to the supply chain, clean and standardize the internal and external data, establish multi-dimensional data associations, and form a unified risk data lake.

[0060] The calculation module 220 is used to parse key clauses of the contract text, map the parsed contract clauses into monitored data indicators, and calculate the data indicators based on the associated data in the risk data lake through preset calculation rules to obtain a standardized risk indicator dataset.

[0061] The identification module 230 is used to perform risk assessment based on the risk indicator dataset, through a hybrid intelligent mode of rule engine and prediction model, to identify risk level and locate risk source, and generate structured early warning events.

[0062] The processing module 240 is used to initiate a preset collaborative handling process based on the structured early warning event, record the risk handling results, and feed the risk handling results back to the data collection, indicator calculation and risk assessment stages.

[0063] The apparatus of this embodiment can be used to perform... Figure 1 The steps of the method embodiment shown are similar in principle and process, and will not be repeated here.

[0064] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0065] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0066] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A supply chain contract core indicator penetrating risk management method, characterized in that, The method includes: Collect internal and external data related to the supply chain, clean and standardize the internal and external data, establish multi-dimensional data associations, and form a unified risk data lake. The contract text is parsed for key clauses, and the parsed contract clauses are mapped to monitored data indicators. Based on the associated data in the risk data lake, the data indicators are calculated according to preset calculation rules to obtain a standardized risk indicator dataset. Based on the aforementioned risk indicator dataset, risk assessment is conducted through a hybrid intelligent mode combining rule engines and prediction models to identify risk levels, locate risk sources, and generate structured early warning events. Based on the structured early warning event, a preset collaborative handling process is initiated, the risk handling results are recorded, and the risk handling results are fed back to the data collection, indicator calculation, and risk assessment stages. Based on the correlated data in the risk data lake, the data indicators are calculated using preset calculation rules to obtain a standardized risk indicator dataset, including: A metrics calculation engine is built; the metrics calculation engine is based on a stream processing engine and extracts related datasets from the risk data lake on demand. The related datasets include order delivery records, quality inspection data, and payment data from internal data, as well as supplier financial data, compliance audit data, and regional risk index from external data. Configure preset calculation rules; the preset calculation rules include SQL query statements or target formulas, and set corresponding aggregation or statistical operation logic for different types of data indicators; The indicator calculation engine calls the corresponding preset calculation rules to perform calculations on the extracted associated dataset and obtain the original values ​​of each data indicator. The calculated raw values ​​of the indicators are mapped to a standardized scoring system of 0-100 points; The normalized indicator values ​​are linked and integrated with the metadata to form a structured dataset; Based on the aforementioned risk indicator dataset, risk assessment is performed using a hybrid intelligent model combining a rule engine and a prediction model. This identifies risk levels, locates risk sources, and generates structured early warning events, including: Configure the rule engine; preset hard warning rules in target form, which are formulated based on contract terms and business needs, including triggering conditions and corresponding warning levels; Build and train the prediction model; The rule engine performs real-time verification of the risk indicator dataset. When the data indicators meet the preset trigger conditions, it directly generates the corresponding level of warning. The prediction model receives real-time input from the risk indicator dataset, calculates using built-in mathematical functions, and outputs the probability of occurrence of various risk events within a preset future time period. The scoring card model comprehensively evaluates the results of the rule engine and the output of the prediction model. First, the individual indicator value is mapped to a sub-score, then the comprehensive risk score is calculated by weighting according to the preset weight, and finally the comprehensive score is mapped to different risk levels. Construct a supply chain knowledge graph that includes entities and relationships. When abnormal risk indicators are detected, use graph computing technology to traverse upstream along the knowledge graph, analyze the capacity, delivery performance, geographical location risk, and financial status of multi-level suppliers, and combine risk transmission path algorithms to infer the risk diffusion logic, locate the root cause of the risk and the transmission path. Integrate risk level, risk type, scope of impact, risk source analysis results, predicted probability, risk transmission path and preliminary response suggestions to form an early warning event and push it to the message queue or generate an early warning work order; For the highest-risk warning events, the impact of different response strategies on costs and delivery time is simulated based on correlated data, and the simulation results are added to the structured warning events.

2. The method according to claim 1, characterized in that, Establish multi-dimensional data relationships, including: Using contracts or orders as the core association key, a unique identifier is assigned to each contract and its corresponding purchase order. Basic information and execution data are associated through the identifier to form a panoramic view of contract performance. Construct a supply chain network map, and associate the capacity, delivery performance, and geographical location risk data of upstream multi-level suppliers based on the supply chain network map; By using the supplier's registered location, production location, and logistics route as key dimensions, contracts or orders are horizontally correlated with external environmental data.

3. The method according to claim 2, characterized in that, The basic information includes supplier master data, procurement categories, product specifications, and pricing terms. The execution data includes order fulfillment status, delivery records, logistics trajectory, warehousing information, quality inspection reports, invoices, and payment status. The external environment data includes macroeconomic data, weather data, maritime news, public opinion data, and geopolitical risk index.

4. The method according to claim 1, characterized in that, The internal data includes purchase orders, receipts, and payment data from the ERP system; supplier master data and performance scores from the SRM system; customer order forecast data from the CRM system; transportation status data from the TMS system; and original contract data from the legal CLM system. The external data includes financial credit and public opinion monitoring data from third-party data providers; weather, port congestion index, and commodity price data from public data sources; and external environmental data collected by IoT sensors.

5. The method according to claim 1, characterized in that, The contract text undergoes key clause analysis, and the analyzed clauses are mapped to monitored data metrics, including: Natural language processing technology was used to extract key clauses from the contract text, including clauses related to delivery time, quality standards, price agreement, liability for damages, service level agreement, payment cycle, and liquidated damages rules. For each key clause extracted, corresponding monitored data indicators are automatically created, and the indicator definitions, calculation dimensions and threshold standards are determined. Among them, on-time delivery rate and delivery delay days indicators are created based on delivery time clauses, batch pass rate indicators are created based on quality standard clauses, and frequency of breach and percentage of liquidated damages are created based on compensation liability and liquidated damages rules clauses. Establish a mapping relationship between the created data metrics and the associated data in the risk data lake to determine the data source required for metric calculation.

6. The method according to claim 1, characterized in that, The data metrics include delivery metrics, quality risk metrics, financial risk metrics, compliance and ESG metrics, and supply chain resilience metrics. Delivery metrics include on-time delivery rate and delivery delay days. Quality risk metrics include batch pass rate, customer complaint rate, and quality cost ratio. Financial risk metrics include price volatility, payment terms deviation, and supplier financial health. Compliance and ESG metrics include contract fulfillment rate, environmental compliance inspection pass rate, and labor standards audit score. Supply chain resilience metrics include dependence on a single supplier, order concentration, and average recovery time.

7. The method according to claim 1, characterized in that, Building and training a prediction model includes: Extract snapshots of risk indicator datasets from historical time points as feature variables; Each historical data snapshot corresponding to a feature variable is labeled; the label is determined based on whether the target risk event actually occurs within a subsequent preset time period, and the label value is 0 or 1. A machine learning classification algorithm is selected as the base model, and the machine learning classification algorithm includes one or more combinations of gradient boosting decision tree, logistic regression, and neural network. The initial parameters of the base model are randomly configured, and the number of iterations, learning rate, and regularization coefficient for model training are set. The prepared training data is divided into a training set and a validation set according to a preset ratio, and the training set is input into the initialized base model. The base model makes predictions on the training set data based on the current parameters and outputs the predicted probability of the occurrence of risk events; The difference between the predicted probability and the true label is calculated by the loss function. The loss value is backpropagated by the gradient descent algorithm. The internal parameters of the model are adjusted to minimize the loss function. The process of prediction, loss calculation and parameter adjustment is repeated until the performance index of the model on the validation set converges to the preset threshold. Historical data that was not used in training was used as a test set and input into the trained model to verify the model's generalization ability.

8. A supply chain contract core indicator penetration-based risk management device, characterized in that, The device includes an acquisition module, a calculation module, an identification module, and a processing module; The acquisition module is used to collect internal and external data related to the supply chain, clean and standardize the internal and external data, establish multi-dimensional data associations, and form a unified risk data lake. The calculation module is used to parse key clauses of the contract text, map the parsed contract clauses into monitored data indicators, and calculate the data indicators based on the associated data in the risk data lake through preset calculation rules to obtain a standardized risk indicator dataset. The identification module is used to assess risks based on the risk indicator dataset through a hybrid intelligent mode of rule engine and prediction model, identify risk levels and locate risk sources, and generate structured early warning events. The processing module is used to initiate a preset collaborative handling process based on the structured early warning event, record the risk handling results, and feed the risk handling results back to the data collection, indicator calculation and risk assessment stages. Based on the correlated data in the risk data lake, the data indicators are calculated using preset calculation rules to obtain a standardized risk indicator dataset, including: A metrics calculation engine is built; the metrics calculation engine is based on a stream processing engine and extracts related datasets from the risk data lake on demand. The related datasets include order delivery records, quality inspection data, and payment data from internal data, as well as supplier financial data, compliance audit data, and regional risk index from external data. Configure preset calculation rules; the preset calculation rules include SQL query statements or target formulas, and set corresponding aggregation or statistical operation logic for different types of data indicators; The indicator calculation engine calls the corresponding preset calculation rules to perform calculations on the extracted associated dataset and obtain the original values ​​of each data indicator. The calculated raw values ​​of the indicators are mapped to a standardized scoring system of 0-100 points; The normalized indicator values ​​are linked and integrated with the metadata to form a structured dataset; Based on the aforementioned risk indicator dataset, risk assessment is performed using a hybrid intelligent model combining a rule engine and a prediction model. This identifies risk levels, locates risk sources, and generates structured early warning events, including: Configure the rule engine; preset hard warning rules in target form, which are formulated based on contract terms and business needs, including triggering conditions and corresponding warning levels; Build and train the prediction model; The rule engine performs real-time verification of the risk indicator dataset. When the data indicators meet the preset trigger conditions, it directly generates the corresponding level of warning. The prediction model receives real-time input from the risk indicator dataset, calculates using built-in mathematical functions, and outputs the probability of occurrence of various risk events within a preset future time period. The scoring card model comprehensively evaluates the results of the rule engine and the output of the prediction model. First, the individual indicator value is mapped to a sub-score, then the comprehensive risk score is calculated by weighting according to the preset weight, and finally the comprehensive score is mapped to different risk levels. Construct a supply chain knowledge graph that includes entities and relationships. When abnormal risk indicators are detected, use graph computing technology to traverse upstream along the knowledge graph, analyze the capacity, delivery performance, geographical location risk, and financial status of multi-level suppliers, and combine risk transmission path algorithms to infer the risk diffusion logic, locate the root cause of the risk and the transmission path. Integrate risk level, risk type, scope of impact, risk source analysis results, predicted probability, risk transmission path and preliminary response suggestions to form an early warning event and push it to the message queue or generate an early warning work order; For the highest-risk warning events, the impact of different response strategies on costs and delivery time is simulated based on correlated data, and the simulation results are added to the structured warning events.

Images