An AI agent-based visual data analysis method and system

Abstract

The application provides a kind of visual data analysis method and system based on AI intelligent agent, it is related to artificial intelligence technical field, the method includes: step 1, through AI intelligent agent initiates multi-source data acquisition task autonomously, obtains multi-source heterogeneous supply chain operation data;Based on multi-source heterogeneous supply chain operation data, through AI intelligent agent constructs dynamic semantic conflict detection and resolution mechanism, identifies and unifies the data semantics of different business concepts, forms standardized data set;Step 2, based on standardized data set, through AI intelligent agent, with the central warehouse in supply chain, regional distribution center and main supplier production base as key entity, a multi-edge analysis framework is constructed;By setting the business boundary constraint of supply chain entity as joint limiting condition, the final division scheme of data partition in multi-edge analysis framework is solved under business boundary constraint.The application meets the operation control needs of supply chain refinement and intelligentization.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a visualization data analysis method and system based on AI intelligent agents. Background Technology

[0002] With the trend of digital operation of supply chain, enterprises have an increasingly urgent need for visualized data analysis of supply chain operation data. Existing supply chain data analysis methods that combine AI still have many shortcomings and are difficult to adapt to the needs of refined management and control.

[0003] When a large-scale FMCG company conducted supply chain operation risk analysis, it manually collected and integrated operational data from regional distribution centers, central warehouses, and upstream core suppliers. However, due to conflicting semantic definitions of inventory-related concepts across different platforms, inconsistent calculation methods for supplier capacity-related indicators, and the lack of data partitioning based on the business boundaries of each supply chain entity, the company used a fixed granularity to process data. While correlation analysis revealed superficial relationships, it failed to identify the complete risk transmission path from supplier supply anomalies to central warehouse allocation delays and regional distribution center inventory shortages. Static visualization reports lacked interactivity, and only general warnings were pushed when risks exceeded thresholds, without targeted handling suggestions. This resulted in untimely responses from the company, leading to delays in end-user supply and impacting market operations. Furthermore, the company demonstrated weak semantic integration capabilities for multi-source heterogeneous data, failed to build a dedicated analysis framework around key supply chain entities, lacked dynamic adjustment basis for data optimization, and its risk analysis lacked causal reasoning capabilities. The visualization and tiered warning mechanisms were also insufficiently targeted and timely. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a visualization data analysis method and system based on AI intelligent agents to meet the needs of refined and intelligent operation and management of the supply chain.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Firstly, a visualization data analysis method based on an AI agent, the method comprising: Step 1: The AI ​​agent autonomously initiates a multi-source data collection task to obtain multi-source heterogeneous supply chain operation data; based on the multi-source heterogeneous supply chain operation data, the AI ​​agent constructs a dynamic semantic conflict detection and resolution mechanism to identify and unify the data semantics of different business concepts, forming a standardized dataset. Step 2: Based on a standardized dataset, a multilateral analysis framework is constructed using an AI agent, with the central warehouse, regional distribution centers, and major supplier production bases in the supply chain as key entities. By setting the business boundary constraints of the supply chain entities as joint limiting conditions, the final partitioning scheme of the data within the multilateral analysis framework is solved under the business boundary constraints. The business scope associated with the multilateral analysis framework is partitioned into multiple feature evaluation blocks, and a dynamic adjustment parameter is obtained based on the data density and distribution pattern within each block. Step 3: Adjust the data field weights and aggregation granularity of the standardized dataset by dynamically adjusting parameters to obtain the optimized dataset. Then, perform data quality optimization processing through an AI agent to obtain a high-quality feature vector set. Step 4: Based on a high-quality feature vector set, conduct multi-dimensional correlation analysis and causal reasoning through an AI agent to identify risk transmission paths and key influencing factors in supply chain operations, and obtain risk assessment results. Step 5: Generate a hierarchical interactive visualization dashboard based on the assessment results, and trigger a tiered early warning and push response report when the risk exceeds the threshold.

[0006] Furthermore, AI agents autonomously initiate multi-source data collection tasks to acquire multi-source heterogeneous supply chain operation data. Based on this multi-source heterogeneous supply chain operation data, AI agents construct a dynamic semantic conflict detection and resolution mechanism to identify and unify the data semantics of different business concepts, forming a standardized dataset, including: The AI ​​agent is scheduled to autonomously detect and plan tasks for multi-source heterogeneous supply chain operation data sources, and generate multi-source data acquisition task instructions that include data source location, acquisition frequency and data format. Based on task instructions, the AI ​​agent acquires structured and unstructured supply chain operation data in real time from enterprise resource planning, logistics tracking platforms, supplier management platforms, and IoT sensor networks. Based on multi-source heterogeneous supply chain operation data, the AI ​​agent is driven to build an ontology mapping mechanism. Through semantic similarity calculation and business rule verification, semantic conflicts of the same business concepts in different data sources are dynamically detected. When a semantic conflict is detected, the AI ​​agent automatically performs semantic resolution operations based on a preset business priority strategy and contextual analysis, mapping the conflicting business concepts to a unified business terminology standard to obtain semantically resolved data; the semantically resolved data is then integrated, and data format standardization conversion and metadata registration are performed to obtain a standardized dataset.

[0007] Furthermore, based on a standardized dataset, a multilateral analysis framework is constructed using AI agents, with central warehouses, regional distribution centers, and major supplier production bases in the supply chain as key entities. By setting the business boundary constraints of supply chain entities as joint limiting conditions, the final data partitioning scheme within the multilateral analysis framework is solved under these constraints. The business scope associated with the multilateral analysis framework is partitioned into multiple feature evaluation blocks, and a dynamically adjusted parameter is obtained based on the data density and distribution pattern within each block, including: Receive standardized datasets and drive AI agents to identify central warehouses, regional distribution centers and major supplier production bases in the supply chain network as key entity nodes, and build a multilateral analysis framework that includes physical connections and business interactions between entities. The configuration of AI agents parameterizes the business boundary constraints of each supply chain entity, setting geographical coverage, inventory capacity thresholds, transportation timeliness requirements, and supplier cooperation terms as joint constraint conditions in the multilateral analysis framework. Under the constraints of joint positioning, the AI ​​agent spatially segments the business scope associated with the multilateral analysis framework and iteratively optimizes the final partitioning scheme of the computational data partition to obtain multiple independent feature evaluation blocks with complete business semantics. The system monitors the real-time data flow within each feature evaluation block, drives the AI ​​agent to calculate the data density index and distribution pattern characteristics of each block, and obtains dynamically adjusted parameters based on a weighted combination of the data density change rate and distribution dispersion.

[0008] Furthermore, configuring the AI ​​agent parameterizes the business boundary constraints of each supply chain entity, setting geographical coverage, inventory capacity thresholds, transportation timeliness requirements, and supplier cooperation terms as joint constraint conditions in the multilateral analysis framework, including: Based on the constructed multilateral analysis framework, the AI ​​agent is driven to extract the business attribute characteristics of each supply chain entity, and parameterize the geographical coverage radius of the central warehouse, the upper limit of the inventory capacity of the regional distribution center, the transportation time commitment of the main supplier's production base, and the service level agreement in the supplier cooperation terms into a numerical constraint vector. Configure the AI ​​agent to establish a constraint mapping table, and bind the parameterized numerical constraint vectors to the entity nodes in the multilateral analysis framework one-to-one to form an entity-constraint relationship; The relationship between entities and constraints is transformed into joint constraints in a multilateral analysis framework, where the geographical coverage radius corresponds to the spatial movement range constraint, the upper limit of inventory capacity corresponds to the capacity threshold constraint, the transportation timeliness commitment corresponds to the time dimension constraint, and the service level agreement corresponds to the quality dimension constraint. By performing conflict detection and coordination optimization on joint limit conditions, when the business boundary constraints of multiple entities overlap, the joint limit parameters are automatically adjusted based on the preset business priority weights and constraint relaxation strategies to obtain conflict-free joint limit conditions.

[0009] Furthermore, by dynamically adjusting the parameters of the data field weights and aggregation granularity of the standardized dataset, an optimized dataset is obtained. Then, an AI agent performs data quality optimization processing to obtain a high-quality feature vector set, including: By dynamically adjusting parameters and a standardized dataset, the AI ​​agent is driven to quantify the importance of each data field in the standardized dataset based on the data density change rate index in the dynamically adjusted parameters, and dynamically allocate field weight coefficients to obtain a weighted set of data fields. Based on the distribution dispersion characteristics in the dynamically adjusted parameters, the spatiotemporal dimension aggregation granularity of the weighted data field set is optimized. According to the geographical distribution density of business entities and the frequency of business operations, the time window granularity and spatial region granularity of data aggregation are automatically adjusted to obtain a multi-granularity optimized dataset. The AI ​​agent is scheduled to perform data quality optimization processing on the multi-granularity optimized dataset. Through outlier detection and correction, intelligent filling of missing values, data consistency verification and noise filtering, outliers, missing items and inconsistent records in the data are eliminated to obtain a cleaned optimized dataset. By performing feature engineering on the purified and optimized dataset, temporal features, spatial features, and business-related features are extracted. Through feature selection and dimensionality reduction operations, a high-quality feature vector set with compressed dimensions and rich information is obtained.

[0010] Furthermore, based on a high-quality feature vector set, multi-dimensional correlation analysis and causal reasoning are performed using AI agents to identify risk transmission paths and key influencing factors in supply chain operations, resulting in risk assessment results, including: By using high-quality feature vector sets, the AI ​​agent is driven to perform multi-dimensional correlation analysis on the temporal features, spatial features and business-related features in the feature vector sets, calculate the feature similarity and dependency strength between different supply chain nodes, and obtain a multi-dimensional correlation network. Based on a multi-dimensional network of relationships, this study identifies the direction and intensity of risk transmission from suppliers to customers in supply chain operations by verifying the direction of relationships and quantifying causal effects, and by using intervention simulation and counterfactual reasoning, thus obtaining a risk transmission path map. By identifying key nodes in the risk transmission path map, calculating the centrality index and vulnerability coefficient of each node in the risk transmission process, and combining business rule constraints and historical risk event data, a set of key influencing factors that have a decisive impact on the overall stability of the supply chain is selected. By quantitatively assessing the risk level of a set of key influencing factors, a comprehensive risk score is calculated based on parameters such as the factor's impact range, transmission speed, and recovery difficulty, generating a risk assessment result that includes risk type, risk level, impact range, and confidence level.

[0011] Furthermore, based on the assessment results, a tiered interactive visual dashboard is generated, and tiered alerts are triggered when risks exceed thresholds, along with response reports, including: Based on the risk assessment results, the AI ​​agent is driven to perform hierarchical and structured processing of the risk assessment results according to the risk type, risk level and impact scope, generating a three-layer data structure including a global overview layer, a regional analysis layer and an entity details layer; Based on a three-layer data structure, a hierarchical interactive visualization dashboard is constructed. The global overview layer displays a heat map of overall supply chain risk, the regional analysis layer presents a risk distribution matrix of distribution centers in each region, and the entity details layer displays risk indicator trend charts of specific suppliers and warehouses. Drill-down interactive controls are set for each layer. The system monitors risk level indicators in a hierarchical interactive visualization dashboard in real time. When any risk indicator exceeds a preset threshold, the AI ​​agent is driven to determine the warning level and trigger the corresponding graded warning mechanism based on the severity of the risk, the breadth of its impact, and the speed of its transmission. Based on the triggered tiered early warning level and the corresponding risk transmission path, a response report is automatically generated, which includes risk root cause analysis, impact range prediction, and handling suggestions. The response report is then pushed to relevant business personnel and management through email and SMS message channels.

[0012] Secondly, a visualization data analysis system based on AI intelligent agents includes: The acquisition module is used to autonomously initiate multi-source data collection tasks through AI agents to acquire multi-source heterogeneous supply chain operation data; based on the multi-source heterogeneous supply chain operation data, a dynamic semantic conflict detection and resolution mechanism is constructed through AI agents to identify and unify the data semantics of different business concepts and form a standardized dataset. The computation module is used to construct a multilateral analysis framework based on a standardized dataset, using AI agents with central warehouses, regional distribution centers, and major supplier production bases in the supply chain as key entities. By setting the business boundary constraints of the supply chain entities as joint limit conditions, the module solves the final partitioning scheme of the data within the multilateral analysis framework under the business boundary constraints. The module partitions the business scope associated with the multilateral analysis framework into multiple feature evaluation blocks, and obtains a dynamically adjusted parameter based on the data density and distribution pattern within each block. The adjustment module is used to dynamically adjust the weights and aggregation granularity of data fields in the standardized dataset to obtain an optimized dataset. The AI ​​agent then performs data quality optimization to obtain a high-quality feature vector set. The assessment module is used to identify risk transmission paths and key influencing factors in supply chain operations by performing multi-dimensional correlation analysis and causal reasoning based on high-quality feature vector sets and AI agents, thereby obtaining risk assessment results. The processing module is used to generate hierarchical interactive visualization dashboards based on the assessment results, and to trigger tiered early warnings and push response reports when the risk exceeds the threshold.

[0013] Thirdly, a computing device includes: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0014] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0015] The above-described solution of the present invention has at least the following beneficial effects: Because it employs a technical approach that uses an AI agent as the main executor throughout the entire process, autonomously initiates multi-source data collection and constructs a dynamic semantic conflict detection and resolution mechanism; builds a multilateral analysis framework based on key supply chain entities and sets business boundary constraints as joint limiting conditions to achieve data partitioning and generate dynamically adjustable parameters; optimizes data field weights and aggregation granularity through these parameters and performs data quality optimization to obtain a high-quality feature vector set; performs multi-dimensional correlation analysis and causal reasoning based on the feature vector set to identify risk transmission paths and key influencing factors; generates hierarchical interactive visualization dashboards and triggers tiered early warnings and pushes response reports, it overcomes the high degree of human intervention required in existing technologies. The technical shortcomings include weak semantic integration capabilities of multi-source heterogeneous data, lack of dedicated analysis frameworks around key entities in the supply chain, lack of dynamic basis for data optimization, risk analysis that can only perform surface correlation analysis and cannot explore causal transmission paths, and insufficient targeting and timeliness of visualization and hierarchical early warning. Therefore, the technical effects achieved include: realizing full-process autonomy and intelligence in supply chain operation data analysis; improving the standardization and quality of data integration; accurately identifying the root causes and complete transmission paths of supply chain risks; achieving intuitive visualization and accurate hierarchical early warning of risks; improving the efficiency of risk response and handling; and meeting the technical needs of refined and intelligent operation data analysis and risk management in the supply chain. Attached Figure Description

[0016] Figure 1This is a flowchart illustrating a visualization data analysis method based on an AI agent, provided by an embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of a visualization data analysis system based on an AI intelligent agent provided by an embodiment of the present invention. Detailed Implementation

[0018] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art.

[0019] like Figure 1 As shown, an embodiment of the present invention proposes a visualization data analysis method based on an AI intelligent agent, the method comprising the following steps: Step 1: The AI ​​agent autonomously initiates a multi-source data collection task to obtain multi-source heterogeneous supply chain operation data; based on the multi-source heterogeneous supply chain operation data, the AI ​​agent constructs a dynamic semantic conflict detection and resolution mechanism to identify and unify the data semantics of different business concepts, forming a standardized dataset. Step 2: Based on a standardized dataset, a multilateral analysis framework is constructed using an AI agent, with the central warehouse, regional distribution centers, and major supplier production bases in the supply chain as key entities. By setting the business boundary constraints of the supply chain entities as joint limiting conditions, the final partitioning scheme of the data within the multilateral analysis framework is solved under the business boundary constraints. The business scope associated with the multilateral analysis framework is partitioned into multiple feature evaluation blocks, and a dynamic adjustment parameter is obtained based on the data density and distribution pattern within each block. Step 3: Adjust the data field weights and aggregation granularity of the standardized dataset by dynamically adjusting parameters to obtain the optimized dataset. Then, perform data quality optimization processing through an AI agent to obtain a high-quality feature vector set. Step 4: Based on a high-quality feature vector set, conduct multi-dimensional correlation analysis and causal reasoning through an AI agent to identify risk transmission paths and key influencing factors in supply chain operations, and obtain risk assessment results. Step 5: Generate a hierarchical interactive visualization dashboard based on the assessment results, and trigger a tiered early warning and push response report when the risk exceeds the threshold.

[0020] In this embodiment of the invention, by employing an AI agent as the core, autonomously initiating multi-source heterogeneous supply chain operation data collection tasks, constructing a dynamic semantic conflict detection and resolution mechanism to form a standardized dataset, building a multilateral analysis framework based on key supply chain entities, setting business boundary constraints as joint limit conditions for data partitioning and obtaining dynamically adjustable parameters, adjusting data field weights and aggregation granularity through these parameters, and obtaining a high-quality feature vector set through data quality optimization, multi-dimensional correlation analysis and causal reasoning based on this vector set to obtain risk assessment results, and finally generating a hierarchical interactive visualization dashboard to trigger tiered early warnings and push response reports, this invention overcomes the technical problems in existing supply chain data analysis, such as high degree of human intervention, inconsistent semantics of multi-source data, lack of a dedicated analysis framework tailored to supply chain entities, lack of dynamic basis for data optimization, inaccurate identification of risk transmission paths, and imperfect visualization and early warning mechanisms. This achieves the technical effects of realizing full-process autonomy in supply chain operation data analysis, improving data standardization and quality, accurately identifying supply chain operation risks and transmission paths, intuitively presenting risk results and timely triggering accurate early warnings and response pushes, improving the efficiency and accuracy of supply chain operation data analysis and risk management, and meeting the technical needs of refined and intelligent supply chain operation.

[0021] In a preferred embodiment of the present invention, step 1 above may include: Step 1.1: Schedule the AI ​​agent to autonomously explore and plan tasks for multi-source heterogeneous supply chain operation data sources, and generate multi-source data acquisition task instructions that include data source location, acquisition frequency and data format. Specifically, this includes: scheduling the AI ​​agent to carry out autonomous exploration of multi-source heterogeneous supply chain operation data sources; the AI ​​agent performs a full-domain scan of supply chain operation-related data sources inside and outside the enterprise, identifies the specific deployment location and data output port of the enterprise resource planning system, logistics tracking platform, supplier management platform and IoT sensor network, and obtains the data output method and collectable data type of each data source. After completing data source detection, the AI ​​agent plans data collection tasks based on the business needs of real-time monitoring of supply chain operations. For structured data, the collection frequency is set to once every five minutes, and for unstructured data, it is set to once every ten minutes. Simultaneously, the output data format requirements for various data sources are obtained, with structured data required to be in tabular form and unstructured data required to be in text form. The AI ​​agent integrates all detected data source location information, the planned collection frequency, and the specified data format requirements, and obtains the corresponding collection execution nodes and data transmission paths for each data source, generating a complete multi-source data collection task instruction. The instruction provides clear and specific regulations on the execution requirements and logic for each collection task. The AI ​​agent in this embodiment is an integrated artificial intelligence system specifically designed for the entire process of supply chain operation data processing. Its core positioning is to replace manual labor in achieving autonomous and precise closed-loop operations for data collection, processing, analysis, early warning, and reporting. It possesses core characteristics of strong scenario adaptability, automated decision-making, and high execution efficiency. Essentially, it is a technological complex integrating four core capabilities: autonomous perception, intelligent decision-making, task execution, and dynamic adaptation. Guided by the business needs of refined supply chain operation and risk management, it incorporates dedicated technical units such as a data source detection core component, task planning and processing logic, data format parsing functional unit, transmission and collaboration component, semantic conflict detection and resolution logic, risk analysis core component, and visualization dashboard construction functional unit. By integrating multi-domain AI technologies and supply chain business rules, this agent autonomously connects to multiple heterogeneous data sources within and outside the enterprise, completing data collection planning, semantic unification, partitioned processing, feature extraction, risk identification, visualization display, and tiered early warning, without requiring manual intervention in decision-making and operations at any stage.

[0022] The construction of the AI ​​agent revolves around the needs of the entire supply chain data processing process. It is implemented step-by-step through four core stages: target alignment, functional decomposition, technology integration, and iterative optimization, ensuring that functions are compatible with each step of the operation. Core objectives are precisely aligned with scenarios, supporting the entire process of processing multi-source heterogeneous supply chain data from collection to early warning without human intervention. It covers seven core scenarios: data collection, semantic integration, partition analysis, feature engineering, risk identification, visualization, and tiered early warning, meeting the needs of refined supply chain operation and risk management. The processing scope is limited to supply chain operation-related data, including all business processes such as procurement, production, warehousing, logistics, and distribution, encompassing both structured and unstructured data, and adapting to various entity nodes such as central warehouses, regional distribution centers, and supplier production bases. Performance indicators are set as follows: data source detection accuracy ≥98%, semantic conflict resolution success rate ≥95%, data partitioning business semantic integrity ≥98%, risk identification confidence ≥80%, and early warning response latency ≤1 minute, ensuring compatibility with the stringent operational requirements of each step.

[0023] Functional breakdown and customized setup: Based on the supply chain data processing workflow, eight core functions are broken down and built. Each function operates independently yet collaboratively, covering the entire process requirements. The data source detection function, a core feature, supports autonomous data source detection. It integrates network scanning technology, port identification technology, and data type parsing algorithms to scan the entire internal and external network of the enterprise, locating the deployment location and output port of data sources such as the Enterprise Resource Planning System and logistics tracking platform. It automatically identifies the data output method and collectable data types, providing basic data for task planning. The task planning and execution function supports data collection task planning and data collection execution. It incorporates data classification and identification algorithms, frequency adaptation rules, and data transmission protocols to distinguish between different data types. Structured and unstructured data are collected at differentiated frequencies: structured data every 5 minutes and unstructured data every 10 minutes. A secure data interaction link is established, and data extraction and aggregation are completed according to rules. Semantic conflict handling functionality supports conflict detection and resolution. A supply chain business concept ontology is built, integrating semantic similarity calculation algorithms, business rule verification logic, and priority strategies. Semantic conflicts from different data sources are dynamically detected, and conflicts are automatically resolved based on business priority and contextual association, achieving data semantic unification. Multilateral analysis and partitioning functionality supports framework construction and data partitioning, integrating entity recognition algorithms, constraint parameterization logic, spatial segmentation, and iterative optimization methods to identify key entities in the supply chain. The system comprises nodes that construct a multilateral analysis framework, transforming business boundary constraints into joint limiting conditions, completing data partitioning with complete business semantics, and generating dynamically adjustable parameters. Data optimization and feature engineering functions support data processing and feature extraction, incorporating weight allocation algorithms, spatiotemporal aggregation granularity optimization logic, data quality optimization tools such as outlier detection and missing value imputation, and feature extraction and dimensionality reduction methods. Based on dynamically adjusted parameters, it optimizes data field weights and aggregation granularity, purifies data, and extracts multi-dimensional core features. Risk analysis functions support risk identification and assessment, integrating multi-dimensional correlation analysis algorithms, causal reasoning logic, and geometric algorithms, such as the algorithm for finding the minimum circumcircle of a point set and risk quantification assessment rules. It constructs a network of relationships, identifies risk transmission paths, and screens key influencing factors to generate multi-dimensional risk assessment results. Its visualization and early warning functions support dashboard construction and tiered early warning, incorporating layered data structuring processing logic, visualization rendering technology, interactive control development tools, and early warning level determination rules to generate a three-tiered interactive visualization dashboard that monitors risk indicators in real time and triggers tiered early warnings. Its response report generation and push functions support report generation and push, integrating report template processing logic and multi-channel push components. Based on early warning levels and risk transmission paths, it automatically generates response reports containing root cause analysis, impact prediction, and handling suggestions, and accurately pushes them through channels such as email and SMS.

[0024] Technology stack integration and scenario adaptation optimization: A technology stack adapted to supply chain data processing scenarios is selected to ensure functional implementation and compatibility. The perception layer technology integrates web crawlers, port scanning tools, and data type recognition algorithms to support full-domain scanning and information extraction for data source detection. The decision layer technology employs rule engines, business requirement matching algorithms, causal reasoning logic, and geometric calculation logic to support core decision-making processes such as task planning, semantic conflict resolution, and risk analysis. The execution layer technology integrates data transmission protocols, path optimization algorithms, data cleaning tools, visualization rendering libraries, and multi-channel push interfaces to support execution operations such as data acquisition, format conversion, dashboard construction, and report push. Adaptability optimization: Multi-data source interface adaptation capabilities are reserved to ensure compatibility with communication protocols and data formats from different data sources such as enterprise resource planning systems and IoT sensor networks. The adaptation logic of geometric algorithms to supply chain scenarios is optimized to ensure the accuracy of spatial analysis and node correlation analysis.

[0025] Scenario-based training and iterative optimization involve training with real-world supply chain business scenario data to continuously optimize the adaptability and stability of the intelligent agent. Training data preparation includes collecting supply chain data source configuration information, historical data collection schemes, business rule documents, and risk event records from multiple industries such as FMCG and manufacturing to construct a scenario-based training dataset covering all business processes. Algorithm performance tuning optimizes the algorithm performance of each function based on the training data, such as improving the identification accuracy of data source detection, adjusting the frequency adaptation logic of task planning, and optimizing the causal inference accuracy of risk analysis, ensuring the rationality and efficiency of each step. Abnormal scenario testing simulates unexpected situations such as data source port changes, abnormal data formats, network interruptions, and constraint conflicts to optimize the agent's adaptive adjustment capabilities, ensuring that core operations can still be completed in abnormal scenarios, or that abnormal prompts can be generated and feedback provided in a timely manner, guaranteeing stable operation throughout the entire process.

[0026] Step 1.2: Based on task instructions, the AI ​​agent acquires structured and unstructured supply chain operation data in real time from the Enterprise Resource Planning (ERP) system, logistics tracking platform, supplier management platform, and IoT sensor network. Specifically, this includes: The AI ​​agent establishes secure and stable data interaction links with the ERP system, logistics tracking platform, supplier management platform, and IoT sensor network based on the generated multi-source data acquisition task instructions; first, it connects with the ERP system to extract structured supply chain operation data such as inventory data, order data, production plan data, and procurement data in real time; then, it connects with the logistics tracking platform to obtain node data, transportation trajectory data, and distribution data of goods. The system delivers timely delivery data, logistics node signing data, and other supply chain operation data; it also connects to the supplier management platform to collect structured supply chain operation data such as each supplier's capacity data, supply plan data, cooperation fulfillment data, and raw material inventory data; finally, it fully connects to the IoT sensor network to collect unstructured supply chain operation data such as warehouse temperature and humidity data, cargo storage status data, transportation vehicle operation status data, and production equipment operating status data; during the data collection process, the AI ​​agent strictly follows the frequency set by the task instructions to perform operations, with structured data being extracted and transmitted every five minutes and unstructured data being collected and summarized every ten minutes, forming a multi-source heterogeneous supply chain operation data set.

[0027] Step 1.3: Based on multi-source heterogeneous supply chain operation data, drive the AI ​​agent to build an ontology mapping mechanism. Through semantic similarity calculation and business rule verification, dynamically detect semantic conflicts of the same business concepts in different data sources. Specifically, this includes: Based on the collected multi-source heterogeneous supply chain operation data, the AI ​​agent drives the construction of an ontology mapping mechanism. First, build the basic architecture of the ontology mapping mechanism, sort out the core business concepts of the entire supply chain operation process, and establish a supply chain business concept ontology library covering all business links of procurement, production, warehousing, logistics, and distribution. The ontology library contains the standard definition, business connotation, and applicable scenarios of each business concept. Then, the AI ​​agent matches the business concepts in different data sources with the standard business concepts in the ontology library one by one. By calculating semantic similarity, the AI ​​agent compares the differences between the expressions and actual business connotations of the same business concepts from different data sources. Simultaneously, it verifies the reasonableness of the matching results by incorporating business rules for supply chain operations. These rules cover both internal supply chain operation management standards and industry-standard supply chain operations. During the verification process, a semantic similarity threshold of 85 is set. When the semantic similarity calculation result for the same business concept from different data sources falls below 85, the AI ​​agent determines that the business concept has a semantic conflict. It then accurately marks this type of semantic conflict, recording the data source, specific expression, and corresponding business process of the conflicting business concept. This enables dynamic detection of semantic conflicts in the same business concept data from different data sources.

[0028] Step 1.4: When a semantic conflict is detected, the AI ​​agent automatically performs semantic resolution based on a preset business priority strategy and contextual analysis, mapping the conflicting business concepts to a unified business terminology standard to obtain semantically resolved data. The semantically resolved data is then integrated, and data format standardization and metadata registration are performed to obtain a standardized dataset. Specifically, after the AI ​​agent detects and marks a semantic conflict, it automatically performs semantic resolution based on a preset business priority strategy and contextual analysis. First, it retrieves the preset business priority strategy, which sets priority levels based on the authority of the data source and the core nature of the business process. Internal enterprise resource planning data has higher priority than external logistics tracking and supplier management platform data; data from core business processes such as procurement and production has higher priority than data from auxiliary business processes such as warehousing and logistics. The AI ​​agent then analyzes the conflicting business concepts within the supply chain operation process using contextual analysis. In practical application scenarios, the core business connotations and actual business needs corresponding to conflicting business concepts are obtained. Then, based on business priority strategies and context analysis results, the conflicting business concepts are uniformly mapped to the unified business terminology standard in the supply chain business concept ontology library, completing the resolution of all semantic conflicts and obtaining semantically resolved supply chain operation data. The AI ​​agent comprehensively integrates the semantically resolved data, performs data format standardization conversion according to preset format standards, and uniformly converts all unstructured data into structured tabular form. At the same time, it unifies the field names, field types, and data units of all data, eliminating data format differences from different data sources. After the format conversion is completed, the AI ​​agent performs metadata registration operations on all converted data, adding metadata information such as data source identifier, collection time, business link affiliation, and data type to each data, completing the unified entry and filing of metadata, and finally forming a standardized dataset with unified semantics, standard format, and complete information.

[0029] In this embodiment of the invention, by employing a scheduling AI agent to autonomously detect multi-source heterogeneous supply chain operation data sources, plan collection tasks, and generate collection instructions, and then acquiring structured and unstructured data in real time from multiple platforms and sensor networks based on these instructions, the AI ​​agent is driven to construct an ontology mapping mechanism. Semantic conflicts are dynamically detected through semantic similarity calculation and business rule verification. Conflicts are automatically resolved by combining preset business priority strategies and context association analysis. Finally, the data is integrated and format standardization conversion and metadata registration are performed to form a standardized dataset. Therefore, this invention overcomes the technical problems of existing technologies, such as reliance on manual labor for multi-source supply chain data collection, low efficiency, inconsistent semantics of the same business concepts from different data sources, conflicting data formats, and difficulty in integration, which prevent the formation of a standardized dataset and thus affect the accuracy and efficiency of subsequent data analysis. As a result, this invention achieves autonomous and intelligent collection of multi-source supply chain operation data, accurately detects and effectively resolves semantic conflicts, unifies business terminology and data formats, and generates a high-quality, standardized dataset.

[0030] In a preferred embodiment of the present invention, step 2 above may include: Step 2.1: Receive the standardized dataset and drive the AI ​​agent to identify the central warehouse, regional distribution centers, and major supplier production bases in the supply chain network as key entity nodes. Construct a multilateral analysis framework that includes physical connections and business interactions between entities. Specifically, the AI ​​agent first receives the standardized dataset and completes data access and verification. After confirming the semantic uniformity and format standardization of the dataset, it performs a full-domain analysis of the supply chain operation-related information in the dataset, extracting core content such as entity identifiers, physical locations, business affiliations, and data interaction records. Based on the business architecture and data correlation of the supply chain network, the AI ​​agent identifies the central warehouse, regional distribution centers, and major supplier production bases as key entity nodes. The identification criteria are the pivotal position of each entity in the entire supply chain procurement, production, warehousing, logistics, and distribution process and the frequency of data interactions. Core hub-type entities are directly identified as key entity nodes. After node identification, the AI ​​agent begins to construct a multilateral analysis framework. First, it builds the basic topology of the framework, spatially arranging each key entity node according to its actual geographical location. Then, it sorts out the physical connections between the nodes, marking the transportation routes between the central warehouse and regional distribution centers, the distribution paths between regional distribution centers and terminal channels, and the supply channels between major supplier production bases and the central warehouse, among other physical connection information. At the same time, it sorts out the business interaction relationships between the nodes, recording the procurement fulfillment relationships between supplier production bases and warehouses, the inventory transfer relationships between warehouses and distribution centers, and the order response relationships between distribution centers, among other business interaction information, to obtain the interaction type, interaction frequency, and data flow direction. Finally, the AI ​​agent integrates the attribute information of all key entity nodes and the physical connections and business interaction relationships between the nodes to form a complete multilateral analysis framework.

[0031] Step 2.2: Configure the AI ​​agent to parameterize the business boundary constraints of each supply chain entity. Geographic coverage, inventory capacity threshold, transportation timeliness requirements, and supplier cooperation terms are set as key constraint conditions in the multilateral analysis framework. Specifically, the AI ​​agent first extracts relevant information on the business boundary constraints of each key entity node from the standardized dataset and enterprise supply chain operation management specifications. For the central warehouse, regional distribution center, and major supplier production bases, four core constraint information categories are extracted: geographic coverage, inventory capacity threshold, transportation timeliness requirements, and supplier cooperation terms. The AI ​​agent then parameterizes the extracted business boundary constraints, converting the geographic coverage of the central warehouse into a radius value (set to 500), the inventory capacity threshold of the regional distribution center into a maximum storage capacity value (set to 10000), the transportation timeliness requirements of major supplier production bases into a delivery cycle value (set to 7), and the supplier cooperation terms into service level-related values, setting the fulfillment rate to 95% and the supply qualification rate to 98%. All constraint information is converted into quantifiable values. Numerical constraint vectors are generated; subsequently, the AI ​​agent establishes a constraint mapping table, matching the corresponding numerical constraint vector to each key entity node in the table, achieving a one-to-one correspondence between entity nodes and constraint conditions, forming the obtained entity-constraint relationship; the AI ​​agent transforms this relationship into joint constraint conditions in a multilateral analysis framework, using the radiation radius value corresponding to the geographical coverage as the spatial movement range constraint, the maximum storage value corresponding to the inventory capacity threshold as the capacity threshold constraint, the delivery cycle value corresponding to the transportation timeliness requirement as the time dimension constraint, and the service level value corresponding to the supplier cooperation terms as the quality dimension constraint; the AI ​​agent performs conflict detection on all joint constraint conditions. If overlapping business boundary constraints of multiple entities are detected, such as overlapping geographical coverage of regional distribution centers, the joint constraint parameters are automatically adjusted based on preset business priority weights and constraint relaxation strategies. The priority weight of core business entities is set to 80, and that of general business entities is set to 20. The constraint parameters of overlapping areas are adjusted according to the weight ratio, ultimately obtaining conflict-free joint constraint conditions.

[0032] Step 2.3: Under the constraint of joint limiting conditions, the AI ​​agent spatially segments the business scope associated with the multilateral analysis framework. Through iterative optimization of the final partitioning scheme for the computational data partitions, multiple independent feature evaluation blocks with complete business semantics are obtained. Specifically, the AI ​​agent uses conflict-free joint limiting conditions as the core constraint to perform spatial segmentation operations on the overall supply chain business scope associated with the multilateral analysis framework. First, based on the spatial movement range limits of each key entity node, the basic business space boundaries of each entity are defined. Then, combined with capacity threshold limits, time dimension limits, and quality dimension limits, the basic boundaries are corrected to ensure that each segmented region conforms to the full-dimensional business boundary constraints of the entity. After completing the initial spatial segmentation, the AI ​​agent obtains multiple initial data partitions. Subsequently, business semantic verification is performed on each initial data partition to check whether the data source within the partition covers the complete business processes of the corresponding entity node, such as procurement, production, warehousing, and logistics, ensuring the business semantic integrity of the partitioned data. The iterative optimization process is initiated, with the iteration termination condition set at 98% business semantic completeness for all data partitions. In each iteration, the AI ​​agent calculates the data correlation degree within each initial data partition, compares the matching degree between the partition boundaries and the entity business boundaries, merges and adjusts data partitions with incomplete business semantics, and splits large partitions with data correlation degrees below a threshold of 85. Simultaneously, it corrects the range of partitions with ambiguous boundaries, repeatedly executing iterative calculations and partition adjustments. When all data partitions meet the business semantic completeness and data correlation threshold requirements, the iterative optimization process terminates. The AI ​​agent calculates the final partitioning scheme for the data partitions. Based on the final partitioning scheme, the AI ​​agent divides the business scope of the multilateral analysis framework into multiple feature evaluation blocks. Each block is independent and possesses complete business semantics. Each block is labeled with information such as the corresponding key entity node business coverage, data type attribution, etc., and each block is precisely associated with a standardized dataset to obtain the data source range for each block.

[0033] Step 2.4: Monitor the real-time data flow within each feature evaluation block, driving the AI ​​agent to calculate the data density index and distribution pattern characteristics of each block. Based on a weighted combination of the data density change rate and distribution dispersion, dynamic adjustment parameters are obtained. Specifically, this includes: continuously monitoring the real-time data flow of supply chain operations within each feature evaluation block, acquiring data flow information such as data access volume, data update speed, and data business type distribution in real time at a collection frequency of once per minute, ensuring full-domain perception of dynamic changes in data within the block; for each feature evaluation block, the AI ​​agent calculates the data density index based on the collected real-time data flow information, statistically analyzing the actual data volume within a unit time and unit business scope to obtain the real-time data density value of each block, while continuously tracking the data density change trend and recording density fluctuations at different time stages; the AI ​​agent performs in-depth analysis of the data distribution pattern characteristics within each block. The system identifies whether the data distribution is centralized or discrete, statistically analyzes the distribution proportions of data across different business segments such as procurement, production, warehousing, and logistics, and examines the distribution patterns over different time periods to extract key information reflecting the core characteristics of the data distribution in each block. The AI ​​agent calculates the data density change rate for each block, which is the ratio of the difference between two consecutive real-time data density values ​​to the previous density value. Simultaneously, it calculates the data distribution dispersion, setting a baseline value of 50, and obtaining specific values ​​based on the deviation of the actual data distribution from the baseline distribution. The AI ​​agent then performs a weighted combination based on the data density change rate and the distribution dispersion, setting the weight of the data density change rate at 60% and the distribution dispersion at 40%. Based on these weights, it calculates the weighted composite value for each block. Finally, considering the overall operational rhythm of the supply chain and data analysis needs, it performs a global calibration of the weighted composite value for all blocks, ultimately obtaining dynamic adjustment parameters that accurately reflect the data characteristics of each feature-evaluated block.

[0034] In this embodiment of the invention, by employing a method that drives an AI agent to receive standardized datasets and identify key entity nodes in the supply chain, constructing a multilateral analysis framework that includes physical connections and business interactions between entities, configuring the AI ​​agent to parameterize the business boundary constraints of each entity, and setting geographical coverage, inventory capacity thresholds, etc., as joint limiting conditions, and completing data partitioning through spatial segmentation and iterative optimization under these conditions to form feature evaluation blocks with complete business semantics, while simultaneously monitoring the real-time data flow of each block, calculating data density and distribution pattern characteristics, and obtaining dynamically adjusted parameters through weighted combination, this invention overcomes the technical problems in existing technologies that fail to construct a dedicated analysis framework around key entities in the supply chain, cannot transform business boundary constraints into effective conditions for data partitioning, have data partitioning that does not fit the supply chain business scenario and has incomplete semantics, and lack a basis for dynamically adjusting subsequent data processing parameters based on the actual data distribution, thus affecting the accuracy of data optimization and risk analysis. This invention achieves the construction of a dedicated multilateral analysis framework that fits the actual business scenario of the supply chain, realizes accurate and semantically relevant data partitioning, and generates dynamically adjusted parameters that reflect the real-time distribution characteristics of the data.

[0035] In a preferred embodiment of the present invention, step 2.2 above may include: Step 2.21: Based on the constructed multilateral analysis framework, drive the AI ​​agent to extract the business attribute features of each supply chain entity. Parameterize the geographical coverage radius of the central warehouse, the inventory capacity limit of the regional distribution center, the transportation timeliness commitment of the major supplier's production base, and the service level agreement in the supplier cooperation terms into numerical constraint vectors. Specifically, this includes: The AI ​​agent, based on the constructed multilateral analysis framework, first performs global positioning of all key supply chain entity nodes within the framework, obtaining the entity type of each node as a central warehouse, regional distribution center, or major supplier's production base. Simultaneously, it retrieves the core business attribute information of each entity node in the framework. Based on the entity type and business attributes, the AI ​​agent selectively extracts the key business attribute features of each supply chain entity, extracting the geographical coverage radius feature for the central warehouse, and the geographical coverage radius feature for the regional distribution center, etc. The system extracts the upper limit of inventory capacity for the regional distribution center, the transportation timeliness commitment for the main supplier's production base, and the service level agreement (SLA) features from the cooperation terms for all cooperating suppliers. After feature extraction, the AI ​​agent parametrically transforms all features, converting abstract business attribute features into quantifiable numerical constraint vectors. The geographical coverage radius of the central warehouse is set to 500, the upper limit of inventory capacity for the regional distribution center is set to 8000, the transportation timeliness commitment for the main supplier's production base is set to 5, and the SLA in the supplier cooperation terms is broken down into three core values: fulfillment rate 95%, on-time delivery rate 98%, and product qualification rate 99%. All parametric values ​​corresponding to each entity are integrated into a set of exclusive numerical constraint vectors, and each entity node has a unique corresponding numerical constraint vector.

[0036] Step 2.22: Configure the AI ​​agent to establish a constraint mapping table, binding parameterized numerical constraint vectors one-to-one with entity nodes in the multilateral analysis framework to form an entity-constraint relationship. Specifically, the AI ​​agent first builds the basic data structure of the constraint mapping table, setting fixed columns such as unique identifiers for entity nodes, entity type, constraint feature type, numerical constraint vectors, and the corresponding business connotations of the vector values, ensuring the table structure can fully carry the association information between entities and constraints. Then, the AI ​​agent enters the unique identifiers of all supply chain entity nodes in the multilateral analysis framework into the corresponding columns of the constraint mapping table, and then binds the parameterized numerical constraint vectors to the entity nodes in the multilateral analysis framework. Numerical constraint vectors are precisely matched according to the unique identifier of the entity node. The numerical constraint vector of each entity is entered into the row corresponding to its identifier, realizing a one-to-one correspondence between the numerical constraint vector and the entity node. After the initial matching is completed, the AI ​​agent performs a full verification of the constraint mapping table to check whether the matching between the entity node and the numerical constraint vector is complete and without mismatch, and whether the vector value is consistent with the actual business boundary constraint of the entity. After the verification is passed, the constraint mapping table is established. The constraint mapping table realizes the firm binding between the numerical constraint vector and the entity node in the multilateral analysis framework, and based on this, a unique and traceable entity-constraint relationship is formed.

[0037] Step 2.23 transforms the entity-constraint relationships into joint constraint conditions within the multilateral analysis framework. Specifically, the geographical coverage radius corresponds to spatial movement range constraints, the inventory capacity limit corresponds to capacity threshold constraints, the transportation timeliness commitment corresponds to time dimension constraints, and the service level agreement corresponds to quality dimension constraints. This involves: using the entity-constraint relationships formed in the constraint mapping table as the core basis, first decomposing the numerical constraint vectors corresponding to each entity node according to constraint feature type to obtain single-dimensional quantitative constraint parameters; then, the AI ​​agent transforms the decomposed quantitative constraint parameters into identifiable and executable joint constraint conditions within the multilateral analysis framework, realizing the transformation of business constraints into framework technical constraints; the numerical parameter corresponding to the geographical coverage radius of the central warehouse serves as the entity's constraint condition within the multilateral analysis framework. The spatial movement range limit within the framework defines the entity's business coverage boundary in the geographic space dimension; the numerical parameter corresponding to the upper limit of the regional distribution center's inventory capacity serves as the entity's capacity threshold limit, defining the entity's business operation boundary in the inventory storage dimension; the numerical parameter corresponding to the transportation timeliness commitment of the main supplier's production base serves as the entity's time dimension limit, defining the entity's business performance boundary in the supply and delivery time dimension; the various numerical parameters after decomposing the service level agreement in the supplier cooperation terms serve as quality dimension limits, defining the entity's business standard boundaries in the quality dimensions such as supply quality and performance efficiency; after the transformation is completed, the AI ​​agent embeds the joint limit conditions of each dimension into the corresponding entity node position in the multilateral analysis framework, configuring exclusive multi-dimensional joint limit conditions for each entity node.

[0038] Step 2.24 involves conflict detection and coordination optimization of joint constraint conditions. When business boundary constraints of multiple entities overlap, the joint constraint parameters are automatically adjusted based on preset business priority weights and constraint relaxation strategies to obtain conflict-free joint constraint conditions. Specifically, the AI ​​agent first initiates a global conflict detection process for joint constraint conditions, performing cross-scanning and matching verification on four types of joint constraint conditions for all entity nodes in the multilateral analysis framework: spatial movement range, capacity threshold, time dimension, and quality dimension. The focus is on detecting whether there are issues such as overlapping geographical ranges, contradictory numerical requirements, or conflicting execution standards in the joint constraint conditions of different entity nodes in the same dimension. A conflict judgment threshold of 10 is set; when the overlap ratio of constraint conditions in the same dimension of two entities reaches 10 or higher, a constraint conflict is determined to exist. When a conflict in joint constraint conditions is detected, the AI ​​agent immediately retrieves the preset business priority weights and constraint relaxation strategies. Assuming the central warehouse has a business priority weight of 90, major supplier production bases have a weight of 80, and regional distribution centers have a weight of 70, the core principle for constraint adjustment is determined based on the business priority weights. Joint constraint conditions for high-priority entity nodes are retained first, while low-priority entity nodes undergo constraint relaxation strategies. The constraint relaxation strategy sets a base relaxation adjustment range of 5. The AI ​​agent adjusts the constraint parameters of low-priority entity nodes step-by-step based on the actual overlap ratio and degree of contradiction in constraint conflicts. After each adjustment, all joint constraint conditions are re-checked for conflict. If conflicts still exist, the adjustment continues according to the strategy until the check result is conflict-free. After coordinating and optimizing all conflicts, the AI ​​agent summarizes, verifies, and files all final joint constraint conditions to obtain conflict-free joint constraint conditions that meet the business requirements of the multilateral analysis framework. Simultaneously, the optimized constraint parameters are updated to the constraint mapping table to achieve data unification of entity and constraint relationships.

[0039] In this embodiment of the invention, by employing a driving AI agent to extract the business attribute features of each supply chain entity according to a multilateral analysis framework, parameterizing relevant business boundary constraints such as the geographical coverage radius of the central warehouse into numerical constraint vectors, configuring the AI ​​agent to establish a constraint mapping table to achieve a one-to-one correspondence between constraint vectors and entity nodes, transforming the association into four-dimensional joint constraint conditions of space, capacity, time, and quality, and through conflict detection and coordination optimization, combined with preset business priority weights and constraint relaxation strategies to adjust overlapping constraints and obtain conflict-free joint constraint conditions, this invention overcomes the technical problems in the prior art where the business boundary constraints of supply chain entities are not parameterized, cannot be accurately bound to entities within the framework, the joint constraint conditions are dimensional and do not meet the needs of entity business, and there are easy conflicts between constraints that cannot provide effective limits for data partitioning, thus affecting the accuracy of data partitioning. Therefore, this invention achieves the standardization and parameterization of the business boundary constraints of supply chain entities, realizes the accurate association between constraints and entity nodes, and constructs multi-dimensional joint constraint conditions that fit the actual business.

[0040] In a preferred embodiment of the present invention, step 3 above may include: Step 3.1: By dynamically adjusting parameters and a standardized dataset, the AI ​​agent is driven to quantify the importance of each data field in the standardized dataset based on the data density change rate index in the dynamically adjusted parameters. Field weight coefficients are dynamically assigned to obtain a weighted data field set. Specifically, the AI ​​agent first completes a two-way connection between the dynamically adjusted parameters and the standardized dataset, performs full-domain analysis on the dynamically adjusted parameters, extracts all numerical information of the data density change rate index, and obtains the supply chain business area and data field coverage corresponding to each index. Simultaneously, the standardized dataset is sorted by field classification, categorizing and labeling all data fields according to core supply chain business links such as procurement, production, warehousing, logistics, and distribution, obtaining the business affiliation and data connotation of each field. The AI ​​agent uses the data density change rate index as its core... Based on this, a quantitative assessment of the importance of each data field in the standardized dataset is conducted. The higher the data density change rate, the higher the business activity and data value of the corresponding field. The quantitative assessment results are presented on a percentage scale. Based on the assessment results, the AI ​​agent dynamically assigns field weight coefficients to each data field. The weight coefficients are set to a range of 1 to 10. Fields with a quantitative assessment result of 80 or above are assigned a weight coefficient of 9 to 10, fields with a result of 50 to 79 are assigned a weight coefficient of 5 to 8, and fields with a result of less than 50 are assigned a weight coefficient of 1 to 4. Each data field is matched with a unique weight coefficient that is appropriate to its importance. After the weight allocation is completed, the AI ​​agent integrates all data fields with weight coefficients, categorizes and organizes them according to business links, and verifies the rationality of the weight allocation to ensure that there are no mismatches or omissions, ultimately forming a weighted set of data fields.

[0041] Step 3.2: Based on the distribution dispersion features in the dynamically adjusted parameters, the weighted data field set is optimized for spatiotemporal aggregation granularity. According to the geographical distribution density and business operation frequency of business entities, the time window granularity and spatial region granularity of data aggregation are automatically adjusted to obtain a multi-granularity optimized dataset. Specifically, this includes: the AI ​​agent first extracts complete numerical information of the distribution dispersion features from the dynamically adjusted parameters to obtain the distribution dispersion values ​​corresponding to each key entity node in the supply chain. Simultaneously, it retrieves the weighted data field set and correlates the distribution dispersion features with the geographical distribution density and business operation frequency of each business entity. Based on this correlation information, the AI ​​agent performs spatiotemporal aggregation granularity optimization on the weighted data field set. Regarding the time window granularity adjustment, entities with a business operation frequency of once per hour or more have their corresponding... The time window granularity for data fields is set to 1 hour, 1 day for entities with daily business operations, and 7 days for entities with weekly business operations, achieving precise adaptation of the time window granularity according to the actual business operation frequency. Regarding spatial region granularity adjustment, regions with high geographical distribution density are divided into the smallest administrative region for the corresponding data fields, regions with medium geographical distribution density are divided into city-level regions, and regions with low geographical distribution density are divided into provincial-level regions, thus aligning with the geographical distribution characteristics of the entities to complete the spatial region granularity adjustment. The AI ​​agent automatically matches the corresponding time window granularity and spatial region granularity for each data field according to the above rules. After adjustment, the data fields are integrated and processed according to the new aggregation granularity, eliminating duplicate aggregated data records, supplementing missing aggregation dimension information, and finally obtaining a multi-granularity optimized dataset.

[0042] Step 3.3: The AI ​​agent is scheduled to perform data quality optimization processing on the multi-granularity optimized dataset. This involves outlier detection and correction, intelligent missing value imputation, data consistency verification, and noise filtering to eliminate outliers, missing items, and inconsistent records, resulting in a purified optimized dataset. Specifically, the AI ​​agent initiates the data quality optimization process, performing a full-domain scan and field-by-field verification of the multi-granularity optimized dataset. First, outlier detection and correction are performed, setting the outlier threshold to 3. Values ​​exceeding 3 are considered outliers by calculating the deviation between each data field's value and the mean of the same dimension. For outliers, the original data source is verified. If it's not a data collection error, it's corrected using the median of the same dimension; if it is a collection error, it's directly removed and the reason for the error is noted. Next, intelligent missing value imputation is performed, matching missing data items according to business relevance. The closest reference data is used to fill in single missing items based on the average value of the same business process, and consecutive missing items are filled by interpolation based on the historical trend of data changes. After all filling operations are completed, the basis for filling and the source of reference data are marked. Then, a data consistency verification operation is performed to check whether the values, units of measurement, and statistical calibers of the same business concept are consistent in different data fields. Inconsistent content is corrected in a unified manner according to the standard business terminology of the supply chain to ensure the consistency of data concepts. Finally, a noise filtering operation is performed to remove invalid and redundant data in the dataset, including duplicate collected null records, garbled data without actual business significance, and invalid values ​​that exceed the reasonable range of business. After completing the four optimization operations, the AI ​​agent performs a full quality review of the processed dataset to ensure that there are no outliers, no missing items, no inconsistent records, and no invalid noise data in the dataset, and finally obtains the purified optimized dataset.

[0043] Step 3.4 involves performing feature engineering on the purified and optimized dataset to extract temporal, spatial, and business-related features. Through feature selection and dimensionality reduction, a high-quality feature vector set with compressed dimensions and rich information is obtained. Specifically, the AI ​​agent performs full-dimensional feature engineering on the purified and optimized dataset. First, temporal features are extracted, analyzing the numerical change patterns, fluctuation trends, peak and trough occurrence times, and data growth or decline rates of each data field within different time windows, covering time-series information across multiple time dimensions such as minutes, hours, days, and weeks. Then, spatial features are extracted, combining the geographical distribution of key entities in the supply chain to analyze the data distribution characteristics in different spatial regions, cross-regional transmission characteristics, and inter-regional data correlation characteristics, aligning with the geographical layout characteristics of the supply chain. Finally, business-related features are extracted, analyzing... The data fields exhibit core business-related features such as correlation, collaboration, and causal characteristics among upstream and downstream business links, including procurement and production, production and warehousing, warehousing and logistics, and logistics and distribution, covering the interconnected information of the entire supply chain business process. After the three types of features are extracted, the AI ​​agent performs a feature selection operation, setting a feature importance threshold of 7. Redundant features with an importance lower than 7 are eliminated through quantitative evaluation, retaining only effective features that have core value for supply chain analysis. Next, a dimensionality reduction operation is performed, fusing the high-dimensional feature set to integrate multiple highly correlated features into a single comprehensive feature, reducing feature dimensions while fully preserving core data information. After feature selection and dimensionality reduction, the AI ​​agent classifies and integrates the processed feature set, labeling each feature with its corresponding feature type, data source, and importance score, ultimately forming a high-quality feature vector set that is both dimensionally compressed and information-rich.

[0044] In this embodiment of the invention, by employing a technique that drives an AI agent to quantitatively evaluate and dynamically allocate data field weights based on the rate of change of data density with dynamically adjusted parameters, optimizes the spatiotemporal aggregation granularity of data by combining distribution dispersion characteristics, performs outlier detection and correction, intelligent filling of missing values, and other data quality optimization processes on the multi-granularity optimized dataset, and finally extracts temporal, spatial, and business-related features through feature engineering and obtains a high-quality feature vector set through feature selection and dimensionality reduction, the invention overcomes the technical problems of existing technologies, such as the lack of dynamic quantitative basis for data field weights, fixed aggregation granularity that does not fit the actual spatiotemporal characteristics of the business, incomplete data quality optimization that is prone to anomalies and missing values, single feature extraction dimension and redundant data dimensions, and inability to provide a high-quality data foundation for subsequent risk analysis. This achieves dynamic and intelligent adjustment of data field weights and spatiotemporal aggregation granularity, improves the targeting and fit of data processing, effectively purifies data and improves the overall data quality, extracts multi-dimensional core business features and compresses data dimensions, resulting in a high-quality feature vector set that is rich in information and concise in dimensions.

[0045] In a preferred embodiment of the present invention, step 4 above may include: Step 4.1: Using a high-quality feature vector set, the AI ​​agent is driven to perform multi-dimensional correlation analysis on the temporal, spatial, and business-related features in the feature vector set. This calculates the feature similarity and dependency strength between different supply chain nodes, resulting in a multi-dimensional correlation network. Specifically, the AI ​​agent first receives the high-quality feature vector set, performs full-domain analysis and classification of the dataset, and extracts the temporal, spatial, and business-related features contained within it. This yields the corresponding supply chain business links, key entity nodes, and data connotations for each feature. Temporal features cover the fluctuation trend of data changes over time, the pattern of peaks and troughs, and the frequency of data updates for each entity node. Spatial features cover the geographical distribution of each entity, the pattern of cross-regional data transmission, and the differences in data distribution across different regions. Business-related features cover the interaction patterns of upstream and downstream business links, the performance relationship between entities, and the data linkage logic. After feature extraction and organization, the AI ​​agent drives multi-dimensional correlation analysis, performing cross-matching for each pair of features, focusing on analyzing the relationship between temporal and spatial features, and temporal features... The system identifies the intrinsic relationships between features and business-related characteristics, as well as between spatial features and business-related characteristics. It also calculates the feature similarity and dependency strength between different supply chain nodes. A feature similarity threshold of 80 is set; when the difference in the value of similar features between two nodes is less than 20%, indicating a similarity of 80 or higher, a strong correlation is considered. Dependency strength is quantified on a scale of 1 to 10. Higher frequency and closer data interaction between nodes result in a higher dependency strength. Nodes with interactions once per hour or more have a dependency strength of 8 to 10, those interacting once per day have a dependency strength of 5 to 7, and those interacting once per week or less have a dependency strength of 1 to 4. The AI ​​agent comprehensively statistically analyzes and classifies the feature similarity and dependency strength of all nodes. Nodes with relationships are integrated according to relationship type, similarity, and dependency strength to build a multi-dimensional relationship network including node identifiers, feature relationship types, similarity values, and dependency strength levels. The network clearly presents the intrinsic relationship logic between each supply chain node and performs preliminary verification of the relationships, eliminating false and invalid relationships to ensure the accuracy and completeness of the network.

[0046] Step 4.2, based on a multi-dimensional relational network, through directional verification and causal effect quantification of the relational relationships, and through intervention simulation and counterfactual reasoning, the direction and intensity of risk transmission from the supplier to the customer in supply chain operations are identified, resulting in a risk transmission path map. Specifically, this includes: the AI ​​agent retrieving and generating a multi-dimensional relational network, using this network as the core analytical foundation to conduct directional verification and causal effect quantification analysis of the relational relationships. In the directional verification stage, the AI ​​agent verifies the causal direction of the relational relationships by sorting out the sequential logic of the supply chain business processes and combining the temporal order of data at each node. For example, it determines whether abnormal supplier production data affects central warehouse inventory data, or whether abnormal central warehouse inventory data affects supplier production plans, eliminating invalid relationships such as reverse correlations and bidirectional fuzzy correlations, and obtaining the causal direction of each set of relational relationships. In the causal effect quantification stage, the AI ​​agent sets the causal effect quantification range to 0 to 100, and analyzes the linkage magnitude of data changes between causal nodes to determine the causal node data changes. The higher the ratio of the magnitude of change to the magnitude of change in the result node data, the larger the quantified value of the causal effect. Causal effects with a ratio of 80 or above are set to 80-100, those with a ratio of 50-79 are set to 50-79, and those below 50 are set to 0-49. Subsequently, the AI ​​agent further verifies the effectiveness of the causal relationship through intervention simulation and counterfactual reasoning. Intervention simulation involves artificially setting up abnormal data at a certain causal node and observing the data changes at the result node to verify the stability of the causal association. Counterfactual reasoning involves assuming that the causal node does not have abnormalities, inferring the normal data state of the result node, and comparing the difference between the actual data and the inferred data to further confirm the authenticity of the causal effect. Through a series of operations, the AI ​​agent accurately identifies the direction of risk transmission from the supplier to the customer in the supply chain operation, obtains the causal direction of each transmission path, quantifies the transmission strength of each path, integrates and sorts out all risk transmission paths, directions, and strengths, marks the key entity nodes and business links corresponding to each path, and finally generates a complete and clear risk transmission path map.

[0047] Step 4.3 involves identifying key nodes in the risk transmission path map, calculating the centrality index and vulnerability coefficient of each node in the risk transmission process, and combining business rule constraints and historical risk event data to screen out a set of key influencing factors that have a decisive impact on the overall stability of the supply chain. Specifically, the AI ​​agent uses the risk transmission path map as an analysis carrier to initiate the key node identification process. First, it classifies and sorts the supply chain entity nodes in the map, extracts the core node cluster consisting of the central warehouse and surrounding core suppliers, retrieves the geographical coordinate data of all nodes within the core node cluster, and uses the minimum circumcircle algorithm to perform spatial analysis on the core node cluster. The results are obtained through calculation. The smallest circumcircle of the core node cluster is obtained, along with its center position and radius. This radius represents the maximum radiation radius of the core cluster. Based on this radius, the AI ​​agent determines whether surrounding terminal nodes are within the spatial coverage area of ​​the core cluster. Nodes within the coverage area are marked as highly correlated nodes, while those outside the coverage area are marked as lowly correlated nodes. Simultaneously, the distance between each node and the center of the circumcircle is recorded, forming a spatial correlation weight table. Nodes closer to the center and within the coverage area have higher spatial correlation weights, and the probability and speed of risk transmission are also higher. Subsequently, the AI ​​agent analyzes all supply chain entity nodes in the graph one by one, focusing on calculating the centrality index and vulnerability coefficient of each node during the risk transmission process. The centrality index measures the coreness of a node in a risk transmission path. Its calculation combines spatial association weights obtained from the minimum circumcircle of the point set, comprehensively considering the number of times a node acts as an intermediate node, the number of connected transmission paths, and the weighting coefficients corresponding to the spatial association weights. The more connected paths, the more times a node acts as an intermediate node, and the higher the spatial association weight, the higher the centrality index value. A centrality index threshold of 75 is set; nodes with a value of 75 or higher are considered core candidate nodes. Nodes near the center of the minimum circumcircle of the core cluster are given an additional spatial core bonus to further improve their centrality index, ensuring that no spatial core nodes are overlooked. The vulnerability coefficient measures the vulnerability of a node after it experiences an anomaly. The impact on the overall risk transmission is calculated by referring to the core cluster coverage area defined by the smallest outer circle of the reference point set. The number of transmission path interruptions caused by node anomalies, the number of affected nodes, and the proportion of affected nodes within the core cluster coverage area are counted. The wider the impact area, the more transmission path interruptions, and the higher the proportion of affected nodes within the coverage area, the higher the vulnerability coefficient value. The vulnerability coefficient threshold is set at 65. Candidate nodes with a value of 65 or above are included in the critical node range. If a node is the center node of the smallest outer circle of the core cluster, such as the central warehouse, its anomaly causes the transmission path interruption of all related nodes within the coverage area. The vulnerability coefficient is calculated at the highest level and it is directly included in the critical node range.After initial screening, the AI ​​agent further optimizes the selection by combining supply chain business rules constraints with historical risk event data. Business rules prioritize hub-type physical nodes such as central warehouses and core supplier production bases. These hub-type physical nodes are typically the center of the smallest circumcircle of the core cluster or key components. Even if some hub nodes' indicators do not fully meet the thresholds, they are included in the candidate range of key nodes and re-verified. Historical risk event data involves retrieving past supply chain risk event records, analyzing nodes that frequently exhibited anomalies and significantly impacted business within the core cluster's coverage area, and matching these significantly impactful nodes with the currently selected candidate nodes. Nodes with risk impact that were not initially selected are added, while nodes with minor historical risk impact are removed. The AI ​​agent performs a final verification of all candidate nodes, confirming that each node has a decisive impact on the overall stability of the supply chain. For example, node anomalies may disrupt multiple risk transmission paths, causing operational anomalies in multiple physical nodes. Spatial analysis based on the smallest circumcircle of the point set shows that its impact range covers the core business area, ultimately selecting the set of core key influencing factors.

[0048] Step 4.4 involves a quantitative risk assessment of the set of key influencing factors. Based on the factors' impact scope, transmission speed, and recovery difficulty parameters, a comprehensive risk score is calculated, generating a risk assessment result that includes risk type, risk level, impact scope, and confidence level. Specifically, the AI ​​agent analyzes each of the selected key influencing factors, initiates the quantitative risk assessment process, and obtains the core dimensions of the assessment as impact scope, transmission speed, and recovery difficulty. Specific quantitative standards and scores are set for each dimension, with a total score of 100 points, and the three dimensions accounting for 40, 30, and 30 points respectively. Regarding the impact scope assessment, it is divided into three levels: large, medium, and small. Impact on all core entity nodes and nationwide business scope receives 30 to 40 points; impact on local entity nodes and regional business scope receives 15 to 29 points; and impact on only a single entity node and single business link receives 0 to 14 points. Regarding the transmission speed assessment, it is divided into three levels: fast, medium, and slow. Transmission across nodes within one day receives 22 to 30 points; transmission within three days receives 11 to 21 points; and transmission within seven days or more receives 40 to 50 points. The completion of the transmission process earns a score of 0 to 10. Regarding the assessment of recovery difficulty, it is divided into three levels: high, medium, and low. A recovery time of 7 days or more earns 22 to 30 points; a recovery time of 3 to 7 days earns 11 to 21 points; and a recovery time of 1 to 2 days earns 0 to 10 points. The AI ​​agent scores each key influencing factor according to the above quantitative standards, summing the scores from the three dimensions to obtain a comprehensive risk score. A risk level classification standard is set: a comprehensive score of 80 or above is high risk, 50 to 79 is medium risk, and below 50 is low risk. Each key influencing factor is matched with a corresponding risk level. At the same time, the AI ​​agent combines historical risk data with the current supply chain operation status to calculate the confidence level of each risk assessment result. For similar risks in historical data with a matching degree of 90 or above, the confidence level is set to 95; for matching degrees of 70 to 89, it is set to 80; and for matching degrees below 70, it is set to 60. Finally, the AI ​​agent integrates the risk type, risk level, scope of impact, and confidence level of all key influencing factors, classifies them by business link, and generates complete and standardized risk assessment results.

[0049] In this embodiment of the invention, by employing a technology that drives an AI agent to conduct multi-dimensional correlation analysis on the temporal, spatial, and business-related features of a high-quality feature vector set, calculates the feature similarity and dependency strength between supply chain nodes to obtain a multi-dimensional correlation network, correlates the directionality and quantifies the causal effect, combines intervention simulation and counterfactual reasoning to identify the direction and intensity of risk transmission and generate a risk transmission path map, identifies key nodes in the map and calculates centrality indicators and vulnerability coefficients, combines business rules and historical risk data to screen key influencing factors, and then quantifies the risk level of these factors according to their impact range, transmission speed, and recovery difficulty, calculates a comprehensive risk score, and generates a multi-dimensional risk assessment result, the invention overcomes the supply chain risk limitations of existing technologies. The analysis currently only focuses on superficial correlation analysis, failing to verify the directionality of the correlation and quantify causal effects. It is difficult to accurately identify the risk transmission path and intensity, and the selection of key influencing factors lacks quantitative basis. Risk assessment also lacks unified quantitative standards, and the results have a single dimension of information and insufficient reference value. Therefore, this approach breaks through the limitations of traditional correlation analysis, enabling causal and quantitative analysis of supply chain operational risks. It accurately uncovers the complete risk transmission path and its direction and intensity, selects core key influencing factors based on quantitative indicators, establishes a scientific quantitative assessment system for supply chain risk levels, and generates comprehensive risk assessment results that include risk type, level, scope of impact, and confidence level. This provides accurate, quantitative, and actionable decision-making basis for supply chain risk management.

[0050] In a preferred embodiment of the present invention, step 5 above may include: Step 5.1: Based on the risk assessment results, the AI ​​agent performs hierarchical structuring processing on the risk assessment results according to risk type, risk level, and impact scope, generating a three-layer data structure including a global overview layer, a regional analysis layer, and an entity details layer. Specifically, this includes: First, receiving the generated complete risk assessment results, performing a full-domain analysis and item-by-item sorting of the assessment results to obtain core information such as all risk types, risk levels, impact scope, and confidence levels. Simultaneously, it retrieves the attribution information of key entity nodes in the supply chain, business segment classification, and geographical distribution data to lay the foundation for hierarchical structuring processing. Then, the AI ​​agent classifies, splits, and structurally integrates the risk assessment results according to the three core dimensions of risk type, risk level, and impact scope, initiating the hierarchical processing flow. Regarding the construction of the global overview layer data, the AI ​​agent summarizes all risk assessment results, statistically analyzes the proportion of high, medium, and low risks in the overall supply chain, integrates the risk distribution overview of each core business segment, extracts the core risk characteristics of the overall supply chain, and forms a global overview layer data that intuitively reflects the overall risk status of the supply chain. Regarding the construction of the regional analysis layer data, the AI ​​agent divides the supply chain into regions... The risk assessment results are categorized by region, and the number of risks of different types and levels within each region is counted. Core risk nodes and risk propagation trends within each region are analyzed to form regional analysis layer data. For entity detail layer data construction, the AI ​​agent precisely binds the risk assessment results to specific entity nodes such as supplier production bases, central warehouses, and regional distribution centers. This yields the risk type, risk level, impact range, and confidence level for each entity node. Historical risk patterns and current risk status of entity nodes are analyzed to form entity detail layer data for individual entities. After the initial construction of the three layers of data, the AI ​​agent verifies each layer, checking the accuracy, completeness, and relevance of the data classification. This ensures that the global overview layer covers the risk overview of all regions and core business processes, the regional analysis layer accurately corresponds to the regional divisions of the global overview layer, and the entity detail layer fully supports the specific data of the regional analysis layer. After verification, the three layers of data are integrated hierarchically to obtain the data mapping relationships between layers, ultimately generating a three-layer data structure containing the global overview layer, regional analysis layer, and entity detail layer.

[0051] Step 5.2: Based on the three-layer data structure, construct a hierarchical interactive visualization dashboard. The global overview layer displays a heatmap of overall supply chain risk; the regional analysis layer presents a risk distribution matrix for distribution centers in each region; and the entity details layer displays risk indicator trend charts for specific suppliers and warehouses. Drill-down interactive controls are set for each layer. Specifically, this includes: using the generated three-layer data structure as the core data source, initiating the construction process of the hierarchical interactive visualization dashboard; first, building the basic architecture of the dashboard, obtaining the one-to-one correspondence between the three-layer data structure and each layer of the dashboard; setting the overall display style and interaction logic of the dashboard; ensuring that the displayed content aligns with the actual needs of supply chain operation and management, while also adapting to different terminal display scenarios; in the global overview... In terms of layered dashboard construction, the AI ​​agent generates a heatmap of overall supply chain risk based on global overview layer data. Different colors are used to distinguish different risk levels: red represents high-risk areas, yellow represents medium-risk areas, and green represents low-risk areas. The heatmap clearly marks the risk distribution density and core risk types of each area, while also indicating the overall supply chain risk score and high-risk links, allowing users to quickly grasp the overall risk status of the supply chain. Regarding the construction of regional analysis layer dashboards, the AI ​​agent generates a risk distribution matrix for each region based on regional analysis layer data. The matrix's rows label the core entity nodes within the region, and the columns label the risk type and risk level. Matrix cells also label the specific details of the corresponding entity nodes. The system provides risk values ​​and confidence levels, along with a simplified diagram of risk transmission paths within the region, clearly presenting the risk relationships between various entity nodes. Regarding the construction of the entity detail layer dashboard, the AI ​​agent generates exclusive risk indicator trend charts for each supplier's production base, central warehouse, and regional distribution center based on the entity detail layer data. These trend charts display recent risk indicator changes for each entity node over time, marking peaks, troughs, and abnormal fluctuations, and include detailed risk descriptions for each entity node, including the initial cause of the risk, its current impact range, and confidence level. After completing the construction of the three-layer dashboard content, the AI ​​agent sets up drill-down interactive controls for each layer, including interactive controls for the global overview layer. The controls support clicking on a specific area to directly drill down to the corresponding area analysis layer dashboard. The interactive controls in the area analysis layer support clicking on a specific entity node to directly drill down to the corresponding entity details layer dashboard. Reverse drill-down is also supported, allowing users to return from the entity details layer to the area analysis layer and from the area analysis layer to the global overview layer, enabling seamless switching between dashboard levels. The AI ​​agent comprehensively tests the dashboard's interactive smoothness and display accuracy, optimizing dashboard loading speed and drill-down response speed, correcting display errors and interactive anomalies, ensuring that the layered interactive visual dashboard clearly, intuitively, and conveniently displays risk information at each level of the supply chain. This addresses the issues of lack of interactivity and unintuitive risk display in existing static visual reports.

[0052] Step 5.3 involves real-time monitoring of risk level indicators in the hierarchical interactive visualization dashboard. When any risk indicator exceeds a preset threshold, the AI ​​agent is driven to determine the warning level and trigger the corresponding tiered warning mechanism based on the severity, scope of impact, and transmission speed of the risk. Specifically, this includes: initiating the real-time monitoring process of the hierarchical interactive visualization dashboard, setting the monitoring frequency to once per minute, continuously scanning and collecting data on all risk level indicators at each level of the dashboard, capturing real-time changes and fluctuations in the values ​​of the risk indicators, and simultaneously retrieving preset risk indicator thresholds to obtain the warning thresholds corresponding to each risk type. The high-risk warning threshold is set to 80, and the medium-risk warning threshold is set to 80. The initial warning threshold is set to 50, and the low-risk warning threshold is set to 30. When the value of any risk indicator exceeds the corresponding warning threshold, the monitoring module immediately sends a trigger signal to notify the AI ​​agent to initiate the tiered warning processing procedure. After receiving the trigger signal, the AI ​​agent immediately retrieves the complete risk assessment information corresponding to the risk indicator that exceeds the threshold, including the severity of the risk level, the breadth of the impact range, and the risk transmission speed. Combined with the preset warning level classification rules, it determines the corresponding warning level. The warning levels are divided into four levels: Level 1 warning corresponds to a high-risk level, the impact range covers the entire area, and the transmission speed is within 1 day; Level 2 warning corresponds to a high-risk level, the impact range is a local area, and the transmission speed is 1 to 30 days. The alert period is 3 days. Level 3 alerts correspond to medium risk, affecting a localized area, with a transmission speed of 3 to 7 days. Level 4 alerts correspond to low risk, affecting a single entity, with a transmission speed of 7 days or more. Once the alert level is determined, the AI ​​agent triggers the corresponding tiered alert mechanism. Level 1 alerts use pop-up notifications, sound alarms, and simultaneous email and SMS notifications to ensure management and relevant core business personnel receive the alert immediately. Level 2 alerts use pop-up notifications and email notifications to notify regional managers and relevant business personnel. Level 3 alerts use email notifications to notify business specialists in the corresponding region. Level 4 alerts use in-system message notifications to notify the corresponding entity. The node's responsible person; simultaneously, the AI ​​agent, at the corresponding level of the hierarchical interactive visual dashboard, specially marks risk indicators exceeding the threshold with a flashing red border, highlighting them and indicating the warning level and trigger time, facilitating users to quickly locate abnormal risk nodes; the AI ​​agent continuously monitors changes in risk indicators exceeding the threshold, updating the warning status in real time. If the risk indicator falls back below the warning threshold, the corresponding warning is immediately lifted, and relevant personnel are notified of the warning cancellation; if the risk indicator continues to rise, the warning level is upgraded, and a risk escalation reminder is pushed out, ensuring the timeliness and relevance of risk warnings, and solving the problems of general, non-tiered, and untimely response in existing technologies.

[0053] Step 5.4. Based on the triggered tiered early warning level and the corresponding risk transmission path, automatically generate a response report including risk root cause analysis, impact scope prediction, and handling suggestions. This report is then pushed to relevant business personnel and management via email and SMS. Specifically, after triggering the tiered early warning mechanism, the AI ​​agent immediately retrieves the corresponding early warning level information and the complete risk transmission path map. Combined with the risk assessment results, it initiates the automatic response report generation process. In the risk root cause analysis phase, the AI ​​agent, based on the risk transmission path map, traces the core root cause of the risk, starting from the initial node of the risk transmission path, and analyzes the specific reasons for the risk at that node, including data anomalies, business operation deviations, and entity operation problems. Simultaneously, by combining historical risk event data and comparing the root causes of similar risks, the accuracy of the risk root causes is confirmed, ensuring that the root cause analysis is realistic, precise, and specific. For example, if the risk is a shortage of inventory in a regional distribution center, the root cause analysis needs to determine whether it is caused by a delay in allocation from the central warehouse or by untimely delivery from suppliers, while also explaining the specific manifestations of the delay or untimely delivery. In the impact prediction stage, the AI ​​agent, based on the current impact range, transmission speed, and risk level of the risk, combined with the correlation between various entity nodes in the supply chain, predicts the spread range of the risk in the future, the number of entity nodes affected, and the specific degree of impact on each business link. The prediction period is set at 7 days, obtaining the risk spread nodes and impact range every day, providing data for subsequent handling work. Prediction Basis: In the response recommendation stage, the AI ​​agent combines the root causes of risks, the predicted scope of impact, and the actual situation of supply chain operations to generate targeted and actionable response recommendations. These recommendations are prioritized, with core recommendations addressing the root causes of risks first, and auxiliary recommendations addressing the spread of risks. For example, in response to a high-risk warning caused by insufficient supplier capacity, the core recommendation is to coordinate with backup suppliers to supplement capacity, while the auxiliary recommendation is to adjust the central warehouse inventory allocation plan to prioritize supply to core areas. The system also provides the implementing entity, implementation steps, and completion deadlines for each recommendation, ensuring direct implementation. After the report content is written, the AI ​​agent standardizes the report format, categorizing it into risk root cause analysis, impact prediction, response recommendations, and risks. The report is organized in order of alert level and confidence level to ensure a clear structure, complete content, and standardized language. It also verifies the accuracy and reasonableness of the report content, correcting any phrasing errors and logical inconsistencies. Subsequently, the AI ​​agent initiates a multi-channel push process, determining the target audience and push channel based on the alert level. Level 1 alerts are pushed to management and core business leaders via both email and SMS. Level 2 alerts are pushed to regional managers and relevant business personnel via email. Level 3 and 4 alerts are pushed to the corresponding business specialists and entity leaders via email or internal system messages. The push content includes the complete response report and an alert level reminder, while recording the push time, target audience, and push status to ensure all relevant personnel receive the response report promptly.

[0054] In this embodiment of the invention, because an AI agent is used to perform hierarchical and structured processing of risk assessment results according to risk type, level, and scope of impact, generating a three-layer data structure of global overview, regional analysis, and entity details, a hierarchical interactive visualization dashboard with a dedicated display format for each layer and drill-down interactive controls is constructed based on this structure. The dashboard's risk indicators are monitored in real time, and warning levels are determined and tiered warnings are triggered according to the severity of the risk level, the breadth of the impact, and the transmission speed. Furthermore, by combining the warning level and the risk transmission path, a response report containing risk root cause analysis, impact scope prediction, and handling suggestions is automatically generated and pushed through multiple channels such as email and SMS. Therefore, this overcomes the limitations of existing technologies where supply chain risk results display lacks hierarchical design, has a single visualization format, and lacks interactivity. The lack of a tiered early warning mechanism, one-sided triggering criteria, and the absence of targeted response reports and limited distribution channels after risks exceed thresholds have led to technical problems such as inconvenient access to risk results, insufficient accuracy of early warnings, slow risk response, and a lack of scientific decision-making references. This solution addresses these issues by enabling a tiered, visualized, and interactive display of risk assessment results. It aims to meet the differentiated viewing needs of business personnel at different levels, making risk distribution and trends more intuitive and verifiable. The goal is to establish a scientific supply chain risk tiered early warning mechanism, improve the accuracy and targeting of risk warnings, automate the generation of risk response reports, and deliver precise information through multiple channels. This provides scientific and specific decision-making basis for risk management, improves the efficiency of risk response and handling, and creates a complete closed loop for supply chain risk management, from result display and risk warnings to handling recommendations.

[0055] like Figure 2 As shown, embodiments of the present invention also provide a visualization data analysis system based on an AI intelligent agent, comprising: The acquisition module is used to autonomously initiate multi-source data collection tasks through AI agents to acquire multi-source heterogeneous supply chain operation data; based on the multi-source heterogeneous supply chain operation data, a dynamic semantic conflict detection and resolution mechanism is constructed through AI agents to identify and unify the data semantics of different business concepts and form a standardized dataset. The computation module is used to construct a multilateral analysis framework based on a standardized dataset, using AI agents with central warehouses, regional distribution centers, and major supplier production bases in the supply chain as key entities. By setting the business boundary constraints of the supply chain entities as joint limit conditions, the module solves the final partitioning scheme of the data within the multilateral analysis framework under the business boundary constraints. The module partitions the business scope associated with the multilateral analysis framework into multiple feature evaluation blocks, and obtains a dynamically adjusted parameter based on the data density and distribution pattern within each block. The adjustment module is used to dynamically adjust the weights and aggregation granularity of data fields in the standardized dataset to obtain an optimized dataset. The AI ​​agent then performs data quality optimization to obtain a high-quality feature vector set. The assessment module is used to identify risk transmission paths and key influencing factors in supply chain operations by performing multi-dimensional correlation analysis and causal reasoning based on high-quality feature vector sets and AI agents, thereby obtaining risk assessment results. The processing module is used to generate hierarchical interactive visualization dashboards based on the assessment results, and to trigger tiered early warnings and push response reports when the risk exceeds the threshold.

[0056] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A visualization data analysis method based on AI intelligent agents, characterized in that, The method includes: Step 1: The AI ​​agent autonomously initiates a multi-source data collection task to obtain multi-source heterogeneous supply chain operation data; based on the multi-source heterogeneous supply chain operation data, the AI ​​agent constructs a dynamic semantic conflict detection and resolution mechanism to identify and unify the data semantics of different business concepts, forming a standardized dataset. Step 2: Based on a standardized dataset, a multilateral analysis framework is constructed using an AI agent, with the central warehouse, regional distribution centers, and major supplier production bases in the supply chain as key entities. By setting the business boundary constraints of the supply chain entities as joint limiting conditions, the final partitioning scheme of the data within the multilateral analysis framework is solved under the business boundary constraints. The business scope associated with the multilateral analysis framework is partitioned into multiple feature evaluation blocks, and a dynamic adjustment parameter is obtained based on the data density and distribution pattern within each block. Step 3: Adjust the data field weights and aggregation granularity of the standardized dataset by dynamically adjusting parameters to obtain the optimized dataset. Then, perform data quality optimization processing through an AI agent to obtain a high-quality feature vector set. Step 4: Based on a high-quality feature vector set, conduct multi-dimensional correlation analysis and causal reasoning through an AI agent to identify risk transmission paths and key influencing factors in supply chain operations, and obtain risk assessment results. Step 5: Generate a hierarchical interactive visualization dashboard based on the assessment results, and trigger a tiered early warning and push response report when the risk exceeds the threshold.

2. The visualization data analysis method based on AI intelligent agents according to claim 1, characterized in that, By autonomously initiating multi-source data collection tasks through AI agents, multi-source heterogeneous supply chain operation data is acquired. Based on this multi-source heterogeneous supply chain operation data, a dynamic semantic conflict detection and resolution mechanism is constructed using AI agents to identify and unify the data semantics of different business concepts, forming a standardized dataset, including: The AI ​​agent is scheduled to autonomously detect and plan tasks for multi-source heterogeneous supply chain operation data sources, and generate multi-source data acquisition task instructions that include data source location, acquisition frequency and data format. Based on task instructions, the AI ​​agent acquires structured and unstructured supply chain operation data in real time from enterprise resource planning, logistics tracking platforms, supplier management platforms, and IoT sensor networks. Based on multi-source heterogeneous supply chain operation data, the AI ​​agent is driven to build an ontology mapping mechanism. Through semantic similarity calculation and business rule verification, semantic conflicts of the same business concepts in different data sources are dynamically detected. When a semantic conflict is detected, the AI ​​agent automatically performs semantic resolution operations based on a preset business priority strategy and contextual analysis, mapping the conflicting business concepts to a unified business terminology standard to obtain semantically resolved data; the semantically resolved data is then integrated, and data format standardization conversion and metadata registration are performed to obtain a standardized dataset.

3. The visualization data analysis method based on AI intelligent agents according to claim 2, characterized in that, Step 2 above includes: Receive standardized datasets and drive AI agents to identify central warehouses, regional distribution centers and major supplier production bases in the supply chain network as key entity nodes, and build a multilateral analysis framework that includes physical connections and business interactions between entities. The configuration of AI agents parameterizes the business boundary constraints of each supply chain entity, setting geographical coverage, inventory capacity thresholds, transportation timeliness requirements, and supplier cooperation terms as joint constraint conditions in the multilateral analysis framework. Under the constraints of joint positioning, the AI ​​agent spatially segments the business scope associated with the multilateral analysis framework and iteratively optimizes the final partitioning scheme of the computational data partition to obtain multiple independent feature evaluation blocks with complete business semantics. The system monitors the real-time data flow within each feature evaluation block, drives the AI ​​agent to calculate the data density index and distribution pattern characteristics of each block, and obtains dynamically adjusted parameters based on a weighted combination of the data density change rate and distribution dispersion.

4. The visualization data analysis method based on AI intelligent agents according to claim 3, characterized in that, The configuration of the AI ​​agent parameterizes the business boundary constraints of each supply chain entity, setting geographical coverage, inventory capacity thresholds, transportation timeliness requirements, and supplier cooperation terms as joint constraint conditions in the multilateral analysis framework, including: Based on the constructed multilateral analysis framework, the AI ​​agent is driven to extract the business attribute characteristics of each supply chain entity, and parameterize the geographical coverage radius of the central warehouse, the upper limit of the inventory capacity of the regional distribution center, the transportation time commitment of the main supplier's production base, and the service level agreement in the supplier cooperation terms into a numerical constraint vector. Configure the AI ​​agent to establish a constraint mapping table, and bind the parameterized numerical constraint vectors to the entity nodes in the multilateral analysis framework one-to-one to form an entity-constraint relationship; The relationship between entities and constraints is transformed into joint constraints in a multilateral analysis framework, where the geographical coverage radius corresponds to the spatial movement range constraint, the upper limit of inventory capacity corresponds to the capacity threshold constraint, the transportation timeliness commitment corresponds to the time dimension constraint, and the service level agreement corresponds to the quality dimension constraint. By performing conflict detection and coordination optimization on joint limit conditions, when the business boundary constraints of multiple entities overlap, the joint limit parameters are automatically adjusted based on the preset business priority weights and constraint relaxation strategies to obtain conflict-free joint limit conditions.

5. The visualization data analysis method based on AI intelligent agents according to claim 4, characterized in that, By dynamically adjusting the parameters to modify the weights and aggregation granularity of the data fields in the standardized dataset, an optimized dataset is obtained. Then, an AI agent performs data quality optimization to produce a high-quality feature vector set, including: By dynamically adjusting parameters and a standardized dataset, the AI ​​agent is driven to quantify the importance of each data field in the standardized dataset based on the data density change rate index in the dynamically adjusted parameters, and dynamically allocate field weight coefficients to obtain a weighted set of data fields. Based on the distribution dispersion characteristics in the dynamically adjusted parameters, the spatiotemporal dimension aggregation granularity of the weighted data field set is optimized. According to the geographical distribution density of business entities and the frequency of business operations, the time window granularity and spatial region granularity of data aggregation are automatically adjusted to obtain a multi-granularity optimized dataset. The AI ​​agent is scheduled to perform data quality optimization processing on the multi-granularity optimized dataset. Through outlier detection and correction, intelligent filling of missing values, data consistency verification and noise filtering, outliers, missing items and inconsistent records in the data are eliminated to obtain a cleaned optimized dataset. By performing feature engineering on the purified and optimized dataset, temporal features, spatial features, and business-related features are extracted. Through feature selection and dimensionality reduction operations, a high-quality feature vector set with compressed dimensions and rich information is obtained.

6. The visualization data analysis method based on AI intelligent agents according to claim 5, characterized in that, Based on a high-quality feature vector set, multi-dimensional correlation analysis and causal reasoning are performed using an AI agent to identify risk transmission paths and key influencing factors in supply chain operations, resulting in risk assessment results, including: By using high-quality feature vector sets, the AI ​​agent is driven to perform multi-dimensional correlation analysis on the temporal features, spatial features and business-related features in the feature vector sets, calculate the feature similarity and dependency strength between different supply chain nodes, and obtain a multi-dimensional correlation network. Based on a multi-dimensional network of relationships, this study identifies the direction and intensity of risk transmission from suppliers to customers in supply chain operations by verifying the direction of relationships and quantifying causal effects, and by using intervention simulation and counterfactual reasoning, thus obtaining a risk transmission path map. By identifying key nodes in the risk transmission path map, calculating the centrality index and vulnerability coefficient of each node in the risk transmission process, and combining business rule constraints and historical risk event data, a set of key influencing factors that have a decisive impact on the overall stability of the supply chain is selected. By quantitatively assessing the risk level of a set of key influencing factors, a comprehensive risk score is calculated based on parameters such as the factor's impact range, transmission speed, and recovery difficulty, generating a risk assessment result that includes risk type, risk level, impact range, and confidence level.

7. The visualization data analysis method based on AI intelligent agents according to claim 6, characterized in that, Step 5: Generate a tiered interactive visualization dashboard based on the assessment results, and trigger tiered alerts and push response reports when risks exceed thresholds, including: Based on the risk assessment results, the AI ​​agent is driven to perform hierarchical and structured processing of the risk assessment results according to the risk type, risk level and impact scope, generating a three-layer data structure including a global overview layer, a regional analysis layer and an entity details layer; Based on a three-layer data structure, a hierarchical interactive visualization dashboard is constructed. The global overview layer displays a heat map of overall supply chain risk, the regional analysis layer presents a risk distribution matrix of distribution centers in each region, and the entity details layer displays risk indicator trend charts of specific suppliers and warehouses. Drill-down interactive controls are set for each layer. The system monitors risk level indicators in a hierarchical interactive visualization dashboard in real time. When any risk indicator exceeds a preset threshold, the AI ​​agent is driven to determine the warning level and trigger the corresponding graded warning mechanism based on the severity of the risk, the breadth of its impact, and the speed of its transmission. Based on the triggered tiered early warning level and the corresponding risk transmission path, a response report is automatically generated, which includes risk root cause analysis, impact range prediction, and handling suggestions. The response report is then pushed to relevant business personnel and management through email and SMS message channels.

8. A visualization data analysis system based on AI intelligent agents, the system implementing the method as described in any one of claims 1 to 7, characterized in that, include: The acquisition module is used to autonomously initiate multi-source data collection tasks through AI agents to acquire multi-source heterogeneous supply chain operation data; based on the multi-source heterogeneous supply chain operation data, a dynamic semantic conflict detection and resolution mechanism is constructed through AI agents to identify and unify the data semantics of different business concepts and form a standardized dataset. The computation module is used to construct a multilateral analysis framework based on a standardized dataset, using AI agents with central warehouses, regional distribution centers, and major supplier production bases in the supply chain as key entities. By setting the business boundary constraints of the supply chain entities as joint limit conditions, the module solves the final partitioning scheme of the data within the multilateral analysis framework under the business boundary constraints. The module partitions the business scope associated with the multilateral analysis framework into multiple feature evaluation blocks, and obtains a dynamically adjusted parameter based on the data density and distribution pattern within each block. The adjustment module is used to dynamically adjust the weights and aggregation granularity of data fields in the standardized dataset to obtain an optimized dataset. The AI ​​agent then performs data quality optimization to obtain a high-quality feature vector set. The assessment module is used to identify risk transmission paths and key influencing factors in supply chain operations by performing multi-dimensional correlation analysis and causal reasoning based on high-quality feature vector sets and AI agents, thereby obtaining risk assessment results. The processing module is used to generate hierarchical interactive visualization dashboards based on the assessment results, and to trigger tiered early warnings and push response reports when the risk exceeds the threshold.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images