An event analysis method, device, terminal device, and program product

Abstract

The application is suitable for the technical field of intelligent investment research, and provides an event analysis method, device, terminal equipment and program product. The method comprises the following steps: determining a target ontology object in a to-be-analyzed event; determining a plurality of target associated objects matched with the target ontology object and a conduction degree number of each target associated object according to preset supply chain relationship information. The supply chain relationship information is used for representing a plurality of candidate associated objects having a supply-demand relationship with each candidate ontology object in a supply chain universe and a conduction degree number of each candidate associated object. The conduction degree number is used for quantifying a supply-demand relationship level between each candidate ontology object and the candidate associated object. Event prediction information of the to-be-analyzed event is generated. The method generates the event prediction information of the to-be-analyzed event according to the supply chain relationship information, provides a structured constraint for event analysis, and improves the accuracy of the event prediction information.

Description

Technical Field

[0001] This application belongs to the field of intelligent investment research technology, and in particular relates to an event analysis method, device, terminal equipment and program product. Background Technology

[0002] It can be argued that in scenarios such as industrial investment research, supply chain risk management, and emergency response to sudden events, when a key event occurs (such as "a company's technological breakthrough", "a natural disaster in a certain region", or "the introduction of a policy for a certain product"), it is necessary to quickly analyze the impact of the event on the industrial chain.

[0003] However, existing technologies mainly perform event analysis by directly inputting event descriptions into large language models. This is prone to creating illusions based on outdated information in the training data, such as unreasonable inferences about the event impact paths, which seriously reduces the accuracy of the analysis results. Summary of the Invention

[0004] This application provides an event analysis method, apparatus, terminal device, and program product, which can provide structured constraints for event analysis and improve the accuracy of event prediction information.

[0005] In a first aspect, embodiments of this application provide an event analysis method, including: In response to an event analysis request, determine the target ontology object in the event to be analyzed; Based on the preset supply chain relationship information, multiple target associated objects that match the target ontology object are identified, as well as the transmission degree of each target associated object. The supply chain relationship information is used to characterize multiple candidate associated objects that have supply and demand relationships with each candidate ontology object in the entire supply chain domain, as well as the transmission degree of each candidate associated object. The transmission degree is used to quantify the supply and demand relationship level between each candidate ontology object and the candidate associated object. Based on the target ontology object, multiple target associated objects, and the transmission degree of each target associated object, event prediction information for the event to be analyzed is generated. The event prediction information is used to characterize the transmission path between the target ontology object and multiple target associated objects affected by the event to be analyzed.

[0006] In one possible implementation of the first aspect, based on preset supply chain relationship information, multiple target associated objects matching the target ontology object are determined, along with the transitivity degree of each target associated object, including: Obtain pre-defined supply chain relationship information; Based on supply chain relationship information, multi-hop transit query is performed on the target ontology object to obtain multiple preliminary screening related objects corresponding to the target ontology object, as well as the transit degree of each preliminary screening related object; Multiple initially screened related objects are deduplicated to obtain multiple target related objects that match the target ontology object.

[0007] In one possible implementation of the first aspect, the event prediction information includes event propagation data and predicted impact level. Based on the target ontology object, multiple target associated objects, and the propagation degree of each target associated object, event prediction information for the event to be analyzed is generated, including: Based on the target ontology object, multiple target associated objects, and the transitivity degree of each target associated object, generate event transitivity data for the event to be analyzed; By inputting the number of objects and event propagation data of multiple target related objects into the large language model, the predicted impact level of the event to be analyzed can be obtained.

[0008] In one possible implementation of the first aspect, event propagation data for the event to be analyzed is generated based on the target ontology object, multiple target associated objects, and the propagation degree of each target associated object, including: For each target associated object, the object identifier and transitivity of each target associated object, as well as the object identifier of the target ontology object, are structured to obtain the path node data of each target associated object; Sort the path node data of all target-related objects according to the degree of transmission to generate event transmission data for the event to be analyzed.

[0009] In one possible implementation of the first aspect, after inputting the number of objects and event propagation data of multiple target related objects into a large language model to obtain the predicted impact level of the event to be analyzed, the method includes: Based on the event to be analyzed, event transmission data, and predicted impact level, an impact analysis report of the event to be analyzed is generated.

[0010] In one possible implementation of the first aspect, after inputting the number of objects and event propagation data of multiple target related objects into a large language model to obtain the predicted impact level of the event to be analyzed, the method includes: In response to the reviewer's request for correction, determine the actual impact level and the data on the reasons for the correction carried in the request; The predicted impact level of the event to be analyzed is adjusted based on the actual impact level. Based on the predicted impact level, the actual impact level, and the reasons for the correction, the parameters of the large language model are adjusted to obtain the adjusted model. The adjusted model is used to generate the predicted impact level of the new event to be analyzed.

[0011] In one possible implementation of the first aspect, determining the target ontology object in the event to be analyzed includes: A large language model is used to extract objects from the event to be analyzed, thereby obtaining the target ontology object in the event to be analyzed.

[0012] Secondly, embodiments of this application provide an event analysis apparatus, including: The first determination module is used to determine the target ontology object in the event to be analyzed in response to the event analysis request; The second determination module is used to determine multiple target associated objects that match the target ontology object and the transmission degree of each target associated object based on the preset supply chain relationship information. The supply chain relationship information is used to characterize multiple candidate associated objects that have supply and demand relationships with each candidate ontology object in the entire supply chain domain, and the transmission degree of each candidate associated object. The transmission degree is used to quantify the supply and demand relationship level between each candidate ontology object and the candidate associated object. The generation module is used to generate event prediction information for the event to be analyzed based on the target ontology object, multiple target associated objects, and the transmission degree of each target associated object. The event prediction information is used to characterize the transmission path between the target ontology object and multiple target associated objects affected by the event to be analyzed.

[0013] Thirdly, embodiments of this application provide a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method as described in any of the first aspects.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method as described in any of the first aspects.

[0015] Fifthly, embodiments of this application provide a computer program product that, when run on a terminal device, causes the terminal device to execute any of the methods described in the first aspect above.

[0016] This application provides an event analysis method, apparatus, terminal device, and program product. The method includes: responding to an event analysis request, determining a target ontology object in the event to be analyzed; determining multiple target associated objects matching the target ontology object, and the transmission degree of each target associated object, based on preset supply chain relationship information, wherein the supply chain relationship information is used to characterize multiple candidate associated objects with supply and demand relationships with each candidate ontology object in the entire supply chain, and the transmission degree of each candidate associated object, the transmission degree being used to quantify the supply and demand relationship hierarchy between each candidate ontology object and the candidate associated objects; generating event prediction information for the event to be analyzed based on the target ontology object, the multiple target associated objects, and the transmission degree of each target associated object, the event prediction information being used to characterize the transmission path between the target ontology object affected by the event to be analyzed and the multiple target associated objects. By using the above technical solution, generating event prediction information for the event to be analyzed based on supply chain relationship information provides structured constraints for event analysis and improves the accuracy of event prediction information. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating an event analysis method provided in an embodiment of this application; Figure 2 This is a flowchart illustrating an event analysis method provided in another embodiment of this application; Figure 3 This is a schematic diagram of the overall architecture of an event analysis method provided in an embodiment of this application; Figure 4 This is a structural block diagram of an event analysis device provided in one embodiment of this application; Figure 5 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. Detailed Implementation

[0019] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0020] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0021] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0022] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0023] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0024] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0025] It can be assumed that in scenarios such as industrial investment research, supply chain risk management, and emergency response to sudden events, when a key event occurs (such as "a company's technological breakthrough", "a natural disaster in a certain region", or "the introduction of a policy for a certain product"), it is necessary to quickly analyze which companies in the industrial chain will be affected by the event, what the transmission path of the impact is, and the overall importance rating of the event.

[0026] On the one hand, existing event analysis methods require analysts to manually query the company's knowledge graph, view the upstream and downstream relationships of the company level by level through visualization tools, and then manually determine the scope and transmission path of the event's impact. However, such methods require manual querying and judgment each time, with a response time of up to several hours, resulting in low efficiency of manual analysis. At the same time, different analysts have inconsistent rating standards for the same event, lacking objective basis, leading to a high degree of subjectivity in importance rating. Furthermore, since the impact of an event often transmits through multiple levels of the supply chain (e.g., A->B->C->D), multi-hop transmission reasoning is quite complex, making it difficult for manual analysis to fully track multi-level transmission paths, especially indirect impacts above level 3, leading to inaccurate analysis results.

[0027] On the other hand, existing event analysis methods generate impact analysis reports (e.g., supplier production stoppage -> downstream enterprise cost increase) by matching predefined rule bases or directly inputting event descriptions into large language models. However, the rule bases have limited coverage, failing to encompass the impact patterns of all event types. This results in an inability to handle complex and novel event scenarios, leading to poor flexibility. Furthermore, these methods require continuous rule base updates to adapt to new industry trends, resulting in high maintenance costs. They also lack automated importance rating mechanisms, making it impossible to prioritize events. Large language models, on the other hand, are prone to misinterpretations based on outdated information in training data, lacking structured knowledge support and the constraints of industry graphs. This leads to illogical impact paths and compromises the logical consistency of reasoning. The analysis results are not stored as structured data, making them unusable for subsequent model training and optimization, thus preventing the formation of a data loop. Finally, the lack of structured impact path storage makes traceability and manual verification difficult.

[0028] Based on this, the event analysis method provided in this application provides an automatic inference of the industry chain impact path of an event through a complete process of entity extraction -> graph query -> multi-hop propagation reasoning -> importance rating -> analysis report generation, thereby realizing automated and structured analysis of the event impact.

[0029] Figure 1 This is a flowchart illustrating an event analysis method according to an embodiment of this application. It is intended as an example and not a limitation; the method can be applied to terminal devices, such as... Figure 1 As shown, the method includes: S101. In response to the event analysis request, determine the target ontology object in the event to be analyzed.

[0030] The event to be analyzed can refer to a specific event that requires an impact assessment, such as "a technological breakthrough by a company," "a natural disaster in a region," or "the introduction of a policy for a product." The target ontology can refer to an entity directly involved in or serving as a core subject in the event to be analyzed, such as the company mentioned in the event.

[0031] This step aims to identify core entities from complex event descriptions, thus initially narrowing down the scope of analysis and laying the foundation for subsequent correlation analysis. The specific method for identifying target ontology objects is not limited. For example, when receiving an event analysis request regarding "a certain automotive chip supplier ceasing production," one could manually read the event text and extract "a certain automotive chip supplier" as the target ontology object. Alternatively, a keyword list could be pre-defined, and potential ontology objects could be identified by matching keywords in the event text.

[0032] In some embodiments, determining the target ontology object in the event to be analyzed includes: A large language model is used to extract objects from the event to be analyzed, thereby obtaining the target ontology object in the event to be analyzed.

[0033] It can be argued that when identifying the target ontology object in an event to be analyzed, it often relies on manual identification or simple keyword matching. This approach is difficult to accurately and comprehensively extract the core ontology object when faced with complex and unstructured event descriptions, resulting in a lack of accurate input basis for subsequent related object queries and impact analysis.

[0034] To address this, this application further proposes utilizing a pre-trained Large Language Model (LLM) for deep semantic understanding and entity recognition of the input text describing the event to be analyzed. This allows for the automatic identification and extraction of core entities involved in the event, i.e., target ontology objects. The aim is to solve the problem that traditional methods struggle to accurately and comprehensively identify key ontology objects when dealing with complex, unstructured event descriptions. Specifically, this can be achieved using any one or a combination of the following methods: One approach is to design prompts that take the text description of the event to be analyzed as input, guiding the LLM to identify and output supply chain-related entities mentioned in the event as target ontology objects. For example, the LLM could be instructed to "extract all potentially affected company or product names from the following event description." Another approach is to perform domain-adaptive fine-tuning on the general LLM model. By training the LLM model using a dataset containing a large number of supply chain event descriptions and their corresponding target ontology object annotations, it can more accurately identify and extract supply chain-specific entity information, thereby improving the model's recognition accuracy and robustness in specific business scenarios. Furthermore, Named Entity Recognition (NER) technology can be combined with large language models as part of or an auxiliary tool for NER models to perform multi-round or multi-level entity recognition on event text, ensuring the comprehensiveness and accuracy of the target ontology. Based on this, leveraging the powerful natural language understanding and entity recognition capabilities of large language models, target ontology objects can be automatically and accurately extracted from complex and varied event descriptions. This overcomes the inefficiencies, omissions, and strong subjectivity of traditional manual recognition or keyword matching methods when dealing with unstructured text, significantly improving the accuracy and automation level of subsequent event prediction information generation.

[0035] S102. Based on the preset supply chain relationship information, determine multiple target associated objects that match the target ontology object, as well as the transmission degree of each target associated object.

[0036] Supply chain relationship information can refer to a pre-built knowledge base or dataset that characterizes the supply and demand relationships between entities across the entire supply chain. It records in detail the supply and demand relationships between candidate ontology objects and multiple candidate related objects, and quantifies the hierarchy of these relationships. Specifically, supply chain relationship information is used to characterize multiple candidate related objects that have supply and demand relationships with each candidate ontology object across the entire supply chain, as well as the transmission degree of each candidate related object.

[0037] Transmission degree is used to quantify the supply and demand relationship hierarchy between each candidate ontology object and candidate related objects. It reflects the distance or strength of the transmission of event impact from the ontology object to the related object. For example, the transmission degree of a direct supplier can be 1, and the transmission degree of a second-tier supplier can be 2. The target related object can refer to an entity that has a supply and demand relationship with the target ontology object in the supply chain.

[0038] In specific implementations, the target ontology object can be used to query supply chain relationship information. The queried data can be directly used as multiple target related objects matched with the target ontology object, along with the transitivity of each target related object. Alternatively, the queried data can be further processed to accurately obtain the relevant data of the target related objects. For example, after obtaining the target ontology object "a certain automotive chip supplier," a pre-built supply chain relationship information can be queried. For instance, the direct customers of "a certain automotive chip supplier" (such as car manufacturers) can be assigned a transitivity of 1, while the downstream dealers of these car manufacturers can be assigned a transitivity of 2. In this way, multiple target related objects directly or indirectly related to "a certain automotive chip supplier" and their corresponding transitivity can be obtained.

[0039] It should be noted that in actual supply chain networks, the relationships often present a complex multi-level, network structure. If only simple direct relationship queries are performed, it is difficult to capture the deep-seated indirect effects. Furthermore, a large number of duplicate relationship nodes are easily generated during multi-hop queries, leading to data redundancy in subsequent analysis and affecting the efficiency and accuracy of the analysis.

[0040] In response, this application further proposes to determine multiple target related objects that match the target ontology object based on preset supply chain relationship information, as well as the transitivity of each target related object, including: obtaining preset supply chain relationship information; performing multi-hop transitivity queries on the target ontology object based on the supply chain relationship information to obtain multiple preliminary screening related objects corresponding to the target ontology object, as well as the transitivity of each preliminary screening related object; and performing deduplication processing on the multiple preliminary screening related objects to obtain multiple target related objects that match the target ontology object.

[0041] Specifically, acquiring pre-defined supply chain relationship information is fundamental for subsequent supply chain event analysis. For example, this can be achieved by extracting and integrating relevant data from a wide range of data sources, such as internal business databases, Enterprise Resource Planning (ERP) systems, and Customer Relationship Management (CRM) systems. These data sources typically contain information such as inter-company purchase orders, sales contracts, and logistics records. Through data cleaning and structuring, a supply chain relationship graph can be constructed. Alternatively, external data can be collected, such as from publicly available corporate registration information, industry reports, and news media, to obtain information on inter-company investment relationships, cooperation agreements, and supplier directories. This information can then be combined with natural language processing technology for entity recognition and relationship extraction to construct or update supply chain relationship information.

[0042] Secondly, within the graph of supply chain relationship information, starting from the target ontology object, multi-level traversal and exploration can be performed along the supply and demand relationship path to discover preliminary related objects that are directly or indirectly associated with the target ontology object. This query method can simulate the hierarchical transmission process of events in the supply chain, thereby revealing deeper potential impacts. For example, a custom graph traversal algorithm can be used to record each visited related object during the traversal process and determine its transmission degree based on its distance from the target ontology object (i.e., hop count). For example, the transmission degree of a direct association is 1, the transmission degree of a second-level association is 2, and so on.

[0043] Furthermore, to avoid data redundancy and duplicate analysis, it is necessary to deduplicate multiple initially screened related objects to remove duplicate entities and ensure that each unique related object appears only once. For example, the initially screened related objects can be stored in a hash set or dictionary, using the unique identifier of the initially screened related objects (such as enterprise ID, unified social credit code, etc.) as the key. The entire list can be traversed to deduplicate duplicate unique identifiers, thus obtaining multiple target related objects that match the target ontology object. Based on this, by introducing a multi-hop transmission query mechanism, the deep relationships in the supply chain network can be effectively mined, solving the problem that single-dimensional queries cannot cover complex transmission paths, and quantifying the depth and intensity of event diffusion in the supply chain. At the same time, by dedupplicating the initially screened related objects, redundant data generated by multiple path intersections is effectively eliminated, ensuring the uniqueness and accuracy of the final determined set of target related objects. This allows event analysis to more accurately focus on key influencing nodes, avoiding analytical biases caused by data duplication or omission, and laying a solid foundation for the subsequent generation of high-quality event transmission data.

[0044] S103. Generate event prediction information for the event to be analyzed based on the target ontology object, multiple target associated objects, and the transitivity of each target associated object.

[0045] Event prediction information can refer to comprehensive data about the potential impact of an event to be analyzed. The specific content included is not limited. For example, event prediction information can be used to characterize the transmission path between the target ontology object affected by the event to be analyzed and multiple target related objects.

[0046] After obtaining the target ontology object, multiple target associated objects, and the transitivity of each target associated object through the above steps, this step can generate event prediction information for the event to be analyzed. For example, the event prediction information for the event to be analyzed can be obtained by performing a series of analyses on the target ontology object, multiple target associated objects, and the transitivity of each target associated object. Alternatively, the event prediction information for the event to be analyzed can be directly generated using a large language model. This embodiment does not limit this approach.

[0047] For example, after determining "a certain automotive chip supplier" and its multiple target related objects (such as "automobile manufacturer A", "automobile manufacturer B", "dealer C") and their respective degrees of transmission, the relationship chain between the ontology object and the related objects can be listed directly as event prediction information for the event to be analyzed, such as "a certain automotive chip supplier -> automobile manufacturer A (degree of transmission 1) -> dealer C (degree of transmission 2)".

[0048] This embodiment provides an event analysis method that generates event prediction information for the event to be analyzed based on supply chain relationship information, providing structured constraints for event analysis and improving the accuracy of event prediction information.

[0049] Figure 2 This is a flowchart illustrating an event analysis method provided in another embodiment of this application. In this embodiment, the event prediction information includes event propagation data and predicted impact level. The process of generating event prediction information for the event to be analyzed based on the target ontology object, multiple target associated objects, and the propagation degree of each target associated object is further optimized as follows: Event propagation data for the event to be analyzed is generated based on the target ontology object, multiple target associated objects, and the propagation degree of each target associated object; the number of objects of multiple target associated objects and the event propagation data are input into a large language model to obtain the predicted impact level of the event to be analyzed. Figure 2 As shown, the method includes: S201. In response to an event analysis request, determine the target ontology object in the event to be analyzed.

[0050] S202. Based on the preset supply chain relationship information, determine multiple target associated objects that match the target ontology object, as well as the transmission degree of each target associated object.

[0051] S203. Generate event transmission data for the event to be analyzed based on the target ontology object, multiple target associated objects, and the transmission degree of each target associated object.

[0052] S204. Input the number of objects and event transmission data of multiple target related objects into the large language model to obtain the predicted impact level of the event to be analyzed.

[0053] Event propagation data refers to structured data used to characterize the propagation path between the target ontology object and multiple related target objects affected by the event being analyzed. It can clearly show the diffusion chain of the event's impact, revealing not only the path of the event's impact but also quantifying the intensity or level of the impact. Predicted impact level refers to a quantitative assessment of the overall impact of the event, which can be divided into different levels such as "low," "medium," and "high" to indicate the potential importance of the event.

[0054] This embodiment does not limit the means of generating event transmission data. One implementation is that the system can construct a directed graph or tree structure based on the target ontology object, multiple target associated objects, and the transmission degree of each target associated object. Nodes represent ontology objects or associated objects, edges represent supply and demand relationships, and the weight or attribute of the edges represents the transmission degree. This graph structure can be directly used as event transmission data for the event to be analyzed, or it can be further serialized, for example, processed into an adjacency list, adjacency matrix, or text representation in a graph database query language, to serve as event transmission data. Another implementation is that the system can extract the complete transmission chain between each target associated object and the target ontology object (i.e., all intermediate nodes from the ontology object to the associated object and their corresponding transmission degrees), and aggregate these chains in the form of structured text (e.g., JSON-formatted path descriptions, including the identifier and transmission degree of each node) to form event transmission data, focusing on describing the specific details of the transmission path.

[0055] Furthermore, the number of objects associated with multiple targets and the obtained event transmission data can be input into a large language model. The aim is to leverage the semantic understanding and reasoning capabilities of the large language model to comprehensively evaluate the breadth (number of objects) and depth (transmission data) of the event's impact, thereby outputting an objective predicted impact level. For example, the number of objects associated with multiple targets and the event transmission data (such as the serialized graph structure or structured text mentioned above) can be integrated into a pre-defined natural language template to form a detailed text prompt. This text prompt is then input into the large language model for processing, allowing the large language model to generate a predicted impact level for the event to be analyzed based on its training knowledge and reasoning capabilities.

[0056] Alternatively, the number of objects and event propagation data can be encoded into numerical vectors or embedded representations. These numerical representations, along with the embedded representations of the textual descriptions of the events to be analyzed, can then be fed into a large language model fine-tuned for a specific domain (such as supply chain risk analysis). This allows the model to directly output a structured prediction, such as a JSON object containing the impact level (e.g., a numerical value from 0 to 100) and a brief explanation, without the need for intermediate natural language prompt generation steps.

[0057] This embodiment provides an event analysis method that generates event transmission data that provides structured and quantified supply chain relationships and transmission hierarchy information, ensuring the accuracy and traceability of model reasoning and avoiding the illusion problem that occurs when traditional large language models lack structured knowledge support.

[0058] Simultaneously, by inputting the number of target-related objects (reflecting the breadth of impact) and event transmission data (reflecting the depth and path of impact) into the big language model, the model can comprehensively consider multiple dimensions of event impact, conduct in-depth analysis and summarization of this complex structured data, and thus output a more accurate and objective prediction of the impact level. This mechanism, which combines structured data with intelligent reasoning of the big language model, transforms complex supply chain relationships and transmission levels into structured information that can be processed and understood by machines. It makes up for the limitations of traditional rule matching or manual assessment in dealing with complex supply chain scenarios, significantly improves the depth and accuracy of event prediction information, and provides a more reliable basis for subsequent risk management and decision-making.

[0059] In some embodiments, event propagation data for the event to be analyzed is generated based on the target ontology object, multiple target associated objects, and the propagation degree of each target associated object, including: For each target associated object, the object identifier and transitivity of each target associated object, as well as the object identifier of the target ontology object, are structured to obtain the path node data of each target associated object; Sort the path node data of all target-related objects according to the degree of transmission to generate event transmission data for the event to be analyzed.

[0060] In specific implementations, discrete information about associated objects related to the target ontology object can be transformed into a standardized data format that is easy for machines to process and subsequently analyze. Through structured processing, the relationship between each associated object and the ontology object is clarified, and the strength of this relationship (transmission degree) is quantified, laying the foundation for constructing a clear transmission path. Specifically, this structured processing can be implemented in various ways. For example, JSON format can be used for structuring, generating a JSON object for each target associated object, which may contain an associated object identifier field, a transmission degree field, and an ontology object identifier field. Alternatively, tuples or lists can be used for structuring, representing the information of each target associated object as a triple such as "(associated object identifier, transmission degree, ontology object identifier)".

[0061] Then, based on the level or intensity of the event's impact, all related objects are logically arranged, so that the generated event transmission data can intuitively show the priority and transmission order of the impact, making it easier for users to understand and make decisions. For example, the transmission degree can be used as a sorting criterion to reflect the distance or intensity of the impact. Specifically, the sorting method can be selected according to actual needs. For example, ascending order can be used, where related objects with lower transmission degrees are listed first, indicating that they are more directly or strongly affected by the event. Building upon this foundation, structured processing integrates previously discrete information into unified path node data. This clearly defines the relationships and influence strength between each associated object and the ontology object. Furthermore, by sorting according to the degree of transmission, complex supply chain transmission paths can be logically arranged according to the depth or level of influence. This not only intuitively displays the priority and transmission order of event impacts, helping users quickly identify key impact paths, but also provides a high-quality, highly logical input foundation for subsequent impact level assessment by large language models. This generates event transmission data with a clear logical structure, effectively solving the problem of lack of structure and readability in event transmission data. Consequently, it significantly improves the accuracy and interpretability of the analysis results, thereby assisting decision-makers in making more accurate judgments.

[0062] In some embodiments, after inputting the object count and event propagation data of multiple target related objects into a large language model to obtain the predicted impact level of the event to be analyzed, the method includes: Based on the event to be analyzed, event transmission data, and predicted impact level, an impact analysis report of the event to be analyzed is generated.

[0063] This embodiment aims to integrate the aforementioned discrete analysis results into a unified and easy-to-read analysis report. One implementation involves the system pre-setting multiple report templates. These templates include fixed modules such as report title, event overview, impact scope analysis, transmission path details, and impact level summary. When generating the report, the system can populate the event overview module with the original information of the event to be analyzed, embed the event transmission data in the form of charts, lists, or structured text into the impact scope analysis and transmission path details modules, and present the predicted impact level in the impact level summary module. Another implementation involves submitting the original description of the event to be analyzed, structured event transmission data (e.g., serialized into text or JSON format), and the predicted impact level as input to a pre-trained large language model. This allows the large language model to automatically generate a complete report including event background, impact analysis, transmission path description, and impact level assessment. This report can be generated according to preset style and format requirements, such as requiring the model to generate a summary, detailed analysis, and conclusions.

[0064] Building upon this foundation, by integrating and textualizing the originally discrete transmission paths and impact levels, and leveraging the path logic inherent in the transmission data, combined with importance judgments in the prediction levels, the generated reports can objectively and comprehensively reflect the potential impact of events on the supply chain. This not only lowers the barrier for decision-makers to read complex data graphs but also provides a standardized medium for subsequent business reviews, knowledge accumulation, and cross-departmental communication. It effectively enhances the practical value and traceability of event analysis results, addresses decision-makers' needs for tracing and archiving the complete analysis process, and provides a means to transform analysis results into standardized, readable business documents. This, in turn, promotes the effective transmission and archiving of analysis conclusions within the organization, thereby achieving a closed loop from data analysis to business output.

[0065] In some embodiments, after inputting the object count and event propagation data of multiple target related objects into a large language model to obtain the predicted impact level of the event to be analyzed, the method includes: In response to the reviewer's request for correction, determine the actual impact level and the data on the reasons for the correction carried in the request; The predicted impact level of the event to be analyzed is adjusted based on the actual impact level. Based on the predicted impact level, the actual impact level, and the reasons for the correction, the parameters of the large language model are adjusted to obtain the adjusted model. The adjusted model is used to generate the predicted impact level of the new event to be analyzed.

[0066] It can be argued that the evaluation results generated by the model often lack a personalized calibration mechanism for specific business scenarios, and the model itself cannot iterate and optimize itself based on actual business feedback. As a result, when faced with complex and ever-changing supply chain events, the model's predictive accuracy is difficult to continuously improve, and it cannot form an effective knowledge loop.

[0067] To address this, this application further incorporates human expertise to calibrate the prediction results generated by the large language model. Specifically, after the large language model generates a predicted impact level for the event to be analyzed, the system presents this prediction to the reviewers. The reviewers then evaluate the model's prediction based on their professional experience in areas such as industry investment research, supply chain risk management, or emergency response to unforeseen events, as well as their in-depth understanding of the event. If the reviewers deem the prediction biased, they will initiate a correction request. This request includes not only the actual impact level that the reviewers believe to be more accurate but also detailed data outlining the reasons for the correction. For example, it might indicate that the model has biases in assessing the impact on a specific industry, has overlooked certain key supply chain transmission paths, or has misjudged the nature of the event. This data outlining the reasons for the correction can be structured label selections or detailed explanations in free text format.

[0068] After receiving the actual impact level provided by the reviewer, the system can directly update the original predicted value to the actual impact level provided by the reviewer, and store and display it as the final impact assessment result of the event to be analyzed.

[0069] The system can then use the original predicted impact level from the large language model, the actual impact level provided by the reviewers, and detailed reasoning data for correction as feedback information to adjust the parameters of the large language model. For example, supervised fine-tuning can be used to construct new training samples from this feedback data, guiding the model to learn how to more accurately infer the actual impact level from the original input (such as the target ontology object, the number of objects associated with multiple targets, and event propagation data). The reasoning data for correction can serve as additional contextual information, helping the model understand the reasons for prediction bias, thereby more accurately adjusting the weights and biases within the model.

[0070] Based on this, it can effectively transform human expertise into quantifiable monitoring signals, making up for the calibration deficiencies of large language models in specific business scenarios. Furthermore, it can correct the predicted impact level of the current event to be analyzed based on the actual impact level, ensuring the real-time accuracy of the analysis results for each event and providing a reliable basis for business decisions.

[0071] Furthermore, by using data on predicted impact levels, actual impact levels, and reasons for correction to adjust the parameters of the large language model, the model can continuously learn and iterate from historical correction experience, resulting in a continuously optimized and adjusted model. This allows for the generation of more accurate predicted impact levels that better align with actual business logic, effectively solving the problem of the model's inability to self-iterate and optimize based on actual business feedback. This significantly improves the system's predictive accuracy, adaptability, and intelligence level in the analysis of complex and ever-changing supply chain events, ultimately forming an effective knowledge loop.

[0072] Figure 3 This is a schematic diagram of the overall architecture of an event analysis method provided in an embodiment of this application, as shown below. Figure 3 As shown, the complete process from event input to final report is illustrated, including the following six core steps: Step S1: Event Input and Core Entity Extraction. External event data (including fields such as event_id, event title, event content, and event_source) is used as input. A large language model is then called to extract core entities, identifying the affected company names from the event content (i.e., using the large language model to extract objects from the event to obtain the target ontology objects). For example, a prompt can be designed: Extract the core company entity names from the following text, return JSON format, input the prompt into the large language model, and output a list of core entities: [Company A, Company B,...]. The model parameter temperature=0.3 can be set to ensure the stability of entity extraction.

[0073] Step S2: Graph multi-hop transmission path query, which means that for each core entity, recursive SQL queries can be used to query multi-level supply chain relationships in the industry graph. For example, the WITH RECURSIVE statement can be used to implement multi-hop transmission queries (up to 3 degrees of transmission). The specific query logic can start from the core company and recursively query upstream and downstream related companies along the supply chain relationship table. The returned results can include: company identifier (company_id), company name (company_name), transmission degree (degree), and source entity (source_entity) (that is, based on the supply chain relationship information, multi-hop transmission queries are performed on the target ontology object to obtain multiple initially screened related objects corresponding to the target ontology object, as well as the transmission degree of each initially screened related object).

[0074] Furthermore, you can limit the recursion depth (e.g., WHERE sc.degree<3) and the number of returned results (e.g., LIMIT 10) to avoid information overload. For example, the SQL query formula is degree(n)=degree(n-1)+1, where degree<3.

[0075] Step S3: Structured storage of influence paths, which means that the query results can be formatted into a structured list of influence paths. Each influence node includes: company ID, company name, transmission degree, source entity (that is, for each target associated object, the object identifier and transmission degree of each target associated object, as well as the object identifier of the target ontology object are structured to obtain the path node data of each target associated object).

[0076] The impact path list can be sorted by the degree field to ensure a clear hierarchy of transmission paths (i.e., sorting the path node data of all target related objects according to the transmission degree to generate the event transmission data of the event to be analyzed); it can also remove duplicate nodes in the impact path list to avoid the same company being recorded multiple times (i.e., deduplicating multiple initially screened related objects to obtain multiple target related objects that match the target ontology object).

[0077] Step S4: Intelligent event importance rating, which involves using a large language model to rate the event's importance (1-5 stars). For example, a Prompt can be designed: based on the following event and its impact path, assess the event's importance (1-5 stars), inputting information such as the event title, event content, number of affected companies, and impact path into the large language model to output a JSON format: {"rating":5,"reason":"Reason for Rating"} (i.e., inputting the number of related objects and event propagation data from multiple targets into the large language model to obtain the predicted impact level of the event to be analyzed). The model parameter temperature=0.5 can be set to improve flexibility.

[0078] Step S5: Automatic Generation of Analysis Report: Integrating event information, impact paths, and importance ratings, the LLM model is invoked to generate an analysis report (i.e., an impact analysis report is generated based on the event to be analyzed, event transmission data, and predicted impact level). Specifically, a Prompt can be designed: A brief analysis report is generated based on the following information: by inputting the event title, event content, impact path (JSON format), and rating into the LLM model, a structured analysis report text of no more than 200 characters is output, including impact analysis and a description of the transmission logic. Model parameters can be set as follows: temperature=0.7, max_tokens=500.

[0079] Step S6: Result storage and feedback loop, that is, the analysis results can be saved to the event_signals_feedback table, which includes fields such as: trigger_event, core_entities, AI inference impact transmission path ai_impact_path, AI risk rating ai_rating, and AI analysis conclusion ai_analysis.

[0080] Meanwhile, the initial state of the analysis results is set to PENDING, awaiting review by domain experts. After expert review, the rating can be corrected and the reasoning logic can be entered, including expert rating (expert_rating, i.e., actual impact level) and expert comment (expert_comment, i.e., data for the reason for correction). The data status after review is updated to REVIEWED, which can be used for subsequent LLM fine-tuning training, thus forming a data closed loop from AI analysis to expert review to data accumulation to model fine-tuning to AI analysis optimization.

[0081] As can be seen from the above description, the event analysis method provided in this embodiment involves the fields of knowledge graph reasoning and intelligent investment research technology. Specifically, it solves the technical problem of how to automatically reason and generate the industrial chain impact path and assess the importance of the event based on the industry graph and external events.

[0082] Compared to manual graph query and analysis methods, which require manual querying and judgment each time, taking several hours and resulting in low analysis efficiency; manual analysis also struggles to fully trace multi-level transmission paths and suffers from significant differences in judgment results among different analysts. The event analysis method provided in this embodiment can automatically discover all influencing nodes within 3 degrees through recursive querying, automatically generate influence paths and analysis reports, and has a unified rating standard and analysis process, ensuring the consistency of analysis results.

[0083] Compared to rule-based impact assessment methods, which suffer from limitations in rule base coverage and flexibility due to their inability to handle complex and novel event scenarios, as well as the need for continuous rule base updates and high maintenance costs, the event analysis method provided in this embodiment can automatically reason based on graph structures and LLM models without the need for predefined rules. It can also flexibly handle various types of events. Furthermore, the graph data can be automatically updated, improving the versatility of the reasoning logic.

[0084] Compared to LLM direct analysis methods, which lack structured knowledge support and are prone to misleading results, particularly due to the lack of graph constraints leading to illogical influence paths and poor reasoning logic, this embodiment's event analysis method can perform multi-hop transmission queries based on real-time industry graphs, ensuring information accuracy. The generated graph transmission paths provide structured constraints, making the reasoning results more reliable. Furthermore, the influence paths are stored in JSONB format, supporting manual review and traceability.

[0085] Furthermore, the analysis results of the event analysis method provided in this embodiment are automatically stored as structured data, supporting subsequent model fine-tuning and optimization, forming a data closed loop; it supports AI preliminary rating and expert review and correction, balancing analysis efficiency and accuracy; recursive queries support propagation reasoning of arbitrary degrees, and the depth can be adjusted according to needs, making the method highly scalable; based on real-time graph data, it reflects the latest industry chain relationships, improving the real-time performance of the event analysis method.

[0086] Corresponding to the event analysis method in the above embodiments, Figure 4 This is a structural block diagram of an event analysis device provided in one embodiment of this application. For ease of explanation, only the parts related to the embodiment of this application are shown.

[0087] Reference Figure 4 The device includes: The first determination module 301 is used to determine the target ontology object in the event to be analyzed in response to the event analysis request; The second determining module 302 is used to determine multiple target associated objects that match the target ontology object and the transmission degree of each target associated object based on the preset supply chain relationship information. The supply chain relationship information is used to characterize multiple candidate associated objects that have supply and demand relationships with each candidate ontology object in the entire supply chain domain, and the transmission degree of each candidate associated object. The transmission degree is used to quantify the supply and demand relationship level between each candidate ontology object and the candidate associated object. The generation module 303 is used to generate event prediction information for the event to be analyzed based on the target ontology object, multiple target associated objects, and the transmission degree of each target associated object. The event prediction information is used to characterize the transmission path between the target ontology object and multiple target associated objects affected by the event to be analyzed.

[0088] This embodiment provides an event analysis device. A first determining module responds to an event analysis request by determining a target ontology object in the event to be analyzed. A second determining module, based on preset supply chain relationship information, determines multiple target associated objects matching the target ontology object, and the transmission degree of each target associated object. The supply chain relationship information characterizes multiple candidate associated objects with supply and demand relationships with each candidate ontology object across the entire supply chain, and the transmission degree of each candidate associated object quantifies the supply and demand relationship hierarchy between each candidate ontology object and its associated objects. A generation module generates event prediction information for the event to be analyzed based on the target ontology object, the multiple target associated objects, and the transmission degree of each target associated object. The event prediction information characterizes the transmission path between the target ontology object affected by the event and the multiple target associated objects. Using this device, by generating event prediction information for the event to be analyzed based on supply chain relationship information, structured constraints are provided for event analysis, improving the accuracy of the event prediction information.

[0089] Optionally, the second determining module is specifically used for: Obtain pre-defined supply chain relationship information; Based on supply chain relationship information, multi-hop transit query is performed on the target ontology object to obtain multiple preliminary screening related objects corresponding to the target ontology object, as well as the transit degree of each preliminary screening related object; Multiple initially screened related objects are deduplicated to obtain multiple target related objects that match the target ontology object.

[0090] Optionally, the event prediction information includes event propagation data and predicted impact level, and the generation module includes: The first generation unit is used to generate event transmission data of the event to be analyzed based on the target ontology object, multiple target associated objects, and the transmission degree of each target associated object; The input unit is used to input the number of objects and event propagation data of multiple target related objects into the large language model to obtain the predicted impact level of the event to be analyzed.

[0091] Optionally, the first generating unit is specifically used for: For each target associated object, the object identifier and transitivity of each target associated object, as well as the object identifier of the target ontology object, are structured to obtain the path node data of each target associated object; Sort the path node data of all target-related objects according to the degree of transmission to generate event transmission data for the event to be analyzed.

[0092] Optionally, the generation module also includes: The second generation unit is used to generate an impact analysis report of the event to be analyzed based on the event to be analyzed, the event transmission data, and the predicted impact level after inputting the number of objects and event transmission data of multiple target related objects into the large language model.

[0093] Optionally, the generation module also includes: The determination unit is used to determine the actual impact level and correction reason data carried in the correction request after inputting the number of objects and event transmission data of multiple target related objects into the large language model to obtain the predicted impact level of the event to be analyzed; The correction unit is used to correct the predicted impact level of the event to be analyzed based on the actual impact level. The parameter adjustment unit is used to adjust the parameters of the large language model based on the predicted impact level, the actual impact level, and the reason for correction, so as to obtain the adjusted model. The adjusted model is used to generate the predicted impact level of the new event to be analyzed.

[0094] Optionally, the first determining module is specifically used for: A large language model is used to extract objects from the event to be analyzed, thereby obtaining the target ontology object in the event to be analyzed.

[0095] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0096] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0097] Figure 5 This is a schematic diagram of the structure of a terminal device provided in one embodiment of this application, as shown below. Figure 5As shown, the terminal device 500 of this embodiment includes: at least one processor 502 ( Figure 5 (Only one is shown) a processor, a memory 501, and a computer program 503 stored in the memory 501 and executable on at least one processor 502. When the processor 502 executes the computer program 503, it implements the steps in the control method embodiments of any of the above-described application programs.

[0098] Terminal device 500 can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. This terminal device may include, but is not limited to, a processor 502 and a memory 501. Those skilled in the art will understand that... Figure 5 This is merely an example of terminal device 500 and does not constitute a limitation on terminal device 500. It may include more or fewer components than shown in the figure, or combine certain components, or different components, such as input / output devices, network access devices, etc.

[0099] The processor 502 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0100] In some embodiments, memory 501 may be an internal storage unit of terminal device 500, such as a hard disk or memory of terminal device 500. In other embodiments, memory 501 may be an external storage device of terminal device 500, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on terminal device 500. Furthermore, memory 501 may include both internal and external storage units of terminal device 500. Memory 501 is used to store operating system, applications, boot loader, data, and other programs, such as program code for computer programs. Memory 501 can also be used to temporarily store data that has been output or will be output.

[0101] This application also provides a computer-readable storage medium storing a computer program, which, when executed by processor 502, can implement the steps in the above-described method embodiments.

[0102] This application provides a computer program product that, when run on a terminal device, enables the terminal device to implement the steps described in the various method embodiments.

[0103] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by the processor 502, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable storage medium can include at least: any entity or device capable of carrying computer program code to a device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0104] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0105] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0106] In the embodiments provided in this application, it should be understood that the disclosed apparatus / terminal devices and methods can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0107] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0108] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. An event analysis method, characterized in that, The method includes: In response to an event analysis request, determine the target ontology object in the event to be analyzed; Based on the preset supply chain relationship information, multiple target associated objects that match the target ontology object are determined, as well as the transmission degree of each target associated object. The supply chain relationship information is used to characterize multiple candidate associated objects that have supply and demand relationships with each candidate ontology object in the entire supply chain, as well as the transmission degree of each candidate associated object. The transmission degree is used to quantify the supply and demand relationship level between each candidate ontology object and the candidate associated object. Based on the target ontology object, the plurality of target associated objects, and the transitivity of each target associated object, event prediction information for the event to be analyzed is generated. The event prediction information is used to characterize the transit path between the target ontology object and the plurality of target associated objects affected by the event to be analyzed.

2. The event analysis method as described in claim 1, characterized in that, The step of determining multiple target associated objects matching the target ontology object based on preset supply chain relationship information, and the transmission degree of each target associated object, includes: Obtain the preset supply chain relationship information; Based on the supply chain relationship information, a multi-hop transit query is performed on the target ontology object to obtain multiple preliminary screening associated objects corresponding to the target ontology object, and the transit degree of each preliminary screening associated object; The multiple initially screened associated objects are deduplicated to obtain multiple target associated objects that match the target ontology object.

3. The event analysis method as described in claim 1, characterized in that, The event prediction information includes event propagation data and predicted impact level. The step of generating event prediction information for the event to be analyzed based on the target ontology object, the multiple target associated objects, and the propagation degree of each target associated object includes: Based on the target ontology object, the multiple target associated objects, and the transitivity of each target associated object, the event transitivity data of the event to be analyzed is generated; The number of objects of the multiple target associated objects and the event transmission data are input into the large language model to obtain the predicted impact level of the event to be analyzed.

4. The event analysis method as described in claim 3, characterized in that, The step of generating event transmission data for the event to be analyzed based on the target ontology object, the multiple target associated objects, and the transmission degree of each target associated object includes: For each of the target associated objects, the object identifier and transitivity of each target associated object, as well as the object identifier of the target ontology object, are structured to obtain the path node data of each target associated object; The path node data of all target-related objects are sorted according to the degree of transmission to generate the event transmission data of the event to be analyzed.

5. The event analysis method as described in claim 3, characterized in that, After inputting the number of objects of the multiple target associated objects and the event transmission data into the large language model to obtain the predicted impact level of the event to be analyzed, the method includes: Based on the event to be analyzed, the event transmission data, and the predicted impact level, an impact analysis report of the event to be analyzed is generated.

6. The event analysis method as described in claim 3, characterized in that, After inputting the number of objects of the multiple target associated objects and the event transmission data into the large language model to obtain the predicted impact level of the event to be analyzed, the method includes: In response to a reviewer's request for correction, determine the actual impact level and reasoning data carried in the request for correction; The predicted impact level of the event to be analyzed is adjusted based on the actual impact level. Based on the predicted impact level, the actual impact level, and the data on the reasons for correction, the parameters of the large language model are adjusted to obtain the adjusted model. The adjusted model is used to generate the predicted impact level of a new event to be analyzed.

7. The event analysis method according to any one of claims 1-6, characterized in that, The process of determining the target ontology object in the event to be analyzed includes: A large language model is used to extract objects from the event to be analyzed, thereby obtaining the target ontology object in the event to be analyzed.

8. An event analysis device, characterized in that, The device includes: The first determination module is used to determine the target ontology object in the event to be analyzed in response to the event analysis request; The second determining module is used to determine, based on preset supply chain relationship information, multiple target associated objects that match the target ontology object, and the transmission degree of each target associated object. The supply chain relationship information is used to characterize multiple candidate associated objects that have supply and demand relationships with each candidate ontology object in the entire supply chain, and the transmission degree of each candidate associated object. The transmission degree is used to quantify the supply and demand relationship level between each candidate ontology object and the candidate associated object. The generation module is used to generate event prediction information for the event to be analyzed based on the target ontology object, the plurality of target associated objects, and the transmission degree of each target associated object. The event prediction information is used to characterize the transmission path between the target ontology object and the plurality of target associated objects affected by the event to be analyzed.

9. A terminal device, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it causes the terminal device to implement the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on a terminal device, it causes the terminal device to perform the method as described in any one of claims 1-7.

Images