An enterprise risk prediction method based on world model and ontology constraint

Abstract

The application belongs to the technical field of enterprise risk management and digital intelligent decision-making, and particularly relates to an enterprise risk prediction method based on a world model and ontology constraint, wherein a hierarchical enterprise ontology and a rule system are constructed; a current organization state is coded into a patternized compressed state graph, internal and external heterogeneous events are accessed through a multi-source event interface, are mapped into state updates through semantic matching, and an initial state graph with ontology semantic constraint is obtained; a forward dynamics deduction is performed on the initial state graph by the world model, ontology rules are embedded in each iteration to check and correct invalid simulation branches in real time, a compliance state trajectory is generated, an organization health degree evolution sequence and multiple types of operation risk indicators are calculated, and a risk event chain is generated through ontology semantic mapping; a candidate intervention is filtered through ontology compliance checking, and a target recommended intervention is output. Thus, the problems of low storage efficiency in a large-scale scene, difficulty in adapting to external heterogeneous events and the like in the prior art are solved.

Description

Technical Field

[0001] This invention belongs to the field of enterprise risk management and digital intelligent decision-making technology, specifically involving an enterprise risk prediction method based on world model and ontology constraints. Background Technology

[0002] As enterprises expand and their organizational structures become increasingly complex, the operational risks they face become more diverse and dynamic. Traditional enterprise risk management primarily relies on historical data statistical analysis for prediction. With the advancement of enterprise digital transformation, risk prediction is gradually shifting from manual experience-based judgment to intelligent solutions. Currently, the mainstream technical approaches fall into three categories: ontology / knowledge graph-driven static risk identification, dynamic health assessment driven by dynamics, and intervention scheme selection driven by simulation sandboxes. However, each technology has inherent shortcomings. Ontology and knowledge graph solutions can identify and match known risks, but only support static correlation reasoning, lacking the ability to infer future risk evolution trends and transmission paths. Dynamics inference solutions can complete dynamic trend prediction, but the state vectors lack business semantic binding, and the inference is not constrained by business and compliance rules, easily leading to problems where numerical results are reasonable but business logic is invalid, resulting in insufficient interpretability of the output results. Simulation sandboxes can compare the effects of multiple intervention schemes, but candidate interventions lack prior verification of compliance and business systems, resulting in poor implementation of the output schemes and potential violations of regulations.

[0003] Existing technologies generally suffer from the problem of shallow splicing between ontology and world model. Ontology fails to play a constraining role by being embedded in the entire lifecycle of world modeling, inference, and intervention. At the same time, there are a series of defects such as difficulty in ensuring rule consistency, low storage efficiency in large-scale scenarios, difficulty in adapting to external heterogeneous events, and high computational cost of inference. These technologies are difficult to meet the actual business needs of enterprises for high-precision, high-reliability, and high-efficiency risk prediction. Summary of the Invention

[0004] This application provides an enterprise risk prediction method based on world model and ontology constraints to solve problems such as low storage efficiency in large-scale scenarios and difficulty in adapting to external heterogeneous events in existing technologies.

[0005] The first aspect of this application provides a method for predicting enterprise risk based on a world model and ontology constraints, comprising the following steps: constructing a hierarchical enterprise ontology and rule system; based on the hierarchical enterprise ontology and rule system, encoding the current organizational state into a patterned compressed state diagram, and accessing internal and external heterogeneous events through a multi-source event interface, mapping them into state updates through semantic matching, to obtain an initial state diagram with ontology semantic constraints; performing forward dynamics deduction on the initial state diagram using a world model, embedding ontology rules for real-time verification at each iteration to correct invalid simulation branches, generating a compliant state trajectory, and calculating the organizational health evolution sequence and multiple types of operational risk indicators, generating a risk event chain through ontology semantic mapping; filtering candidate interventions based on the risk event chain through ontology compliance verification, evaluating the effect through parallel counterfactual deduction, and outputting a target recommended intervention.

[0006] Preferably, the hierarchical enterprise ontology and rule system is constructed, including: building a shared core entity layer; constructing a business ontology, a risk ontology, and a compliance ontology on the shared core entity layer, and formally defining each type of ontology using four elements: concept set, attribute set, relation set, and axiom set; configuring business rules, risk rules, and compliance rules for each type of ontology based on the formal definitions; and establishing a three-level verification mechanism of syntax verification, semantic verification, and consistency verification using the configured business rules, risk rules, and compliance rules to complete the construction of the hierarchical enterprise ontology and rule system.

[0007] Preferably, the business ontology describes the organizational entity, assets, and business processes; the risk ontology describes the risk type, transmission relationship, and degree of impact; and the compliance ontology describes the regulatory provisions, control points, and responsible entities. The business rules adopt SHACL syntax, the risk rules adopt SWRL syntax, and the compliance rules adopt OWL2 restricted syntax, and the rules are divided into two categories: hard constraints and soft constraints. The consistency check is used to detect conflicts between different rules and resolve them according to a preset priority.

[0008] Preferably, obtaining an initial state graph with ontology semantic constraints includes: extracting entities and aligning them with ontology for entities, assets, business processes, and risk events within the organization; eliminating redundant dimensions with a shared core entity layer as a reference to obtain a compressed state graph with entities as nodes and ontology relationships as edges; based on the compressed state graph, accessing internal business system events and external public information events through a multi-source event interface; mapping events to corresponding nodes and edges of the compressed state graph through semantic matching, entity linking, and relationship extraction, and superimposing them in chronological order to complete the state update; and performing a three-level verification mechanism of syntax verification, semantic verification, and consistency verification on the updated compressed state graph to eliminate mapping results that conflict with the layered enterprise ontology, thereby obtaining an initial state graph with ontology semantic constraints.

[0009] Preferably, forward dynamics deduction is performed on the initial state diagram using a world model. Each iteration embeds ontology rules for real-time verification to correct invalid simulation branches and generate a compliant state trajectory. This includes: starting from the initial state diagram, dividing the deduction iteration at each time step into three sequential stages: numerical iteration, rule verification, and branch processing. In the numerical iteration stage, the next time step value of each entity's state is calculated based on the organizational dynamics model to obtain a candidate state set. Based on the candidate state set, in the rule verification stage, hard and soft constraints are invoked to determine the compliance of each candidate state. In the branch processing stage, simulation branches that violate hard constraints are terminated, and states that violate soft constraints are corrected to the feasible region according to the principle of minimum perturbation and potential risk factors are marked. Compliant states that pass the verification are retained. Based on the compliant states retained step by step, a continuous time evolution sequence is generated to obtain the compliant state trajectory.

[0010] Preferably, the candidate interventions are first filtered through ontology compliance verification based on the risk event chain, and their effectiveness is evaluated through parallel counterfactual inference to output the target recommended intervention. This includes: enumerating a set of candidate interventions based on the risk event chain, and performing ontology compliance verification on each candidate intervention according to the compliance ontology. If the execution premise or execution consequence of a candidate intervention violates any hard constraint, it is filtered out, and a subset of candidate interventions that pass the verification is retained. For each candidate intervention in the subset of candidate interventions, parallel counterfactual inference is initiated with the current compliance state trajectory as the baseline. One or more simulation acceleration methods are used, such as baseline state sharing and write-time copy, common prefix trajectory reuse, and early pruning, to calculate the risk indicator reduction and compliance cost after the implementation of the candidate intervention. Based on the risk indicator reduction and compliance cost of each candidate intervention, the ratio of the risk indicator reduction to the compliance cost is used as the comprehensive utility, and the candidate intervention with the largest comprehensive utility is selected as the target recommended intervention and output.

[0011] The second aspect of this application provides an enterprise risk prediction system based on a world model and ontology constraints, comprising: an ontology construction module for constructing a hierarchical enterprise ontology and rule system; a state modeling module for encoding the current organizational state into a patterned compressed state diagram based on the hierarchical enterprise ontology and rule system, and accessing internal and external heterogeneous events through a multi-source event interface, mapping them into state updates through semantic matching, to obtain an initial state diagram with ontology semantic constraints; a deduction and prediction module for performing forward dynamic deduction on the initial state diagram using the world model, embedding ontology rules for real-time verification at each iteration to correct invalid simulation branches, generating a compliant state trajectory, and calculating the organizational health evolution sequence and multiple types of operational risk indicators, generating a risk event chain through ontology semantic mapping; and an intervention decision module for filtering candidate interventions based on the risk event chain through ontology compliance verification, evaluating the effect through parallel counterfactual deduction, and outputting a target recommended intervention.

[0012] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the program to implement an enterprise risk prediction method based on a world model and ontology constraints as described in the above embodiments.

[0013] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement an enterprise risk prediction method based on a world model and ontology constraints as described in the above embodiments.

[0014] The fifth aspect of this application provides a computer program product, including a computer program or instructions, for implementing an enterprise risk prediction method based on a world model and ontology constraints as described in the above embodiments.

[0015] Therefore, this application includes the following beneficial effects: By constructing a hierarchical enterprise ontology and rule system, the organizational state is encoded into a patterned compressed state diagram and semantic mapping updates of multi-source heterogeneous events are completed. Forward dynamics deduction is carried out based on the world model, and real-time verification of ontology rules is introduced during the iteration process to correct invalid simulation branches and generate compliant state trajectories. Combined with ontology semantic mapping, a traceable risk event chain is output. Simultaneously, pre-emptive ontology compliance filtering and parallel counterfactual deduction evaluation are implemented for candidate interventions. This overcomes the shortcomings of static risk identification in being unable to deduce risk evolution, dynamic deduction lacking business semantics and compliance constraints, intervention schemes easily violating regulations, and shallow splicing of ontology and world model. It balances large-scale scenario storage efficiency with heterogeneous event adaptability, realizing dynamic deduction of enterprise risks, interpretable analysis, and intelligent recommendation of compliant intervention schemes, providing reliable decision support for enterprise operational risk management. Thus, it solves the problems of low large-scale scenario storage efficiency and difficulty in adapting to external heterogeneous events in existing technologies.

[0016] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0017] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart illustrating a method for predicting enterprise risk based on a world model and ontology constraints, according to an embodiment of this application. Figure 2 This is a system architecture diagram provided according to one embodiment of the present application; Figure 3 This is a flowchart of an overall method provided according to an embodiment of this application; Figure 4 This is a flowchart illustrating a compliance-preliminary counterfactual deduction and intervention recommendation process according to one embodiment of this application; Figure 5 This is a schematic diagram of the structure of an enterprise risk prediction system based on a world model and ontology constraints according to an embodiment of this application; Figure 6 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0018] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0019] The following describes an enterprise risk prediction method based on a world model and ontology constraints, according to an embodiment of this application, with reference to the accompanying drawings. Addressing the issue of low storage efficiency in large-scale scenarios mentioned in the background, this application provides an enterprise risk prediction method based on a world model and ontology constraints. This method constructs a hierarchical enterprise ontology and rule system, encodes organizational states into a patterned compressed state diagram, and completes semantic mapping updates for multi-source heterogeneous events. It relies on the world model to conduct forward dynamics deduction and introduces real-time verification of ontology rules during the iteration process, correcting invalid simulation branches to generate compliant state trajectories. Combined with ontology semantic mapping, it outputs a traceable risk event chain. Simultaneously, it implements pre-emptive ontology compliance filtering and parallel counterfactual deduction evaluation for candidate interventions. This overcomes the shortcomings of static risk identification, which cannot deduce risk evolution; dynamics deduction, which lacks business semantics and compliance constraints; intervention plans, which are prone to violating regulations; and the shallow splicing of ontology and world models. It balances large-scale scenario storage efficiency with heterogeneous event adaptability, achieving dynamic deduction of enterprise risks, interpretable analysis, and intelligent recommendation of compliant intervention plans, providing reliable decision support for enterprise operational risk management. Thus, it solves the problems of low storage efficiency in large-scale scenarios and difficulty in adapting to external heterogeneous events in existing technologies.

[0020] Specifically, Figure 1 This application provides a method for predicting enterprise risk based on a world model and ontology constraints.

[0021] like Figure 1 As shown, this enterprise risk prediction method based on a world model and ontology constraints includes the following steps: In step S101, a hierarchical enterprise ontology and rule system is constructed.

[0022] Among them, the hierarchical enterprise ontology is a formal knowledge model that uses hierarchical inheritance and composition relationships, such as parent class and child class, whole and part, to structurally organize the core concepts in the enterprise and their interrelationships, so as to achieve a unified description and reasoning of the static structure and dynamic semantics of the enterprise.

[0023] It is understood that the embodiments of this application, through a layered architecture of shared core entity layer plus multiple specialized ontologies of business, risk, and compliance, coupled with differentiated rule syntax and a three-level verification mechanism, can unify the definitions of concepts, attributes, and relationships in the business, risk, and compliance domains of an enterprise. This eliminates heterogeneous ambiguities from multi-source information, completes entity standardization alignment, and removes redundant dimensions, providing a unified, rigorous, and verifiable semantic foundation for enterprise risk prediction. It can accurately undertake semantic mapping and state diagram updates of internal and external multi-source events, ensuring the standardization and consistency of initial state modeling. It can also provide real-time rule constraints for forward dynamics extrapolation of the world model, filter invalid simulation branches, correct extrapolation deviations, support the accurate generation of compliance state trajectories and risk event chains, and conduct compliance verification and conflict resolution for intervention strategies. From modeling, extrapolation, risk tracing to intervention recommendation, it solidifies semantic compliance and logical rigor throughout the entire process, effectively solving the problems of fragmented information, inconsistent rules, unconstrained extrapolation, and uninterpretable results in traditional enterprise risk prediction.

[0024] In this embodiment, a hierarchical enterprise ontology and rule system is constructed, including: building a shared core entity layer; constructing a business ontology, a risk ontology, and a compliance ontology on top of the shared core entity layer, and formally defining each type of ontology using four elements: concept set, attribute set, relation set, and axiom set; configuring business rules, risk rules, and compliance rules for each type of ontology based on the formal definitions; and establishing a three-level verification mechanism of syntax verification, semantic verification, and consistency verification using the configured business rules, risk rules, and compliance rules to complete the construction of the hierarchical enterprise ontology and rule system.

[0025] Among them, the shared core entity layer is a key intermediate layer in the layered enterprise ontology architecture. It abstracts and encapsulates the core concepts and relationships that are universally shared across various business domains and systems. It serves as a common foundation for semantic integration and reuse, and connects the top-level general ontology with the bottom-level domain-specific ontology.

[0026] It is understood that the shared core entity layer in this application serves as a unified foundation for the layered enterprise ontology, enabling standardized, unified definitions and structured encapsulation of enterprise organizational entities, assets, and core business entities. This provides a consistent entity semantic benchmark for upper-layer business ontology, risk ontology, and compliance ontology. It effectively opens up entity association channels between multi-domain ontology, eliminating conceptual fragmentation, entity ambiguity, and definition conflicts caused by independent modeling of various specialized ontology. In the process of multi-source heterogeneous event access, entity extraction, and ontology alignment, it relies on the unified core entity benchmark to eliminate redundant dimensions, achieve accurate entity links and state mapping, and significantly improve the standardization and uniformity of state diagram construction. At the same time, it provides general entity support for the full-domain three-level verification mechanism, world model dynamics deduction, and counterfactual intervention assessment, ensuring the consistency, reusability, and scalability of risk prediction modeling throughout the entire process.

[0027] In this embodiment, the business ontology describes the organizational entity, assets, and business processes; the risk ontology describes the risk type, transmission relationship, and degree of impact; and the compliance ontology describes the regulatory provisions, control points, and responsible entities. Business rules adopt SHACL syntax, risk rules adopt SWRL syntax, and compliance rules adopt OWL2 restricted syntax, and the rules are divided into two categories: hard constraints and soft constraints. Consistency verification is used to detect conflicts between different rules and resolve them according to preset priorities.

[0028] It is understood that the embodiments of this application decouple modeling through three dedicated ontology domains—business, risk, and compliance—to accurately define the core semantic elements of each domain, achieving a comprehensive and structured characterization of enterprise business operations, risk evolution, and compliance control. Simultaneously, it matches differentiated rule syntaxes of SHACL, SWRL, and OWL2 to the characteristics of different domains, aligning with the different operational mechanisms of business data verification, risk logic reasoning, and rigid compliance constraints, while ensuring both the accuracy and adaptability of rule expression. By distinguishing between hard and soft constraints, it differentiates between absolute compliance bottom lines and optimization risk indicators, enhancing the hierarchy and rationality of simulation and risk assessment. Coupled with cross-rule consistency verification and priority conflict resolution mechanisms, it effectively avoids logical contradictions, semantic conflicts, and reasoning paradoxes arising from the superposition of rules across multiple domains, ensuring the rigorous unity, compatibility, and synergy of the overall ontology rule system. This provides layered, precise, and conflict-free semantic rule support for subsequent state diagram update verification, world model compliance simulation, risk event chain generation, and compliance screening of intervention strategies, significantly improving the accuracy, interpretability, and feasibility of enterprise risk prediction.

[0029] In step S102, based on the hierarchical enterprise ontology and rule system, the current organizational state is encoded into a patterned compressed state diagram, and internal and external heterogeneous events are accessed through a multi-source event interface and mapped to state updates through semantic matching to obtain an initial state diagram with ontology semantic constraints.

[0030] Among them, the pattern-compressed state diagram is a formal graphical representation that abstracts the original complex state machine into a simplified general template structure through preset pattern matching and state equivalence reduction, aiming to reduce the state space complexity in system verification and model detection.

[0031] It is understood that the modular compressed state diagram in this application uses a hierarchical enterprise ontology and shared core entities as a basis. It unifies the complex organizational entities, assets, business and risk information of the enterprise, structures the relationships and removes redundancies. It transforms the fragmented, high-dimensional and heterogeneous original enterprise operation data into a standardized topological graph structure with entities as nodes and ontology semantic relationships as edges. This not only greatly simplifies the representation dimensions of the enterprise's dynamic state and reduces the computational complexity and redundant operations of the world model forward inference, but also achieves compact, standardized and computable modeling of the enterprise's complex operation state. Furthermore, it relies on ontology semantic constraints to solidify business logic and relationships, ensuring accurate alignment of multi-source internal and external event access, semantic mapping and state iteration updates. This provides a unified, low-redundancy, semantically reliable and time-series iterative standardized state carrier for subsequent dynamic simulation, risk trajectory evolution, risk transmission link mining and counterfactual inference of intervention strategies. It effectively solves the problems of messy data, dimensional redundancy, semantic missing and inferability in traditional enterprise state modeling.

[0032] In this embodiment, obtaining an initial state graph with ontology semantic constraints includes: extracting entities and aligning them with ontology for entities, assets, business processes, and risk events within the organization; eliminating redundant dimensions with a shared core entity layer as a reference to obtain a compressed state graph with entities as nodes and ontology relationships as edges; based on the compressed state graph, accessing internal business system events and external public information events through a multi-source event interface; mapping events to corresponding nodes and edges of the compressed state graph through semantic matching, entity linking, and relationship extraction, and superimposing them in chronological order to complete the state update; and verifying the updated compressed state graph using a three-level verification mechanism of syntax verification, semantic verification, and consistency verification to eliminate mapping results that conflict with the layered enterprise ontology, thereby obtaining an initial state graph with ontology semantic constraints.

[0033] Among them, ontology semantic constraints refer to formal rules applied to entity types, attribute value ranges, and relation topologies based on conceptual connotations and logical axioms, in order to ensure the consistency, completeness, and reasoning effectiveness of domain knowledge representation.

[0034] It is understood that the ontology semantic constraints in this application rely on a hierarchical ontology system and multi-type domain rules to provide standardized semantic logic and rigid compliance boundaries for the entire enterprise risk prediction process. This enables the standardized constraint and precise verification of the semantic mapping, entity linking, and state update results of multi-source heterogeneous events, eliminating erroneous state mapping results that violate business logic, risk mechanisms, and compliance requirements, and ensuring the semantic correctness and logical rationality of the initial state diagram. At the same time, real-time rule auditing is implemented in each step of the forward dynamics deduction of the world model, actively terminating invalid simulation branches, correcting non-compliant state disturbances, and marking potential risk factors. This avoids the problems of logic being detached from reality, results being unexplainable, and deduction trajectory distortion that exist in pure data-driven deduction. It also provides a unified semantic judgment basis for risk event chain generation, candidate intervention compliance filtering, and rule conflict resolution, making model simulation deduction no longer a simple numerical calculation, but an interpretable, verifiable, and implementable intelligent prediction process that conforms to enterprise business mechanisms, risk transmission laws, and legal and regulatory requirements, significantly improving the credibility and accuracy of risk deduction and intervention decisions.

[0035] In step S103, forward dynamics deduction is performed on the initial state diagram using the world model. In each iteration, ontology rules are embedded for real-time verification to correct invalid simulation branches, generating a compliant state trajectory. The organizational health evolution sequence and multiple types of operational risk indicators are calculated, and a risk event chain is generated through ontology semantic mapping.

[0036] Among them, the world model is a core concept in the field of artificial intelligence. It refers to the abstract representation system that an intelligent agent builds internally through learning the inherent dynamic laws of the environment, which can simulate, predict and imagine future states to support complex reasoning and decision-making.

[0037] It is understood that the world model in this application takes an initial state diagram with ontology semantic constraints as the starting point for deduction. Based on organizational dynamics, it completes the temporal evolution simulation of the enterprise system state. It can perform multi-step forward dynamic deduction of the enterprise's future business direction and risk transmission process, and generate a multi-branch candidate state set. Unlike the traditional static risk assessment method, it can reproduce the dynamic evolution process of the enterprise's complex system over time. Combined with the real-time verification mechanism of ontology rules, it terminates, corrects and marks the simulation branches with risks, and outputs the compliance state trajectory. It supports the generation of organizational health evolution sequence, multiple types of business risk indicators and risk event chains. At the same time, it can also undertake parallel counterfactual deduction of candidate interventions, simulate the enterprise risk evolution effect after the implementation of different intervention measures, and realize the quantitative evaluation of intervention plans. It breaks through the limitations of traditional risk prediction, which can only make static judgments and is difficult to predict the risk evolution path and the long-term effect of intervention. It provides a simulateable and quantifiable deduction base for the enterprise's dynamic risk prediction and intervention decision-making.

[0038] In this embodiment, forward dynamics deduction is performed on the initial state diagram using a world model. Each iteration embeds ontology rules for real-time verification to correct invalid simulation branches and generate a compliant state trajectory. This includes: starting from the initial state diagram, dividing the deduction iteration at each time step into three sequential stages: numerical iteration, rule verification, and branch processing. In the numerical iteration stage, the next time step value of each entity's state is calculated based on the organizational dynamics model to obtain a candidate state set. Based on the candidate state set, in the rule verification stage, hard and soft constraints are invoked to determine the compliance of each candidate state. In the branch processing stage, simulation branches that violate hard constraints are terminated, and states that violate soft constraints are corrected to the feasible region according to the principle of minimum perturbation and potential risk factors are marked. Compliant states that pass the verification are retained. Based on the compliant states retained step by step, a continuous time evolution sequence is generated to obtain the compliant state trajectory.

[0039] Among them, the organizational dynamics model refers to a theoretical analysis framework that uses a systems science perspective to comprehensively examine the nonlinear dynamic process of interaction, evolution, and adaptation between an organization's internal structure, culture, individual behavior, and external environment over time.

[0040] It is understood that the organizational dynamics model in this application is oriented towards the dynamic evolution of interactions between entities within an enterprise, business processes, and risk transmission. Using the attribute parameters of each entity in the compressed state diagram as input, it achieves numerical iterative calculation of the state of each entity at different time steps, outputting a set of candidate states for the next moment. This transforms the operational mechanism of a complex organizational system into a computable dynamic evolution relationship, providing underlying numerical iterative capabilities for forward simulation of the world model. This model can characterize the mutual influence and temporal changes between entities, assets, and business processes, capturing the transmission and diffusion effects of risks along with business processes. In conjunction with the ontology rule verification mechanism, the dynamics model is responsible for state numerical deduction, while the ontology rules are responsible for logical compliance checks. This compensates for the shortcomings of pure ontology reasoning in conducting continuous temporal numerical simulations, providing dynamic numerical calculation support for generating compliant state trajectories, calculating organizational health and operational risk indicators, and constructing risk event chains. This realizes the transformation of enterprise risk from static judgment to continuous dynamic evolution simulation.

[0041] In step S104, candidate interventions are first filtered through ontology compliance verification based on the risk event chain, and their effectiveness is evaluated through parallel counterfactual inference to output the target recommended intervention.

[0042] In risk management, a risk event chain refers to the overall interconnected process in which an initial risk event triggers a series of secondary and derivative events and chain effects through a causal transmission mechanism.

[0043] It is understood that the risk event chain in this application embodiment is generated by mapping compliance status trajectory with ontology semantics. It can connect discrete operational risk indicators and risk factors according to temporal logic and the risk transmission relationship defined by ontology, restore the complete evolution path from the source cause, intermediate transmission nodes to the final risk consequence, and change the traditional risk output mode that only provides isolated risk points. It relies on ontology semantics to realize the traceability and interpretability of risk elements, clearly present the sequence of risk occurrence, the transmission link between entities and the process of impact transmission. This not only facilitates the location of the root cause of the risk, but also provides clear risk tracing basis for the enumeration of candidate intervention strategies. It supports subsequent ontology compliance filtering and parallel counterfactual inference evaluation of intervention measures, so that the intervention plan can accurately target the key nodes in the risk transmission link, effectively improve the pertinence of risk disposal, and make the risk prediction result no longer an abstract indicator, but a traceable risk analysis carrier with complete causal time sequence.

[0044] In this embodiment, candidate interventions are first filtered through ontology compliance verification based on the risk event chain, and their effectiveness is evaluated through parallel counterfactual inference to output the target recommended intervention. This includes: enumerating a set of candidate interventions based on the risk event chain, and performing ontology compliance verification on each candidate intervention according to the compliance ontology. When the execution premise or execution consequence of a candidate intervention violates any hard constraint, it is filtered out, and a subset of candidate interventions that pass the verification is retained. For each candidate intervention in the subset of candidate interventions, parallel counterfactual inference is initiated with the current compliance state trajectory as the baseline. One or more simulation acceleration methods are used, such as baseline state sharing and write-time copy, common prefix trajectory reuse, and early pruning, to calculate the risk indicator reduction and compliance cost after the implementation of the candidate intervention. Based on the risk indicator reduction and compliance cost of each candidate intervention, the ratio of the risk indicator reduction to the compliance cost is used as the comprehensive utility, and the candidate intervention with the largest comprehensive utility is selected as the target recommended intervention and output.

[0045] In this context, the candidate intervention subset refers to a limited set of alternative intervention measures selected from all possible intervention measures in a causal inference or personalized decision-making scenario, based on specific constraints or optimization objectives, for use in subsequent evaluation or final decision-making.

[0046] It is understood that the candidate intervention subset in this application embodiment is a set of valid interventions retained after the compliance ontology has completed compliance verification and filtered out invalid interventions that violate hard constraints. It conducts pre-compliance screening based on all candidate interventions obtained from the risk event chain enumeration, eliminating infeasible solutions in advance that violate business, risk, and compliance hard constraints in terms of execution premises or execution consequences. This avoids meaningless simulation calculations for a large number of invalid interventions in subsequent counterfactual deductions, reducing the task scale and computing power consumption of parallel deductions. At the same time, it ensures that all interventions in the subset meet the compliance bottom line defined by the ontology, avoiding the output of non-compliant and unimplementable intervention suggestions from the source. It provides evaluation objects with controllable boundaries and realistic executability for subsequent counterfactual deductions, so that the quantitative evaluation of the reduction of risk indicators and compliance costs focuses on feasible solutions, ensuring the effectiveness and realistic implementation of comprehensive utility calculation and target recommendation intervention output.

[0047] The following will illustrate a method for enterprise risk prediction based on a world model and ontology constraints through a specific embodiment, including: This embodiment targets a medium-sized software R&D and management entity, comprising 480 business personnel, 23 ongoing projects, 48 ​​business skills, and 4 business departments. This entity frequently faces multiple operational risks, including personnel turnover, skills gaps, imbalanced project resource allocation, and compliance violations. Traditional risk statistical analysis can only statically identify existing risks, failing to simulate the dynamic propagation and evolution of risks as business operations continue, and also struggling to quantify the actual effects of different mitigation measures. This embodiment introduces an enterprise risk prediction method based on a world model and ontology constraints, fully executing the entire process of ontology construction, state coding, event access, simulation, risk analysis, and intervention assessment. The system architecture is as follows: Figure 2 As shown, the overall business execution process is as follows: Figure 3 As shown.

[0048] Figure 2 This is a system architecture diagram provided according to one embodiment of the present application. Figure 2 The system architecture modules include: an enterprise ontology construction module, a state encoding module, a multi-source event interface, a world model simulation module, a health assessment module, a risk prediction module, and a counterfactual inference module. The enterprise ontology construction module outputs data to the state encoding module and constraint projections to the multi-source event interface. The multi-source event interface inputs events to the world model simulation module. The state encoding module outputs to the world model simulation module. The world model simulation module drives the health assessment module and the risk prediction module respectively. The risk prediction module outputs to the counterfactual inference module. The health assessment module provides feedback to the recommended intervention output, and the counterfactual inference module outputs to the recommended intervention output, ultimately yielding the recommended intervention output result.

[0049] Figure 3The overall method flowchart provided according to one embodiment of this application shows the following steps in the overall business execution process: Start → Construct a hierarchical enterprise ontology and rule system → Encode organizational state into a patterned compressed state diagram → Access internal and external multi-source events → Map to state update operations → Forward simulation and deduction of the world model with ontology constraints. After the simulation and deduction, the process splits into two branches: one branch generates state trajectory → calculates the organizational health evolution sequence; the other branch generates state trajectory → calculates enterprise business risk indicators → identifies risk event chains. The risk event chain identification results are sent to candidate intervention compliance verification and generation, and then enter parallel counterfactual deduction → effect evaluation → output recommended intervention.

[0050] The first step is to construct a hierarchical enterprise ontology and rule system. This is done according to the four-tuple model. Complete the enterprise domain ontology definition, among which For concept sets, For attribute collection, For relation sets, It is a set of axioms and rules. The ontology is broken down into business ontology. Risk ontology Compliance ontology The three types of sub-ontologies share the same core entity layer, and the merged ontology is denoted as... , The business ontology represents a set of shared concepts, attributes, and relationships among the three entities. It defines personnel, projects, skills, and departments, configures business constraint rules, and uses SHACL syntax to set a minimum of 3 and a maximum of 20 project team members, ensuring project personnel configuration remains within a reasonable range. Business rules are described using a four-part structure: scope—condition—action—priority. The risk ontology defines risk concepts such as personnel turnover, skills gaps, project delays, and key entity dependencies. It uses SWRL syntax to write risk inference rules, which include triggering conditions, risk level functions, risk propagation operators, and risk decay parameters. The personnel turnover risk calculation logic is triggered when personnel performance scores are low, workloads are high, or salary benchmarks are lower than market benchmarks. The compliance ontology introduces internal company policies and industry regulatory provisions, using OWL2 restricted syntax combined with moral logic operators. Expressing obligations Indicates prohibition The compliance rules indicate permission and include elements such as clause identification, source, scope of application, binding expression, effect, and penalty level. They distinguish between hard and soft constraints, setting the monthly overtime hours of no more than 36 hours as a hard constraint, which cannot be exceeded by any scenario or intervention plan.

[0051] All business rules, risk rules, and compliance rules undergo three levels of validation before being added to the database. Syntax validation completes BNF grammar parsing and intercepts syntax errors. Semantic validation checks whether the entities, attributes, and relationships referenced by the rules are defined in the ontology and verifies the legality of data value domains. Consistency validation detects rule conflicts and redundancy, resolving conflicts in the order of compliance rules having higher priority than risk rules, and risk rules having higher priority than business rules. If conflicts cannot be resolved, a conflict report is output. The three types of sub-ontologies are integrated through terminology alignment mapping, business-to-risk mapping, risk-to-compliance mapping, and compliance-to-business mapping. Business-to-risk mapping transforms business rule violation records into triggering evidence for risk rules. Compliance-to-business mapping compiles hard constraints into the constraint projection layer of the state transition operator, restricting the generated state space from the bottom layer and avoiding unreasonable deduction results from business logic. When the ontology is updated, a fusion consistency check is automatically performed to detect concept satisfiability and mapping loops. If anomalies are found, the change is rolled back directly. The ontology version number is synchronously retained along with the state diagram and deduction results, ensuring full-chain traceability.

[0052] The second step, based on the hierarchical enterprise ontology and rule system, encodes the current organizational state into a patterned compressed state diagram. This diagram then connects to internal and external heterogeneous events through a multi-source event interface, and through semantic matching and mapping, is converted into state updates, resulting in an initial state diagram with ontology semantic constraints. This state diagram is denoted as... , Represents a set of nodes. This represents the set of edges. The node set includes personnel nodes, project nodes, skill nodes, and department nodes. The edge set includes relationship edges such as personnel's department, personnel's project participation, personnel's skills, and project-required skills. Each node and edge carries corresponding business attributes. Personnel node attributes include performance score, workload coefficient, salary benchmark, and turnover tendency value; project node attributes include team size, project progress, budget consumption ratio, and basic value of delay risk.

[0053] Faced with the massive amount of entity data brought about by a scale of 480 personnel, we carried out pattern-based encoding and compression processing. Nodes only store entity type number and attribute value tuples, while edges store relationship type number, source entity number, target entity number and edge attribute tuples. The global entity number is converted into a dense integer through dictionary encoding. The topology relationship adopts CSR sparse row storage. Attribute data is compressed in columnar form. The time-series status adopts a hybrid storage of baseline snapshots and incremental logs. At the memory level, write-on-write replication is enabled, and subgraphs that have not been modified are reused in adjacent time steps, which greatly reduces memory usage and storage overhead.

[0054] The multi-source event interface simultaneously handles events from internal business systems and external industry events. Internal events include business information such as personnel performance updates, project milestone changes, and job adjustments. These are directly matched with ontology entities and relationships, transforming them into state update operations such as modifying node attributes, adding / deleting nodes, and adding / deleting relationship edges. External industry events include unstructured information such as changes in industry salary benchmarks, changes in talent market supply and demand, and updates to regulatory systems. A two-level mapping process is performed. The first level extracts standardized elements of the events based on a large model, completing semantic matching between the events and ontology entities, and identifying the set of entities affected by the events. The second level distinguishes event types and performs state updates. Market salary changes are treated as attribute batch modification events, batch updating the salary benchmark attributes of personnel entities. New regulatory regulations are rule constraint injection events, directly updating the rule entries of the compliance ontology. After incremental verification, the constraints are applied to all subsequent deduction processes. After semantic mapping updates of internal and external events, an initial state diagram carrying complete ontology semantic constraints is finally obtained. .

[0055] The third step is to use the world model in the initial state diagram. The forward dynamics derivation is performed, and the core of the world model is the parameterized state transition operator. The operator mapping relationship is In the formula This is the state diagram at the current moment. For the external context vector, As counterfactual intervention parameters, during normal forward inference Leave blank, This is the set of trainable parameters for a neural network. The operator is decomposed into a personnel state transition sub-operator. Project state transition sub-operators relational evolution sub-operators The personnel and project sub-operators rely on recurrent neural networks to update temporal attributes, while the relationship evolution sub-operator uses a graph attention network to facilitate the propagation and diffusion of risk between entity networks. Internally, each operator performs three layers of processing, with the encoding layer following an ontology model. Encode the state diagram into a tensor Candidate tensors are calculated from the neural evolution layer. via decoder Restore to candidate state diagram The ontology correction layer performs rule verification and correction on the candidate states, and outputs compliant states. The simulation and derivation process of a world model with ontological constraints is as follows: Figure 4 As shown.

[0056] Figure 4 The flowchart for compliance-preliminary counterfactual deduction and intervention recommendation provided according to an embodiment of this application is as follows: Initial state Input is entered, and the numerical iteration phase begins. Then, rule validation is performed: the ontology rule engine is invoked to sequentially execute compliance rule validation → business rule validation. If a hard constraint is violated, the branch terminates; if a soft constraint is violated, it is corrected to the feasible region. After successful validation, it is determined whether N steps have been reached. If N steps have not been reached, the numerical iteration continues; if N steps have been reached, the state trajectory is output. The state trajectory is output in three ways: the first way calculates the evolution sequence of organizational health and outputs the health evolution sequence; the second way calculates the operational risk indicators and outputs the risk distribution; and the third way identifies the risk event chain (ontology semantic mapping) and outputs the risk event chain.

[0057] The derivation iteration executes three unit steps: numerical iteration, rule verification, and branch processing. Numerical iteration calls the state transition operator to complete the state tensor calculation and decodes it to generate a candidate state graph. The rule verification process invokes the ontology rule engine, sequentially executing compliance hard rule verification and business rule verification. The branch processing stage strictly distinguishes between hard and soft constraints. If the simulation result triggers a hard constraint such as monthly overtime exceeding 36 hours, the simulation branch is terminated, and invalid simulation paths are removed. For soft constraint scenarios where the project team size is 2 people, violating the minimum project personnel limit of 3 people, the system performs minimum disturbance correction, adjusting the project team size to 3 people, simultaneously recording anomaly logs, and marking the manpower allocation gap as a risk factor for subsequent risk calculations. Soft constraint conflict scenarios are iteratively corrected according to rule priority. If the iteration reaches a threshold and still cannot completely eliminate violations, the remaining violations are included in the compliance risk indicators, and conflict diagnosis records are retained.

[0058] After multiple rounds of iterative deduction, the future is generated. The compliance status trajectory sequence at each moment Based on the state diagram corresponding to each time point, the evolution sequence of organizational health is calculated. The formula for calculating organizational health is as follows: , Represents the weight of each sub-indicator. represent Each sub-indicator at any given time includes personnel structure rationality, skill coverage, project resource matching, and compliance compliance. The compliance compliance is directly derived from the ontology rule verification results. In this embodiment, the initial simulation shows an overall downward trend in the health sequence. Simultaneously, multi-dimensional operational risk indicators are calculated to determine the risk of personnel turnover. Skills gap risk Project delay risk Key entity dependence risk Risk distribution changes with the simulation cycle. Based on the definition of ontology entity relationships, purely numerical risk indicators are mapped to business semantics to generate a complete and traceable risk event chain. In this embodiment, the typical risk transmission chain is simulated as follows: the tendency for some personnel to leave increases, causing the project team's manpower load to exceed the limit, further increasing the probability of project delay. Project delays will amplify the dependence risk of key entities in turn. Each transmission node in the entire chain is associated with a corresponding ontology rule number, making the risk process explainable.

[0059] The fourth step involves filtering candidate interventions based on the risk event chain through ontology compliance verification, evaluating their effectiveness through parallel counterfactual reasoning, and outputting recommended interventions. Combining the identified risk event chain, multiple sets of candidate interventions are generated. For the risk of employee turnover, three sets of candidate interventions are set: the first set is a 15% salary increase for core personnel; the second set is optimizing workload and reducing ineffective overtime; and the third set is increasing long-term incentive benefits. For the risk of skills gaps, three sets of candidate interventions are set: external recruitment of personnel for corresponding positions, conducting internal business skills training, and outsourcing certain business modules.

[0060] All candidate interventions are first fed into the ontology rule engine for pre-compliance screening to verify whether the intervention will violate hard constraints. Solutions that violate hard constraints, such as causing monthly overtime hours to exceed 36 hours, are directly eliminated. Only interventions that pass the compliance verification are retained for parallel counterfactual inference. The inference process reuses the state transition operators trained in the forward inference. Different intervention programs only modify the initial state. or intervention parameters This ensures the consistency of the underlying logic of the model.

[0061] At the computing power level, a two-level scheduling mechanism is employed. The global scheduler allocates inference branches to compute node groups, while the node-internal scheduler assigns tasks to CPU workgroups and GPU batch inference groups. Combined with lightweight model pre-scoring, reward-based scheduling is implemented. The inference container supports elastic scaling and checkpoint fault tolerance. Simulation acceleration utilizes baseline state write-time copy-and-share. Each intervention branch shares the same read-only copy of the baseline, and the memory overhead is borne by Decrease to For intervention schemes that take effect in the middle of the simulation, the common prefix trajectory before the intervention takes effect is reused to avoid repeated calculations; an early stopping pruning mechanism is introduced, and the confidence intervals of the risk indicators of each branch are compared at intervals of several simulation steps to eliminate simulation paths with significantly worse effects than the optimal branch in advance. In this embodiment, early stopping pruning can eliminate about 70% of inferior candidate branches; at the same time, it is combined with lightweight model pre-screening and multi-fidelity two-level simulation to ensure the reliability of simulation results while controlling computing power consumption.

[0062] After each intervention branch completes its simulation, a completely new state trajectory is obtained after the intervention. The organizational health sequence and various risk indicators after the intervention are calculated. A multi-dimensional effect evaluation is conducted to compare the improvement in health before and after the intervention. The decline in various risk indicators At the same time, the comprehensive implementation cost corresponding to the intervention is calculated, and the cost-benefit ratio is determined.

[0063] The simulation and evaluation data in this embodiment show that the intervention plan of increasing salaries by 15% can reduce the risk of employee turnover from 30% to 10%, while increasing the overall human resource cost by 8%. The internal skills training intervention plan can reduce the skills gap risk from 25% to 15%, with an overall implementation cost of 500,000. Further counterfactual simulations were conducted using the combination of salary increase and internal skills training interventions as candidate solutions. The combined intervention plan can reduce the risk of employee turnover to 8%, the skills gap risk to 12%, and the project delay risk from 40% to 20%. Its overall cost-effectiveness is superior to other single intervention plans, and this combined plan is ultimately recommended as the target intervention output.

[0064] In summary, this invention ensures rigorous logical compliance by constructing multiple integrated ontologies and performing multi-level rule verification. It leverages patterned encoding and multi-source event semantic mapping to achieve efficient processing of massive amounts of business data and restoration of real business states. Through dynamic deduction of a world model with ontology rule correction, it accurately simulates the dynamic propagation process of various risks such as personnel turnover and skill gaps, generating interpretable risk event chains. Finally, through compliance pre-screening and efficient parallel counterfactual deduction, it quantitatively evaluates the cost-effectiveness of multiple single and combined intervention schemes, accurately selecting the optimal risk management solution. This effectively solves the multi-dimensional operational risk management challenges of personnel, projects, resources, and compliance for enterprises, breaking through the limitations of traditional static risk identification. It provides software development companies with a practical, traceable, and highly reliable intelligent risk management method for dynamic risk prediction, risk source tracing and explanation, and scientific selection of optimal intervention schemes.

[0065] Figure 5 This is a schematic diagram of the structure of an enterprise risk prediction system based on a world model and ontology constraints, according to an embodiment of this application.

[0066] like Figure 5 As shown, the enterprise risk prediction system 10 based on world model and ontology constraints includes: ontology construction module 100, state modeling module 200, inference and prediction module 300, and intervention decision module 400.

[0067] The system comprises the following modules: 100, an ontology construction module, and 200, an ontology modeling module, which encodes the current organizational state into a patterned compressed state diagram based on the hierarchical enterprise ontology and rule system. It then connects to internal and external heterogeneous events through a multi-source event interface, maps them to state updates via semantic matching, and obtains an initial state diagram with ontology semantic constraints. 300, a deduction and prediction module, performs forward dynamics deduction on the initial state diagram using a world model. Each iteration embeds ontology rules for real-time verification to correct invalid simulation branches, generating a compliant state trajectory. It also calculates the organizational health evolution sequence and various operational risk indicators, generating a risk event chain through ontology semantic mapping. Finally, 400, an intervention decision module, filters candidate interventions based on the risk event chain through ontology compliance verification, evaluates the effectiveness through parallel counterfactual deduction, and outputs a target recommended intervention.

[0068] It should be noted that the foregoing explanation of an embodiment of an enterprise risk prediction method based on world model and ontology constraints also applies to an enterprise risk prediction system based on world model and ontology constraints in this embodiment, and will not be repeated here.

[0069] This application proposes an enterprise risk prediction system based on a world model and ontology constraints. By constructing a hierarchical enterprise ontology and rule system, it encodes organizational states into a patterned compressed state diagram and updates the semantic mapping of multi-source heterogeneous events. It conducts forward dynamics deduction based on the world model and introduces real-time verification of ontology rules during the iteration process, correcting invalid simulation branches to generate compliant state trajectories. Combined with ontology semantic mapping, it outputs a traceable risk event chain. Simultaneously, it performs pre-emptive ontology compliance filtering and parallel counterfactual deduction evaluation on candidate interventions. This overcomes the shortcomings of static risk identification in failing to predict risk evolution, dynamics deduction lacking business semantics and compliance constraints, intervention schemes easily violating regulations, and shallow splicing of ontology and world models. It balances large-scale scenario storage efficiency with heterogeneous event adaptability, enabling dynamic deduction of enterprise risks, interpretable analysis, and intelligent recommendation of compliant intervention schemes, providing reliable decision support for enterprise operational risk management. Thus, it solves the problems of low large-scale scenario storage efficiency and difficulty in adapting to external heterogeneous events in existing technologies.

[0070] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: The memory 601, the processor 602, and the computer program stored on the memory 601 and capable of running on the processor 602.

[0071] When the processor 602 executes the program, it implements the enterprise risk prediction method based on world model and ontology constraints provided in the above embodiments.

[0072] Furthermore, electronic devices also include: Communication interface 603 is used for communication between memory 601 and processor 602.

[0073] The memory 601 is used to store computer programs that can run on the processor 602.

[0074] The memory 601 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0075] If the memory 601, processor 602, and communication interface 603 are implemented independently, then the communication interface 603, memory 601, and processor 602 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 6 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0076] Optionally, in a specific implementation, if the memory 601, processor 602, and communication interface 603 are integrated on a single chip, then the memory 601, processor 602, and communication interface 603 can communicate with each other through an internal interface.

[0077] The processor 602 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0078] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described enterprise risk prediction method based on a world model and ontology constraints.

[0079] Furthermore, this application also provides a computer program product, including a computer program or instructions, which, when executed, implement the aforementioned enterprise risk prediction method based on a world model and ontology constraints.

[0080] In the description of this specification, the references to the terms "an embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0081] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of those features. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0082] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0083] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by suitable instructions. For example, if implemented in hardware as in another embodiment, it can be implemented using any of the following techniques known in the art, or a combination thereof: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0084] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0085] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for predicting enterprise risk based on a world model and ontology constraints, characterized in that, include: Constructing a hierarchical enterprise ontology and rule system; Based on the hierarchical enterprise ontology and rule system, the current organizational state is encoded into a patterned compressed state diagram, and internal and external heterogeneous events are accessed through a multi-source event interface and mapped to state updates through semantic matching to obtain an initial state diagram with ontology semantic constraints. The world model is used to perform forward dynamics deduction on the initial state diagram. Each iteration embeds ontology rules for real-time verification to correct invalid simulation branches, generate compliant state trajectories, and calculate the organizational health evolution sequence and multiple types of operational risk indicators. Risk event chains are generated through ontology semantic mapping. Based on the risk event chain, candidate interventions are first filtered through ontology compliance verification, and their effectiveness is evaluated through parallel counterfactual inference, resulting in the output of target recommended interventions.

2. The enterprise risk prediction method based on world model and ontology constraints according to claim 1, characterized in that, Constructing a layered enterprise ontology and rules system, including: Establish a shared core entity layer; On top of the shared core entity layer, business ontology, risk ontology and compliance ontology are constructed respectively, and each type of ontology is formally defined using four elements: concept set, attribute set, relation set and axiom set. Based on the formal definitions, business rules, risk rules, and compliance rules are configured for various ontologies; By utilizing the configured business rules, risk rules, and compliance rules, a three-level verification mechanism of syntax verification, semantic verification, and consistency verification is established to complete the construction of the layered enterprise ontology and rule system.

3. The enterprise risk prediction method based on world model and ontology constraints according to claim 2, characterized in that, The business ontology describes the organizational entity, assets, and business processes; the risk ontology describes risk types, transmission relationships, and impact levels; and the compliance ontology describes regulatory provisions, control points, and responsible entities. The business rules use SHACL syntax, the risk rules use SWRL syntax, and the compliance rules use OWL2 restricted syntax, and the rules are divided into hard constraints and soft constraints. The consistency check is used to detect conflicts between different rules and resolve them according to a preset priority.

4. The enterprise risk prediction method based on world model and ontology constraints according to claim 1, characterized in that, The initial state graph with ontology semantic constraints is obtained, including: The entity extraction and ontology alignment of the main entities, assets, business processes and risk matters within the organization are performed. Redundant dimensions are eliminated with the shared core entity layer as a reference, resulting in a compressed state diagram with entities as nodes and ontology relationships as edges. Based on the compressed state graph, internal business system events and external public information events are accessed through a multi-source event interface. After semantic matching, entity linking and relationship extraction, the events are mapped to the corresponding nodes and edges of the compressed state graph and superimposed in chronological order to complete the state update. The updated compressed state graph is validated using a three-level validation mechanism of syntax validation, semantic validation, and consistency validation. Mapping results that conflict with the hierarchical enterprise ontology are removed, resulting in an initial state graph with ontology semantic constraints.

5. The enterprise risk prediction method based on world model and ontology constraints according to claim 1, characterized in that, The world model is used to perform forward dynamics derivation on the initial state diagram. Each iteration embeds ontology rules for real-time verification to correct invalid simulation branches and generate compliant state trajectories, including: Starting from the initial state diagram, the deduction iteration of each time step is divided into three stages: numerical iteration, rule verification, and branch processing, which are executed sequentially. In the numerical iteration stage, the next time value of each entity state is calculated based on the organizational dynamics model to obtain a candidate state set. Based on the candidate state set, in the rule verification stage, hard constraints and soft constraints are invoked to judge the compliance of each candidate state. In the branch processing stage, simulation branches that violate hard constraints are terminated, states that violate soft constraints are corrected to the feasible region according to the principle of minimum perturbation and potential risk factors are marked, and compliant states that pass the verification are retained. Based on the compliance state retained step by step, a continuous time evolution sequence is generated to obtain the compliance state trajectory.

6. The enterprise risk prediction method based on world model and ontology constraints according to claim 1, characterized in that, Based on the aforementioned risk event chain, candidate interventions are first filtered through ontology compliance verification, and their effectiveness is evaluated through parallel counterfactual inference. The resulting recommended interventions are then output, including: Based on the risk event chain, a set of candidate interventions is enumerated, and each candidate intervention is verified for compliance according to the compliance ontology. When the execution premise or execution consequence of a candidate intervention violates any hard constraint, it is filtered out, and a subset of candidate interventions that pass the verification is retained. For each candidate intervention in the subset of candidate interventions, parallel counterfactual simulation is initiated with the current compliance status trajectory as the baseline. One or more simulation acceleration methods are used, including baseline state sharing and write-time copy, common prefix trajectory reuse, and early stop pruning, to obtain the risk index reduction and compliance cost after the implementation of the candidate intervention. Based on the decrease in risk indicators and compliance costs of each candidate intervention, the ratio of the decrease in risk indicators to compliance costs is used as the comprehensive utility, and the candidate intervention with the largest comprehensive utility is selected as the target recommended intervention and output.

7. A business risk prediction system based on a world model and ontology constraints, characterized in that, include: The ontology construction module is used to build a layered enterprise ontology and rule system; The state modeling module is used to encode the current organizational state into a patterned compressed state diagram based on the hierarchical enterprise ontology and rule system, and to access internal and external heterogeneous events through a multi-source event interface, and to map them into state updates through semantic matching, so as to obtain an initial state diagram with ontology semantic constraints. The deduction and prediction module is used to perform forward dynamic deduction on the initial state diagram using the world model. Each iteration embeds ontology rules for real-time verification to correct invalid simulation branches, generates compliant state trajectories, and calculates the organizational health evolution sequence and multiple types of operational risk indicators, and generates risk event chains through ontology semantic mapping. The intervention decision module is used to filter candidate interventions based on the risk event chain, first through ontology compliance verification, then through parallel counterfactual inference to evaluate the effect, and finally output the target recommended intervention.

8. An electronic device, characterized in that, The invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the program to implement the enterprise risk prediction method based on a world model and ontology constraints as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When a computer program or instruction is executed, it implements the enterprise risk prediction method based on a world model and ontology constraints as described in any one of claims 1-6.

10. A computer program product, comprising a computer program or instructions, characterized in that, When a computer program or instruction is executed, it implements the enterprise risk prediction method based on a world model and ontology constraints as described in any one of claims 1-6.

Images