A method and system for intelligent decision-making of predictive personnel transfer based on multi-hypothesis evidence deduction

Abstract

The application discloses a kind of based on multiple hypothesis evidence deduction's foresighted personnel transfer intelligent decision method and system, the method includes: obtaining the evidence input information corresponding to each space unit, the hierarchical information is converted into the initial transfer decision evidence related to personnel transfer;Initial transfer decision evidence is given corresponding uncertainty representation, and is uniformly included in evidence life cycle management process;For the dynamic updating and convergence of effective state evidence, form multiple-source evidence set;Multi-source evidence set is carried out fusion calculation and consistency check, generates corresponding foresighted personnel transfer decision result;Based on multiple hypothesis evidence deduction mechanism, construct multiple reasonable evidence state hypothesis, consistency and stability evaluation is carried out to foresighted personnel transfer decision result, confirms final personnel transfer decision result.The present application solves the problems, such as the risk classification result in the existing personnel transfer decision technology is difficult to directly convert into executable decision basis, etc.

Description

Technical Field

[0001] This invention relates to the field of disaster emergency decision-making technology, and in particular to a predictive intelligent decision-making method and system for personnel evacuation based on multi-hypothesis evidence deduction. Background Technology

[0002] In the event of natural disasters such as floods, timely and scientific evacuation of people is the core means of reducing casualties. Existing disaster early warning and evacuation technologies can be mainly divided into four categories: The first category comprises comprehensive watershed disaster prevention and early warning schemes based on multi-source data, such as the Chinese patent document with application number "202510412197.8". This patent focuses on watersheds, dividing them into grids and sub-regions, constructing runoff generation models and critical rainfall models based on rainfall, topographic and geological features, and then providing disaster early warnings for different regions and comprehensively assessing the overall risk level of the watershed. Its core technical path is "multi-source monitoring data → runoff generation calculation → threshold discrimination → risk classification and early warning". This method emphasizes hydrophysical mechanisms and spatial overlay analysis, and is suitable for disaster monitoring and early warning management at the watershed scale. However, its decision output mainly remains at the level of risk level and alarm triggering, without explicitly modeling the management decision of "whether to implement personnel evacuation", nor does it characterize the uncertainty introduced by the lead time for prediction, nor does it address how the risk classification results are transformed into fusionable decision evidence. Therefore, it is difficult to support predictable and interpretable personnel evacuation decisions. This patent mainly focuses on risk classification and early warning judgment based on rainfall, topography and runoff generation mechanism. Its decision-making logic is based on critical rainfall and water volume threshold, focusing on judging the probability of disaster occurrence. However, it does not model how the risk classification results are transformed into specific management decisions (especially whether to transfer personnel), nor does it explicitly characterize the uncertainty changes brought about by the lead time of prediction. As a result, the early warning results are difficult to directly support predictable and explainable personnel transfer decisions.

[0003] The second category comprises basin-wide flood control and drought relief coordination schemes based on precipitation forecast big data. For example, Chinese patent application number "202511291428.0" describes a system that constructs an end-to-end collaborative prevention and control and scheduling system based on precipitation forecast big data. Through generative precipitation forecasting models, multi-timescale risk quantification and dynamic correction mechanisms, combined with reservoir, river channel, and soil conditions, it achieves joint peak-shaving scheduling and emergency resource allocation across the entire basin. Its key technologies lie in big data processing, intelligent precipitation forecasting, and engineering scheduling optimization. This scheme, driven by physical quantities and threshold criteria, directly derives scheduling instructions from precipitation and hydrological conditions, emphasizing engineering control and resource scheduling efficiency. However, it treats personnel relocation as merely an auxiliary output of the scheduling results, lacking an evidence expression mechanism for the "relocation / non-relocation" decision itself, and failing to structurally model the conflicts and uncertainties between knowledge rules, case experience, and model results. This existing technology focuses on precipitation forecasting and hydrological conditions, generating scheduling instructions through multi-timescale risk indices and engineering scheduling rules. It emphasizes physical quantity-driven and scheduling efficiency, but its decision-making model is highly dependent on model output and threshold triggering. It does not independently model the personnel transfer decision itself, nor does it solve the conflict and credibility issues between different model results and knowledge experience. It is difficult to provide stable and interpretable transfer decision support under uncertain forecast conditions.

[0004] The third category is general conflict evidence fusion algorithms, such as the Chinese patent document with application number "CN202211132564.1". This patent belongs to a general evidence fusion method, mainly targeting the credibility and uncertainty issues of multi-source conflict evidence. It calculates the trust factor, divergence, belief entropy, and information weight of the evidence vector, performs a weighted average of the evidence, and outputs the final decision result using the DS combination rule. Its innovation lies in the evidence weight correction and conflict suppression mechanism. This method is relatively complete at the algorithm level, but its application scenario is represented by automotive system fault diagnosis. It belongs to the paradigm of "constructed evidence → optimized fusion rules", and does not involve the evidence generation process, let alone consider issues such as risk classification, prediction lead time, or decision timeliness. The existing technology focuses on the weight correction and fusion rule improvement among constructed evidence, mainly to solve the problem of the impact of evidence conflict on the fusion result. However, its application premise is that the evidence has been clearly given. It does not involve the evidence generation mechanism in engineering decision-making scenarios such as risk classification and timeliness changes, nor can it express the impact of prediction lead time on decision uncertainty. Therefore, it is difficult to directly apply to decision-making problems with strong timeliness and predictability requirements, such as disaster personnel transfer.

[0005] The fourth category is dynamic personnel evacuation early warning schemes for flood storage and detention areas. For example, Chinese patent document with application number "202511398733.X" uses flood storage and detention areas as a typical scenario. It integrates hydrological monitoring, population distribution, geographic information, and flood evolution models in real time to generate a gridded real-time risk level. Based on this, it triggers graded early warnings, plans the optimal evacuation path, and pushes evacuation information to individual terminals. Its core technology lies in the dynamic early warning and execution closed loop of "real-time perception + flood simulation + path planning". This scheme relies heavily on physical simulation results such as water depth and flow velocity. The decision-making logic is mainly based on risk threshold triggering and path optimization. Although it solves the problem of "how to evacuate and where to evacuate", it does not answer the decision-making problem of "whether evacuation should be carried out in advance under uncertain prediction conditions", nor does it construct an evidence expression mechanism that integrates risk classification and knowledge experience. This patent triggers early warnings for population evacuation and plans evacuation routes through flood evolution simulation and real-time population distribution analysis, focusing on solving the problems of "how to evacuate" and "optimizing evacuation routes". However, its decision-making basis mainly comes from real-time or short-term physical simulation results, lacking a comprehensive expression of evidence on risk classification results, knowledge and experience and prediction uncertainty, making it difficult to support predictive population evacuation decisions under conditions of long lead time. Summary of the Invention

[0006] This invention provides a predictive intelligent decision-making method and system for personnel relocation based on multi-hypothesis evidence deduction, aiming to solve the problems that are common in existing personnel relocation decision-making technologies, such as the difficulty in directly converting risk classification results into actionable decision-making basis, the lack of systematic characterization of the uncertainty introduced by the lead time for prediction, the difficulty in synergistic use of model evaluation results and knowledge experience from different sources, and the lack of stability and predictability of relocation decisions under complex and uncertain conditions.

[0007] This invention provides a predictive intelligent decision-making method for personnel transfer based on multi-hypothesis evidence deduction, the method comprising the following steps: Step S10: Using spatial units as the basic decision-making objects, obtain the evidence input information corresponding to each spatial unit. The evidence input information includes hazard classification, vulnerability classification, and personnel transfer knowledge input information. Based on the preset classification mapping rules, the classification information is transformed into initial transfer decision evidence related to personnel transfer. Step S20: Combining the prediction lead information, assign corresponding uncertainty characteristics to the initial transfer decision evidence and uniformly incorporate it into the evidence lifecycle management process; Step S30: Dynamically update and aggregate the evidence of valid states from different spatial units to form a multi-source evidence set for personnel transfer decisions; Step S40: Perform fusion calculation and consistency verification on the multi-source evidence set to generate corresponding predictive personnel transfer decision results; Step S50: Based on the multi-hypothesis evidence deduction mechanism, while keeping the current environmental constraints unchanged, construct a variety of reasonable evidence state hypotheses, evaluate the consistency and stability of the predictive personnel transfer decision under different evidence hypothesis conditions, and confirm the final personnel transfer decision based on the evaluation results. Step S60: Output the corresponding personnel transfer decision instruction based on the final personnel transfer decision result. At the same time, continuously monitor the decision execution effect and subsequent data changes, and re-incorporate the relevant feedback information into the evidence lifecycle management process for subsequent evidence updates and decision optimization.

[0008] A further technical solution of the present invention is that step S10 includes: Step S101: Convert the evidence input information into standardized evidence objects, and configure each evidence object with evidence identifier, source information, timestamp, uncertainty parameter, weight parameter and status flag field to realize the object-oriented encapsulation of evidence; Step S102: Based on the pre-defined level mapping rules, map the level information to the corresponding initial value of the transfer tendency. Step S20 includes: Step S201: Based on the initial value of the transfer tendency, the uncertainty of the evidence object is assigned and corrected in combination with the prediction lead time to form an evidence state with time-sensitivity attribute; Step S202: Perform core calculation and update operations on the evidence object, identify graded changes, invalidation changes or feedback anomalies through the change detection mechanism, and trigger evidence update, decay or reconstruction. Step S203: Output the current set of valid evidence and its metadata index information, and generate a complete evidence version chain and state transition record to support the analysis of decision interpretability and subsequent audit traceability, providing basic support for the formation of a closed-loop decision-making mechanism driven by the evidence lifecycle.

[0009] A further technical solution of the present invention is that step S30 includes: Step S301: Obtain the set of valid evidence, extract the transfer tendency, uncertainty, weight and version information of each evidence object, and perform consistent representation of evidence from different sources and scales through evidence normalization processing, while initially screening out missing fields, expired evidence or outlier evidence. Step S302: Aggregate and group the evidence objects according to spatial units, time windows or decision-making scenarios to form corresponding multi-source evidence subsets, and identify the diversity and redundancy of the sources of evidence in each subset. Step S303: Generate fusion results based on evidence support, uncertainty parameters and weight information, output the corresponding candidate transfer decision evidence, and simultaneously calculate the fusion credibility index. Step S40 includes: Step S401: Perform consistency verification and conflict analysis on the fusion results. Identify high-conflict or low-consistency scenarios by calculating the consistency or conflict degree between evidence, and generate fusion diagnostic information and corresponding predictive personnel transfer decision results. The fusion diagnostic information includes at least fusion consistency, uncertainty level and conflict marker. Specifically, when consistency is acceptable, the fusion result is retained and the result is corrected by increasing uncertainty or adjusting weights; when consistency is insufficient or conflict is unacceptable, abnormal evidence is downweighted or removed, or the fusion result is marked as uncertain.

[0010] A further technical solution of the present invention is that step S50 includes: Step S501: Predict the outcome of the anticipated personnel transfer decision, and determine whether the uncertainty exceeds a preset threshold and whether there is a conflict or anomaly in the fusion result, so as to determine whether it is necessary to enter the multi-hypothesis deduction path; when the uncertainty is low and the fusion result is consistent, skip the deduction process and directly enter the decision confirmation stage; when the uncertainty is high or there is a conflict, determine the number of hypotheses and deduction dimensions required for the deduction, and construct a variety of reasonable evidence state hypotheses under the premise of meeting business rules and constraints. Step S502: Based on uncertainty disturbances, evidence missing simulations, and low-credibility evidence replacement methods, an evidence state set is formed, and the outcome of the predictive personnel transfer decision is evaluated under each evidence state to obtain the corresponding hypothetical decision outcome. Step S503: Compare and analyze the hypothetical decision results under each hypothetical condition, calculate the decision consistency ratio, decision change magnitude and sensitivity to evidence disturbance, and classify the decision stability into stable, substable or unstable levels accordingly. Step S504: For different stability levels, implement corresponding decision processing strategies: For sub-stable decisions, increase decision uncertainty and add risk warnings; for unstable decisions, mark them as requiring manual review or postpone execution; for stable decisions, confirm the final personnel transfer decision results and generate decision explanation information including a summary of the deduction process, the number of assumptions, and the stability assessment conclusions, to provide a basis for decision issuance, execution, and auditing.

[0011] A further technical solution of the present invention is that step S60 includes: Step S601: The final personnel transfer decision result is converted into a final personnel transfer decision instruction. Based on the preset instruction parsing rules and the decision interpretation information, the final personnel transfer decision instruction is parsed into a standard execution instruction format and sent to the corresponding execution terminal. Step S602: Obtain the execution status information returned by the execution terminal, monitor whether the instruction was successfully issued, whether it was responded to, and the execution progress, and encapsulate the collected execution status data in a structured manner to form an execution feedback package. The execution feedback package includes at least information such as instruction identifier, execution status, response time, and anomaly marker, and is sent back to the evidence lifecycle management module to update the status, uncertainty parameters, or validity markers of relevant evidence objects, thereby realizing a closed-loop linkage between decision generation, execution feedback, and evidence update.

[0012] To achieve the above objectives, the present invention also proposes a predictive intelligent decision-making system for personnel transfer based on multi-hypothesis evidence deduction. The system includes an evidence construction and lifecycle management module, a multi-source evidence fusion and consistency constraint module, and a multi-hypothesis evidence deduction and decision output module. The evidence construction and lifecycle management module is used to obtain evidence input information corresponding to each spatial unit, taking spatial units as the basic decision objects. The evidence input information includes hazard classification, vulnerability classification, and personnel transfer knowledge input information. Based on the preset classification mapping rules, the classification information is transformed into initial transfer decision evidence related to personnel transfer. The evidence construction and lifecycle management module is also used to combine the prediction lead information to assign corresponding uncertainty characteristics to the initial transfer decision evidence and uniformly incorporate it into the evidence lifecycle management process. The multi-source evidence fusion and consistency constraint module is used to dynamically update and aggregate evidence from valid states of different spatial units to form a multi-source evidence set for personnel transfer decisions. The multi-source evidence fusion and consistency constraint module is also used to perform fusion calculation and consistency verification on the multi-source evidence set to generate corresponding predictive personnel transfer decision results. The multi-hypothesis evidence deduction and decision output module is used to construct a variety of reasonable evidence state hypotheses based on the multi-hypothesis evidence deduction mechanism, while keeping the current environmental constraints unchanged, to evaluate the consistency and stability of the predictive personnel transfer decision results under different evidence hypothesis conditions, and to confirm the final personnel transfer decision results based on the evaluation results. The multi-hypothesis evidence deduction and decision output module is also used to output corresponding personnel transfer decision instructions based on the final personnel transfer decision results. At the same time, it continuously monitors the decision execution effect and subsequent data changes, and re-incorporates relevant feedback information into the evidence lifecycle management process for subsequent evidence updates and decision optimization.

[0013] A further technical solution of the present invention is that the evidence construction and lifecycle management module is also used to convert the evidence input information into standardized evidence objects, and configure each evidence object with evidence identifier, source information, timestamp, uncertainty parameter, weight parameter and status flag field to realize the object-oriented encapsulation of evidence; Based on the pre-defined level mapping rules, the level information is mapped to the corresponding initial value of the transfer tendency. The multi-source evidence fusion and consistency constraint module is also used to assign and correct the uncertainty of the evidence object based on the initial value of the transfer tendency and the prediction lead, so as to form an evidence state with time-sensitive attributes. Perform core calculations and update operations on the evidence object, identify graded changes, invalid changes or feedback anomalies through the change detection mechanism, and trigger evidence updates, decay or reconstruction; It outputs the current set of valid evidence and its metadata index information, and generates a complete evidence version chain and state transition record to support the analysis of decision interpretability and subsequent audit traceability, providing basic support for the formation of a closed-loop decision-making mechanism driven by the evidence lifecycle.

[0014] A further technical solution of the present invention is that the multi-source evidence fusion and consistency constraint module is also used to obtain the effective evidence set, extract the transfer tendency, uncertainty, weight and version information of each evidence object, and perform consistent representation of evidence from different sources and at different scales through evidence normalization processing, while initially screening out missing fields, expired evidence or outlier evidence. Evidence objects are aggregated and grouped according to spatial units, time windows, or decision-making scenarios to form corresponding multi-source evidence subsets, and the diversity and redundancy of the sources of evidence within each subset are identified. The fusion result is generated based on the evidence support, uncertainty parameters and weight information, and the corresponding candidate transfer decision evidence is output. The fusion credibility index is calculated simultaneously. Consistency verification and conflict analysis are performed on the fusion results. High-conflict or low-consistency scenarios are identified by calculating the consistency or conflict degree between evidence, and fusion diagnostic information and corresponding predictive personnel transfer decision results are generated. The fusion diagnostic information includes at least fusion consistency, uncertainty level and conflict marker. Specifically, when consistency is acceptable, the fusion result is retained and the result is corrected by increasing uncertainty or adjusting weights; when consistency is insufficient or conflict is unacceptable, abnormal evidence is downweighted or removed, or the fusion result is marked as uncertain.

[0015] A further technical solution of the present invention is that the multi-hypothesis evidence deduction and decision output module is also used to predict the outcome of the predictive personnel transfer decision, and to determine whether the uncertainty exceeds a preset threshold and whether there is a conflict or anomaly in the fusion result, thereby determining whether it is necessary to enter the multi-hypothesis deduction path; when the uncertainty is low and the fusion result is consistent, the deduction process is skipped and the decision confirmation stage is directly entered; when the uncertainty is high or there is a conflict, the number of hypotheses and the deduction dimension required for the deduction are determined, and under the premise of meeting the business rules and constraints, a variety of reasonable evidence state hypotheses are constructed. Based on uncertainty disturbances, evidence deficiency simulation, and low-reliability evidence replacement methods, an evidence state set is formed, and the results of the predictive personnel transfer decision are evaluated under each evidence state to obtain the corresponding hypothetical decision results. The hypothetical decision-making results under various assumptions are compared and analyzed to calculate the decision consistency rate, decision change magnitude, and sensitivity to evidence disturbances. Based on this, the decision stability is classified into stable, substable, or unstable levels. For different stability levels, corresponding decision-making strategies are implemented: for sub-stable decisions, decision uncertainty is increased and risk warnings are attached; for unstable decisions, they are marked as requiring manual review or postponed; for stable decisions, the final personnel transfer decision result is confirmed, and decision explanation information including a summary of the deduction process, the number of assumptions, and the stability assessment conclusions is generated to provide a basis for decision issuance, execution, and auditing.

[0016] A further technical solution of the present invention includes a decision execution and feedback module, which is used to convert the final personnel transfer decision result into a final personnel transfer decision instruction, and based on the preset instruction parsing rules and the decision interpretation information, parse the final personnel transfer decision instruction into a standard execution instruction format and send it to the corresponding execution terminal; The system acquires the execution status information returned by the execution terminal, monitors whether the instruction was successfully issued, whether it was responded to, and the execution progress, and encapsulates the collected execution status data into an execution feedback package. The execution feedback package includes at least the instruction identifier, execution status, response time, and anomaly marker, and is sent back to the evidence lifecycle management module to update the status, uncertainty parameters, or validity markers of the relevant evidence objects, thereby achieving a closed-loop linkage between decision generation, execution feedback, and evidence updating.

[0017] Compared to existing personnel transfer decision-making technologies that generally rely on single risk indicators, static rules, or one-off judgments, this invention effectively overcomes the problems of unstable, unreliable, and unpredictable decision results in complex and uncertain environments by introducing an evidence-based multi-stage decision-making mechanism. This invention models the results of risk classification, vulnerability classification, and knowledge recommendation using evidence, and introduces an evidence lifecycle management mechanism to dynamically control the timeliness, uncertainty, and validity of decision evidence, avoiding interference from outdated information or low-reliability evidence. Simultaneously, this invention uses a multi-source evidence fusion and consistency constraint mechanism to uniformly aggregate and analyze decision evidence from different sources, significantly reducing the risk of misjudgment due to evidence conflicts while ensuring full utilization of information. Furthermore, this invention introduces a multi-hypothesis evidence deduction mechanism to evaluate the stability of candidate personnel transfer decisions under various reasonable evidence states, making the decision results independent of a single evidence state, thereby significantly improving the robustness and reliability of personnel transfer decisions under uncertain conditions. Furthermore, by constructing a closed-loop mechanism for decision execution and feedback, this invention achieves dynamic linkage between decision results and execution effectiveness, providing a basis for updating evidence status and optimizing subsequent decisions, thereby enhancing the system's engineering deployability and continuous operation capability. In summary, this invention can provide more robust, interpretable, and predictive personnel transfer decision support in complex and ever-changing emergency scenarios, demonstrating significant technological advancements and practical application value. Attached Figure Description

[0018] Figure 1 A schematic diagram of the overall process of a preferred embodiment of the intelligent decision-making method for predictive personnel transfer based on multi-hypothesis evidence deduction of the present invention; Figure 2 This is a detailed flowchart of step S10; Figure 3 This is a detailed flowchart of step S20; Figure 4 This is a detailed flowchart of step S30; Figure 5 This is a detailed flowchart of step S50; Figure 6 This is a detailed flowchart of step S60; Figure 7 This is the overall architecture diagram of the intelligent decision-making system for predictive personnel transfer based on multi-hypothesis evidence deduction of the present invention; Figure 8 This is a schematic diagram of the operation of the evidence lifecycle management module; Figure 9 This is a schematic diagram of the operation of the multi-source evidence fusion and consistency constraint module; Figure 10This is the flowchart of the multi-hypothesis evidence deduction module. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0020] This invention proposes a predictive intelligent decision-making method for personnel relocation based on multi-hypothesis evidence deduction. The technical solution employed in this invention primarily uses hazard classification, vulnerability classification, and personnel relocation knowledge recommendation results as inputs. By constructing evidence objects oriented towards personnel relocation and their lifecycle management mechanism, the classification information and knowledge recommendation results are uniformly transformed into decision evidence with timeliness and uncertainty characteristics. Based on this, a multi-hypothesis evidence deduction mechanism is introduced. While keeping the current environmental constraints unchanged, multiple reasonable evidence state hypotheses are constructed, and the performance of candidate personnel relocation decisions under different hypothetical conditions is comprehensively evaluated. This allows for the selection and confirmation of predictive personnel relocation decision results with high stability and robustness. Through the above technical solution, this invention realizes the transformation of personnel relocation decision-making from static rule-based judgment to predictive intelligent decision-making with deductive capabilities, continuous updates, and interpretability, providing more scientific and reliable support for emergency command in personnel relocation decision-making.

[0021] Specifically, such as Figure 1 As shown, a preferred embodiment of the predictive intelligent decision-making method for personnel transfer based on multi-hypothesis evidence deduction of the present invention includes the following steps: Step S10: Using spatial units as the basic decision-making objects, obtain the evidence input information corresponding to each spatial unit. The evidence input information includes hazard classification, vulnerability classification, and personnel transfer knowledge input information. Based on the preset classification mapping rules, the classification information is transformed into initial transfer decision evidence related to personnel transfer.

[0022] Step S20: Combining the prediction lead information, assign corresponding uncertainty characteristics to the initial transfer decision evidence and uniformly incorporate it into the evidence lifecycle management process.

[0023] This embodiment of the predictive intelligent decision-making method for personnel relocation based on multi-hypothesis evidence deduction first receives hazard classification, vulnerability classification, and personnel relocation knowledge input information corresponding to each spatial unit during operation. Based on preset classification mapping rules, it transforms the classification information into initial decision evidence related to personnel relocation. After generating the initial evidence, the system combines predictive lead time information to assign corresponding uncertainty characteristics to the decision evidence and incorporates it into the evidence lifecycle management process to characterize the credible state of the evidence as it evolves over time.

[0024] Step S30: Dynamically update and aggregate the evidence of valid states from different spatial units to form a multi-source evidence set for personnel transfer decisions.

[0025] Step S40: Perform fusion calculation and consistency verification on the multi-source evidence set to generate corresponding predictive personnel transfer decision results.

[0026] This embodiment continuously maintains valid evidence. When the grading results, prediction conditions, or knowledge inputs change, the corresponding evidence is dynamically updated, attenuated, or withdrawn. Valid evidence from different spatial units is aggregated to form a multi-source evidence set for personnel relocation decisions. Based on this, the system performs fusion calculations and consistency checks on the multi-source evidence set, comprehensively considering the support strength and uncertainty information of various types of evidence to generate corresponding predictive personnel relocation decision results.

[0027] Step S50: Based on the multi-hypothesis evidence deduction mechanism, while keeping the current environmental constraints unchanged, construct a variety of reasonable evidence state hypotheses, evaluate the consistency and stability of the predictive personnel transfer decision under different evidence hypothesis conditions, and confirm the final personnel transfer decision based on the evaluation results.

[0028] Step S60 outputs the corresponding personnel transfer decision instruction based on the final personnel transfer decision result. At the same time, it continuously monitors the decision execution effect and subsequent data changes, and re-incorporates the relevant feedback information into the evidence lifecycle management process for subsequent evidence updates and decision optimization.

[0029] To assess the stability and rationality of the decision-making results under uncertain conditions, this embodiment introduces a multi-hypothesis evidence deduction mechanism. While keeping the current environmental constraints unchanged, it constructs multiple reasonable evidence state hypotheses and evaluates the consistency and stability of the candidate personnel transfer decision under different evidence hypothesis conditions. After completing the multi-hypothesis evidence deduction evaluation, the final personnel transfer decision result is confirmed, and the corresponding personnel transfer decision instruction is output. Simultaneously, the decision execution effect and subsequent data changes are continuously monitored, and relevant feedback information is reintegrated into the evidence lifecycle management process for subsequent evidence updates and decision optimization, thereby forming a continuously iterative, adaptive, and predictive intelligent decision-making operation mechanism for personnel transfer.

[0030] Furthermore, such as Figure 2 As shown, in this embodiment, step S10 specifically includes the following steps: Step S101: Convert the evidence input information into standardized evidence objects, and configure each evidence object with evidence identifier, source information, timestamp, uncertainty parameter, weight parameter and status flag field to realize the object-oriented encapsulation of evidence.

[0031] Step S102: Based on the pre-defined level mapping rules, the level information is mapped to the corresponding initial value of the transfer tendency.

[0032] like Figure 3 As shown, step S20 includes the following steps: Step S201: Based on the initial value of the transfer tendency, the uncertainty of the evidence object is assigned and corrected by combining the prediction lead amount to form an evidence state with time-sensitivity attribute.

[0033] Step S202: Perform core calculation and update operations on the evidence object, identify graded changes, invalidation changes or feedback anomalies through the change detection mechanism, and trigger evidence updates, decay or reconstruction.

[0034] This embodiment implements automatic migration control for evidence objects between states such as "generated, valid, decayed, withdrawn or invalid". Evidence objects in the valid state are allowed to participate in subsequent evidence fusion and decision deduction, while evidence objects in the decayed or withdrawn state will have their weight reduced or be removed from the decision-making process.

[0035] Step S203: Output the current set of valid evidence and its metadata index information, and generate a complete evidence version chain and state transition record to support the analysis of decision interpretability and subsequent audit traceability, providing basic support for the formation of a closed-loop decision-making mechanism driven by the evidence lifecycle.

[0036] After completing the evidence status update, this embodiment outputs the current valid evidence set and its metadata index information, and generates a complete evidence version chain and state transition record to support decision interpretability analysis and subsequent audit traceability, and provides basic support for the system to form a closed-loop decision-making mechanism driven by the evidence lifecycle.

[0037] Furthermore, such as Figure 4 As shown, in this embodiment, step S30 specifically includes the following steps: Step S301: Obtain the set of valid evidence, extract the transfer tendency, uncertainty, weight and version information of each evidence object, and perform evidence normalization processing to make evidence from different sources and scales consistent, while initially filtering out missing fields, expired evidence or outlier evidence.

[0038] Step S302: Aggregate and group the evidence objects according to spatial units, time windows or decision-making scenarios to form corresponding multi-source evidence subsets, and identify the diversity and redundancy of the sources of evidence in each subset.

[0039] Step S303: Generate fusion results based on evidence support, uncertainty parameters and weight information, output the corresponding candidate transfer decision evidence, and simultaneously calculate the fusion credibility index.

[0040] Step S40 includes: Step S401: Perform consistency verification and conflict analysis on the fusion results. Identify high-conflict or low-consistency scenarios by calculating the consistency or conflict degree between evidence, and generate fusion diagnostic information and corresponding predictive personnel transfer decision results. The fusion diagnostic information includes at least fusion consistency, uncertainty level, and conflict marker.

[0041] Specifically, when consistency is acceptable, the fusion result is retained and the result is corrected by increasing uncertainty or adjusting weights; when consistency is insufficient or conflict is unacceptable, abnormal evidence is downweighted or removed, or the fusion result is marked as uncertain.

[0042] This embodiment performs unified processing on valid evidence objects to generate candidate transfer decision results with reliability constraints.

[0043] First, a set of valid evidence objects is received. Transfer tendency, uncertainty, weight, and version information are extracted from each evidence object. Evidence normalization is then performed to ensure consistency across different sources and scales, while missing fields, expired evidence, or outlier evidence are initially filtered out. After normalization, evidence objects are aggregated and grouped according to spatial units, time windows, or decision scenarios to form corresponding multi-source evidence subsets. The diversity and redundancy of evidence sources within each subset are identified. Subsequently, the fusion calculation unit generates fusion results based on evidence support, uncertainty parameters, and weight information, outputting corresponding candidate transfer decision evidence and simultaneously calculating the fusion credibility index. To avoid instability in decision results caused by conflicts between multi-source evidence, consistency verification and conflict analysis are further performed on the fusion results. High-conflict or low-consistency scenarios are identified by calculating the degree of consistency or conflict between evidence, and fusion diagnostic information is generated. If consistency is acceptable, the fusion results are retained, and the results are corrected by increasing uncertainty or adjusting weights. If consistency is insufficient or conflict is unacceptable, outlier evidence is downweighted or removed, or the fusion results are marked as uncertain. Finally, the candidate personnel transfer decision results and corresponding fusion diagnostic information are output separately, providing structured input for subsequent stability assessment.

[0044] Furthermore, such as Figure 5 As shown, in this embodiment, step S50 specifically includes the following steps: Step S501: Predict the outcome of the anticipated personnel transfer decision, and determine whether the uncertainty exceeds a preset threshold and whether there is a conflict or anomaly in the fusion result, so as to determine whether it is necessary to enter the multi-hypothesis deduction path; when the uncertainty is low and the fusion result is consistent, skip the deduction process and directly enter the decision confirmation stage; when the uncertainty is high or there is a conflict, determine the number of hypotheses and deduction dimensions required for the deduction, and construct a variety of reasonable evidence state hypotheses under the premise of meeting business rules and constraints.

[0045] Step S502: Based on uncertainty disturbances, evidence missing simulations, and low-credibility evidence replacement methods, an evidence state set is formed, and the outcome of the predictive personnel transfer decision is evaluated under each evidence state to obtain the corresponding hypothetical decision outcome.

[0046] Step S503: Compare and analyze the hypothetical decision results under each assumption, calculate the decision consistency ratio, decision change magnitude and sensitivity to evidence disturbance, and classify the decision stability into stable, substable or unstable levels accordingly.

[0047] Step S504: For different stability levels, implement corresponding decision processing strategies: For sub-stable decisions, increase decision uncertainty and add risk warnings; for unstable decisions, mark them as requiring manual review or postpone execution; for stable decisions, confirm the final personnel transfer decision results and generate decision explanation information including a summary of the deduction process, the number of assumptions, and the stability assessment conclusions, to provide a basis for decision issuance, execution, and auditing.

[0048] like Figure 6 As shown, step S60 specifically includes the following steps: Step S601: The final personnel transfer decision result is converted into a final personnel transfer decision instruction. Based on the preset instruction parsing rules and the decision interpretation information, the final personnel transfer decision instruction is parsed into a standard execution instruction format and sent to the corresponding execution terminal.

[0049] Step S602: Obtain the execution status information returned by the execution terminal, monitor whether the instruction was successfully issued, whether it was responded to, and the execution progress, and encapsulate the collected execution status data in a structured manner to form an execution feedback package. The execution feedback package includes at least information such as instruction identifier, execution status, response time, and anomaly marker, and is sent back to the evidence lifecycle management module to update the status, uncertainty parameters, or validity markers of relevant evidence objects, thereby realizing a closed-loop linkage between decision generation, execution feedback, and evidence update.

[0050] In this embodiment, when there is uncertainty or conflict of evidence in the candidate transfer decision, the stability and consistency of the decision result under multiple reasonable evidence states are systematically evaluated to avoid the risk of randomness in the decision result under a single evidence state.

[0051] First, candidate decisions and their corresponding fusion diagnostic information are received. The fusion diagnostic information includes at least fusion consistency, uncertainty level, and conflict markers. The candidate decisions are pre-judged to determine whether the uncertainty exceeds a preset threshold and whether there are conflicts or anomalies in the fusion results, thereby determining whether to proceed to the multi-hypothesis deduction path. When the uncertainty is low and the fusion results are consistent, the deduction process is skipped, and the decision confirmation stage is entered directly. When the uncertainty is high or conflicts exist, the number of hypotheses and deduction dimensions required for the deduction are determined, and multiple reasonable evidence state hypotheses are constructed under the premise of meeting business rules and constraints. Subsequently, in the multi-hypothesis evidence state generation unit, an evidence state set is formed based on uncertainty perturbation, evidence missing simulation, and low-confidence evidence replacement. The candidate personnel transfer decision is evaluated under each evidence state to obtain the corresponding hypothetical decision result. After completing the multi-hypothesis evaluation, the decision results under each hypothetical condition are compared and analyzed. The decision consistency ratio, decision change magnitude, and sensitivity to evidence perturbation are calculated, and the decision stability is classified into stable, substable, or unstable levels accordingly. For different stability levels, corresponding decision-making strategies are implemented: for stable decisions, the final personnel transfer decision instruction is confirmed and output; for sub-stable decisions, the decision uncertainty is increased and risk warnings are attached; for unstable decisions, they are marked as requiring manual review or postponed. Finally, the final personnel transfer decision result is output, along with decision explanation information including a summary of the deduction process, the number of assumptions, and the stability assessment conclusions, providing a basis for decision issuance, execution, and auditing.

[0052] The confirmed final personnel transfer decision is transformed into executable control commands, and the execution status of the decision is collected and fed back to construct a closed-loop operation mechanism for personnel transfer decision-making. First, the final personnel transfer decision and its corresponding explanatory information are received. According to preset command mapping rules, the decision result is parsed into a standardized execution command format and sent to the corresponding emergency execution terminal. The emergency execution terminal includes, but is not limited to, emergency command systems, information dissemination systems, or on-site handling systems. After the command is issued, the execution status information returned by the execution terminal is continuously acquired. The success of the command issuance, whether it has been responded to, and the execution progress are monitored. The collected execution status data is structured and encapsulated into an execution feedback package. The execution feedback package includes at least command identifier, execution status, response time, and anomaly markers and is transmitted back to update the status, uncertainty parameters, or validity markers of relevant evidence objects, thereby achieving closed-loop linkage between decision generation, execution feedback, and evidence updating. This allows for the engineering implementation and continuous adaptive optimization of personnel transfer decisions while maintaining the independence of the decision logic.

[0053] The following section further elaborates on the intelligent decision-making method for predictive personnel transfer based on multi-hypothesis evidence deduction.

[0054] The main implementation of this invention is as follows: First, an intelligent decision-making system for personnel relocation is deployed in an emergency management platform or urban disaster prevention information system. Disaster prevention areas are divided according to spatial units, and upstream-output hazard classification maps, vulnerability classification maps, and personnel relocation knowledge recommendation results are accessed. Simultaneously, corresponding prediction lead time information is obtained as a timeliness input. Based on this, the classification information and knowledge recommendation results are mapped to decision evidence related to personnel relocation. The evidence objects are standardized and encapsulated, and the uncertainty of the evidence is assigned and dynamically updated in conjunction with the prediction lead time, forming a valid evidence set with timeliness attributes and status markers. Subsequently, the valid evidence set is normalized, grouped, aggregated, and fused to generate candidate personnel relocation decision results. Consistency checks and conflict analysis are used to constrain and correct the fusion results, outputting candidate decisions and fusion diagnostic information. In cases where candidate decisions have high uncertainty or evidence conflicts, multiple reasonable evidence status assumptions are constructed under the premise of meeting business rules and constraints. The stability and consistency of candidate personnel relocation decisions under different evidence assumptions are evaluated. Based on the deduction results, decisions are classified into stable, substable, or unstable levels, and the final personnel relocation decision result is confirmed or the decision result requiring manual review is marked accordingly. Finally, the confirmed personnel transfer decision is sent to the emergency command or information release system for execution, and the execution status and feedback information are continuously collected. The feedback results are then sent back to update the evidence status and uncertainty parameters, thereby forming a closed-loop operation system for predictive intelligent decision-making on personnel transfer with evidence as the core and multiple hypothesis extrapolation as the guarantee.

[0055] The beneficial effects of the predictive intelligent decision-making method for personnel transfer based on multi-hypothesis evidence deduction in this invention are: Compared to existing personnel transfer decision-making technologies that generally rely on single risk indicators, static rules, or one-off judgments, this invention effectively overcomes the problems of unstable, unreliable, and unpredictable decision results in complex and uncertain environments by introducing an evidence-based multi-stage decision-making mechanism. This invention models the results of risk classification, vulnerability classification, and knowledge recommendation using evidence, and introduces an evidence lifecycle management mechanism to dynamically control the timeliness, uncertainty, and validity of decision evidence, avoiding interference from outdated information or low-reliability evidence. Simultaneously, this invention uses a multi-source evidence fusion and consistency constraint mechanism to uniformly aggregate and analyze decision evidence from different sources, significantly reducing the risk of misjudgment due to evidence conflicts while ensuring full utilization of information. Furthermore, this invention introduces a multi-hypothesis evidence deduction mechanism to evaluate the stability of candidate personnel transfer decisions under various reasonable evidence states, making the decision results independent of a single evidence state, thereby significantly improving the robustness and reliability of personnel transfer decisions under uncertain conditions. Furthermore, by constructing a closed-loop mechanism for decision execution and feedback, this invention achieves dynamic linkage between decision results and execution effectiveness, providing a basis for updating evidence status and optimizing subsequent decisions, thereby enhancing the system's engineering deployability and continuous operation capability. In summary, this invention can provide more robust, interpretable, and predictive personnel transfer decision support in complex and ever-changing emergency scenarios, demonstrating significant technological advancements and practical application value.

[0056] To achieve the above objectives, this invention also proposes a predictive intelligent decision-making system for personnel transfer based on multi-hypothesis evidence deduction, such as... Figures 7 to 10 As shown, the system includes an evidence construction and lifecycle management module, a multi-source evidence fusion and consistency constraint module, and a multi-hypothesis evidence deduction and decision output module.

[0057] The evidence construction and lifecycle management module is used to obtain evidence input information corresponding to each spatial unit, taking spatial units as the basic decision objects. The evidence input information includes hazard classification, vulnerability classification, and personnel transfer knowledge input information. Based on preset classification mapping rules, the classification information is transformed into initial transfer decision evidence related to personnel transfer.

[0058] The evidence construction and lifecycle management module is also used to combine the prediction lead information to assign corresponding uncertainty characteristics to the initial transfer decision evidence and uniformly incorporate it into the evidence lifecycle management process. The multi-source evidence fusion and consistency constraint module is used to dynamically update and aggregate evidence from valid states of different spatial units to form a multi-source evidence set for personnel transfer decisions.

[0059] The multi-source evidence fusion and consistency constraint module is also used to perform fusion calculation and consistency verification on the multi-source evidence set to generate corresponding predictive personnel transfer decision results.

[0060] The multi-hypothesis evidence deduction and decision output module is used to construct a variety of reasonable evidence state hypotheses based on the multi-hypothesis evidence deduction mechanism, while keeping the current environmental constraints unchanged, to evaluate the consistency and stability of the predictive personnel transfer decision results under different evidence hypothesis conditions, and to confirm the final personnel transfer decision results based on the evaluation results.

[0061] The multi-hypothesis evidence deduction and decision output module is also used to output corresponding personnel transfer decision instructions based on the final personnel transfer decision results. At the same time, it continuously monitors the decision execution effect and subsequent data changes, and re-incorporates relevant feedback information into the evidence lifecycle management process for subsequent evidence updates and decision optimization.

[0062] Furthermore, the evidence construction and lifecycle management module is also used to convert the evidence input information into standardized evidence objects, and configure evidence identifier, source information, timestamp, uncertainty parameter, weight parameter and status flag field for each evidence object to realize the object-oriented encapsulation of evidence.

[0063] Based on the pre-defined level mapping rules, the level information is mapped to the corresponding initial value of the transfer tendency.

[0064] The multi-source evidence fusion and consistency constraint module is also used to assign and correct the uncertainty of the evidence object based on the initial value of the transfer tendency and the prediction lead, so as to form an evidence state with time-sensitive attributes.

[0065] Perform core calculations and update operations on the evidence object, identify graded changes, invalid changes or feedback anomalies through a change detection mechanism, and trigger evidence updates, decay or reconstruction.

[0066] It outputs the current set of valid evidence and its metadata index information, and generates a complete evidence version chain and state transition record to support the analysis of decision interpretability and subsequent audit traceability, providing basic support for the formation of a closed-loop decision-making mechanism driven by the evidence lifecycle.

[0067] Furthermore, the multi-source evidence fusion and consistency constraint module is also used to obtain the effective evidence set, extract the transfer tendency, uncertainty, weight and version information of each evidence object, and perform consistent representation of evidence from different sources and at different scales through evidence normalization processing, while initially screening out missing fields, expired evidence or outlier evidence.

[0068] Evidence objects are aggregated and grouped according to spatial units, time windows, or decision-making scenarios to form corresponding multi-source evidence subsets, and the diversity and redundancy of the sources of evidence within each subset are identified.

[0069] The fusion result is generated based on the evidence support, uncertainty parameters and weight information, and the corresponding candidate transfer decision evidence is output. The fusion credibility index is calculated simultaneously.

[0070] Consistency verification and conflict analysis are performed on the fusion results. High-conflict or low-consistency scenarios are identified by calculating the degree of consistency or conflict between evidence, and fusion diagnostic information and corresponding predictive personnel transfer decision results are generated. The fusion diagnostic information includes at least fusion consistency, uncertainty level and conflict marker.

[0071] Specifically, when consistency is acceptable, the fusion result is retained and the result is corrected by increasing uncertainty or adjusting weights; when consistency is insufficient or conflict is unacceptable, abnormal evidence is downweighted or removed, or the fusion result is marked as uncertain.

[0072] Furthermore, the multi-hypothesis evidence deduction and decision output module is also used to predict the outcome of the predictive personnel transfer decision, and to determine whether the uncertainty exceeds a preset threshold and whether there is a conflict or anomaly in the fusion result, thereby determining whether it is necessary to enter the multi-hypothesis deduction path; when the uncertainty is low and the fusion result is consistent, the deduction process is skipped and the decision confirmation stage is directly entered; when the uncertainty is high or there is a conflict, the number of hypotheses and the deduction dimension required for the deduction are determined, and under the premise of meeting the business rules and constraints, a variety of reasonable evidence state hypotheses are constructed.

[0073] Based on uncertainty disturbances, evidence deficiency simulation, and low-credibility evidence replacement methods, an evidence state set is formed, and the outcome of the predictive personnel transfer decision is evaluated under each evidence state to obtain the corresponding hypothetical decision outcome.

[0074] The hypothetical decision-making results under various assumptions are compared and analyzed to calculate the decision consistency rate, decision change magnitude, and sensitivity to evidence disturbances. Based on this, the decision stability is classified into stable, substable, or unstable levels.

[0075] For different stability levels, corresponding decision-making strategies are implemented: for sub-stable decisions, decision uncertainty is increased and risk warnings are attached; for unstable decisions, they are marked as requiring manual review or postponed; for stable decisions, the final personnel transfer decision result is confirmed, and decision explanation information including a summary of the deduction process, the number of assumptions, and the stability assessment conclusions is generated to provide a basis for decision issuance, execution, and auditing.

[0076] Furthermore, the predictive personnel transfer intelligent decision-making system based on multi-hypothesis evidence deduction also includes a decision execution and feedback module. The decision execution and feedback module is used to convert the final personnel transfer decision result into a final personnel transfer decision instruction. Based on the preset instruction parsing rules and the decision interpretation information, the final personnel transfer decision instruction is parsed into a standard execution instruction format and sent to the corresponding execution terminal.

[0077] The system acquires the execution status information returned by the execution terminal, monitors whether the instruction was successfully issued, whether it was responded to, and the execution progress, and encapsulates the collected execution status data into an execution feedback package. The execution feedback package includes at least the instruction identifier, execution status, response time, and anomaly marker, and is sent back to the evidence lifecycle management module to update the status, uncertainty parameters, or validity markers of the relevant evidence objects, thereby achieving a closed-loop linkage between decision generation, execution feedback, and evidence updating.

[0078] The following combination Figures 7 to 10 The present invention provides a further explanation of the intelligent decision-making system for predictive personnel transfer based on multi-hypothesis evidence deduction.

[0079] like Figure 7 As shown, the present invention discloses a predictive intelligent decision-making system for personnel relocation based on multi-hypothesis evidence deduction. This system achieves predictive generation and dynamic optimization of personnel relocation decisions under complex and uncertain conditions by constructing a technical architecture that integrates hierarchical perception, evidence-driven mechanisms, multi-hypothesis deduction, and closed-loop feedback. The system includes an evidence construction and lifecycle management module, a multi-source evidence fusion and consistency constraint module, and a multi-hypothesis evidence deduction and decision output module. The evidence construction and lifecycle management module receives input information such as hazard classification, vulnerability classification, personnel relocation knowledge recommendation, and prediction lead time, and maps this input information to generate personnel relocation-related decision evidence. Simultaneously, it assigns uncertainty representation to the decision evidence and performs evidence objectification and lifecycle management. The multi-source evidence fusion and... The consistency constraint module is used to dynamically aggregate decision evidence in a valid state across spatial units, and to perform fusion calculation and consistency verification on the aggregated multi-source evidence set to generate candidate predictive personnel transfer decision results. The multi-hypothesis evidence inference and decision output module is used to construct a variety of reasonable evidence state hypotheses while keeping the current environmental constraints unchanged, to evaluate the stability and consistency of candidate personnel transfer decisions under different evidence hypothesis conditions, to confirm the final personnel transfer decision instruction and output it to the emergency execution terminal, and at the same time collect the execution feedback information of personnel transfer decision and send it back to the evidence construction and life cycle management module for subsequent update and optimization of decision evidence, thereby forming a predictive personnel transfer intelligent decision-making system with closed-loop collaborative operation of perception-evidence-inference-decision-feedback.

[0080] like Figure 8 As shown, the evidence lifecycle management module is used to uniformly model, dynamically update, and manage the status of various types of evidence generated during the personnel transfer decision-making process, ensuring that the evidence involved in the decision-making process always possesses timeliness, traceability, and credibility. This module first receives evidence input information from upstream sources, including risk classification, vulnerability classification, knowledge recommendation results, and execution feedback. It converts this evidence input into standardized evidence objects and configures each evidence object with fields such as evidence identifier, source information, timestamp, uncertainty parameters, weight parameters, and status markers, achieving object-oriented encapsulation of the evidence. Based on this, the module maps the classification information to the corresponding initial value of the transfer tendency according to pre-defined level mapping rules, and assigns and corrects the uncertainty of the evidence objects based on the prediction lead time, forming an evidence status with timeliness attributes. Subsequently, the module performs core calculations and update operations on the evidence objects, identifying level changes, timeliness changes, or feedback anomalies through a change detection mechanism, and triggering evidence updates, attenuation, or reconstruction. The evidence lifecycle management module incorporates an evidence state machine to automatically control the transition of evidence objects between states such as "generated, valid, decayed, withdrawn, or invalid." Evidence objects in the valid state are allowed to participate in subsequent evidence fusion and decision deduction, while evidence objects in the decayed or withdrawn state will have their weight reduced or be removed from the decision-making process. After updating the evidence state, the module outputs the current set of valid evidence and its metadata index information, while generating a complete evidence version chain and state transition record. This supports decision interpretability analysis and subsequent audit traceability, and provides a foundation for the system to form a closed-loop decision-making mechanism driven by the evidence lifecycle.

[0081] like Figure 9As shown, the multi-source evidence fusion and consistency constraint module is used to uniformly process the valid evidence objects output by the evidence lifecycle management module to generate candidate transfer decision results with reliability constraints. This module first receives a set of valid evidence objects, extracts the transfer tendency, uncertainty, weight, and version information from each evidence object, and performs evidence normalization to uniformly represent evidence from different sources and scales, while initially filtering out missing fields, expired evidence, or outlier evidence. After normalization, the module aggregates and groups the evidence objects according to spatial units, time windows, or decision scenarios to form corresponding multi-source evidence subsets, and identifies the source diversity and redundancy of evidence within each subset. Subsequently, the module generates fusion results in the fusion calculation unit based on evidence support, uncertainty parameters, and weight information, outputting the corresponding candidate transfer decision evidence, and simultaneously calculating the fusion credibility index. To avoid the instability of decision results caused by conflicts between multi-source evidence, the module further performs consistency verification and conflict analysis on the fusion results, identifying high-conflict or low-consistency scenarios by calculating the degree of consistency or conflict between evidence, and generating fusion diagnostic information. When consistency is acceptable, the module retains the fusion results and corrects them by increasing uncertainty or adjusting weights. When consistency is insufficient or conflicts are unacceptable, the module reduces or eliminates anomalous evidence, or marks the fusion results as uncertain. Finally, the multi-source evidence fusion and consistency constraint modules output candidate transfer decision results and corresponding fusion diagnostic information, providing structured input for subsequent multi-hypothesis evidence deduction and decision stability assessment.

[0082] like Figure 10As shown, the multi-hypothesis evidence deduction module is used to systematically evaluate the stability and consistency of decision results under multiple reasonable evidence states when there is uncertainty or evidence conflict in the candidate transfer decision, in order to avoid the risk of randomness in the decision results under a single evidence state. This module first receives candidate decisions and their corresponding fusion diagnostic information output from the multi-source evidence fusion and consistency constraint module, where the fusion diagnostic information includes at least fusion consistency, uncertainty level, and conflict markers. The module uses a deducibility judgment unit to pre-judge the candidate decisions, determining whether the uncertainty exceeds a preset threshold and whether there are conflicts or anomalies in the fusion results, thereby determining whether to enter the multi-hypothesis deduction path. When the uncertainty is low and the fusion results are consistent, the module skips the deduction process and directly enters the decision confirmation stage. When the uncertainty is high or there are conflicts, the module activates the hypothesis generation control unit to determine the number of hypotheses and deduction dimensions required for the deduction, and constructs multiple reasonable evidence state hypotheses under the premise of meeting business rules and constraints. Subsequently, the multi-hypothesis evidence deduction module, within the multi-hypothesis evidence state generation unit, forms a set of evidence states based on methods such as uncertainty disturbances, evidence missing simulations, and low-confidence evidence replacement. Under each evidence state, it evaluates the candidate personnel transfer decision, obtaining the corresponding hypothetical decision result. After completing the multi-hypothesis evaluation, the module compares and analyzes the decision results under each hypothesis condition through the stability and sensitivity analysis unit, calculating the decision consistency ratio, decision change magnitude, and sensitivity to evidence disturbances. Based on this, the decision stability is classified into stable, sub-stable, or unstable levels. For different stability levels, the module executes corresponding decision processing strategies: for stable decisions, it confirms and outputs the final personnel transfer decision instruction; for sub-stable decisions, it increases decision uncertainty and includes risk warnings; for unstable decisions, it marks them as requiring manual review or postponing execution. Finally, the multi-hypothesis evidence deduction module outputs the final personnel transfer decision instruction and generates decision explanation information including a summary of the deduction process, the number of hypotheses, and the stability assessment conclusions, providing a basis for decision issuance, execution, and auditing.

[0083] The decision execution and feedback module transforms the final personnel transfer decision confirmed by the multi-hypothesis evidence deduction module into executable control commands, and collects and provides feedback on the decision execution status to construct a closed-loop operation mechanism for personnel transfer decisions. This module first receives the final personnel transfer decision command and its corresponding explanatory information output by the multi-hypothesis evidence deduction module. Based on preset command mapping rules, it parses the decision command into a standardized execution command format and sends it to the corresponding emergency execution terminal via the system interface. The emergency execution terminal includes, but is not limited to, emergency command systems, information dissemination systems, or on-site handling systems. After the command is issued, the decision execution and feedback module continuously acquires the execution status information returned by the execution terminal, monitoring whether the command was successfully issued, whether it was responded to, and the execution progress. It then structures and encapsulates the collected execution status data into an execution feedback package. The execution feedback package includes at least command identifier, execution status, response time, and anomaly markers, and is sent back to the evidence lifecycle management module to update the status, uncertainty parameters, or validity markers of relevant evidence objects, thereby achieving closed-loop linkage between decision generation, execution feedback, and evidence updating. By introducing this decision execution and feedback module, the system can achieve the engineering implementation and continuous adaptive optimization of personnel transfer decisions while maintaining the independence of decision logic.

[0084] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A predictive intelligent decision-making method for personnel transfer based on multi-hypothesis evidence deduction, characterized in that, The method includes the following steps: Step S10: Using spatial units as the basic decision-making objects, obtain the evidence input information corresponding to each spatial unit. The evidence input information includes hazard classification, vulnerability classification, and personnel transfer knowledge input information. Based on the preset classification mapping rules, the classification information is transformed into initial transfer decision evidence related to personnel transfer. Step S20: Combining the prediction lead information, assign corresponding uncertainty characteristics to the initial transfer decision evidence and uniformly incorporate it into the evidence lifecycle management process; Step S30: Dynamically update and aggregate the evidence of valid states from different spatial units to form a multi-source evidence set for personnel transfer decisions; Step S40: Perform fusion calculation and consistency verification on the multi-source evidence set to generate corresponding predictive personnel transfer decision results; Step S50: Based on the multi-hypothesis evidence deduction mechanism, while keeping the current environmental constraints unchanged, construct a variety of reasonable evidence state hypotheses, evaluate the consistency and stability of the predictive personnel transfer decision under different evidence hypothesis conditions, and confirm the final personnel transfer decision based on the evaluation results. Step S60: Output the corresponding personnel transfer decision instruction based on the final personnel transfer decision result. At the same time, continuously monitor the decision execution effect and subsequent data changes, and re-incorporate the relevant feedback information into the evidence lifecycle management process for subsequent evidence updates and decision optimization.

2. The predictive intelligent decision-making method for personnel transfer based on multi-hypothesis evidence deduction as described in claim 1, characterized in that, Step S10 includes: Step S101: Convert the evidence input information into standardized evidence objects, and configure each evidence object with evidence identifier, source information, timestamp, uncertainty parameter, weight parameter and status flag field to realize the object-oriented encapsulation of evidence; Step S102: Based on the pre-defined level mapping rules, map the level information to the corresponding initial value of the transfer tendency. Step S20 includes: Step S201: Based on the initial value of the transfer tendency, the uncertainty of the evidence object is assigned and corrected in combination with the prediction lead time to form an evidence state with time-sensitive attributes; Step S202: Perform core calculation and update operations on the evidence object, identify graded changes, invalidation changes or feedback anomalies through the change detection mechanism, and trigger evidence update, decay or reconstruction. Step S203: Output the current set of valid evidence and its metadata index information, and generate a complete evidence version chain and state transition record to support the analysis of decision interpretability and subsequent audit traceability, providing basic support for the formation of a closed-loop decision-making mechanism driven by the evidence lifecycle.

3. The predictive intelligent decision-making method for personnel transfer based on multi-hypothesis evidence deduction according to claim 2, characterized in that, Step S30 includes: Step S301: Obtain the set of valid evidence, extract the transfer tendency, uncertainty, weight and version information of each evidence object, and perform consistent representation of evidence from different sources and scales through evidence normalization processing, while initially screening out missing fields, expired evidence or outlier evidence. Step S302: Aggregate and group the evidence objects according to spatial units, time windows or decision-making scenarios to form corresponding multi-source evidence subsets, and identify the diversity and redundancy of the sources of evidence in each subset. Step S303: Generate fusion results based on evidence support, uncertainty parameters and weight information, output the corresponding candidate transfer decision evidence, and simultaneously calculate the fusion credibility index. Step S40 includes: Step S401: Perform consistency verification and conflict analysis on the fusion results. Identify high-conflict or low-consistency scenarios by calculating the consistency or conflict degree between evidence, and generate fusion diagnostic information and corresponding predictive personnel transfer decision results. The fusion diagnostic information includes at least fusion consistency, uncertainty level and conflict marker. Specifically, when consistency is acceptable, the fusion result is retained and the result is corrected by increasing uncertainty or adjusting weights; when consistency is insufficient or conflict is unacceptable, abnormal evidence is downweighted or removed, or the fusion result is marked as uncertain.

4. The predictive intelligent decision-making method for personnel transfer based on multi-hypothesis evidence deduction according to claim 3, characterized in that, Step S50 includes: Step S501: Predict the outcome of the anticipated personnel transfer decision, and determine whether the uncertainty exceeds a preset threshold and whether there is a conflict or anomaly in the fusion result, so as to determine whether it is necessary to enter the multi-hypothesis deduction path; when the uncertainty is low and the fusion result is consistent, skip the deduction process and directly enter the decision confirmation stage; when the uncertainty is high or there is a conflict, determine the number of hypotheses and deduction dimensions required for the deduction, and construct a variety of reasonable evidence state hypotheses under the premise of meeting business rules and constraints. Step S502: Based on uncertainty disturbances, evidence missing simulations, and low-credibility evidence replacement methods, an evidence state set is formed, and the outcome of the predictive personnel transfer decision is evaluated under each evidence state to obtain the corresponding hypothetical decision outcome. Step S503: Compare and analyze the hypothetical decision results under each hypothetical condition, calculate the decision consistency ratio, decision change magnitude and sensitivity to evidence disturbance, and classify the decision stability into stable, substable or unstable levels accordingly. Step S504: For different stability levels, implement corresponding decision processing strategies: For sub-stable decisions, increase decision uncertainty and add risk warnings; for unstable decisions, mark them as requiring manual review or postpone execution; for stable decisions, confirm the final personnel transfer decision results and generate decision explanation information including a summary of the deduction process, the number of assumptions, and the stability assessment conclusions, to provide a basis for decision issuance, execution, and auditing.

5. The predictive intelligent decision-making method for personnel transfer based on multi-hypothesis evidence deduction according to claim 4, characterized in that, Step S60 includes: Step S601: The final personnel transfer decision result is converted into a final personnel transfer decision instruction. Based on the preset instruction parsing rules and the decision interpretation information, the final personnel transfer decision instruction is parsed into a standard execution instruction format and sent to the corresponding execution terminal. Step S602: Obtain the execution status information returned by the execution terminal, monitor whether the instruction was successfully issued, whether it was responded to, and the execution progress, and encapsulate the collected execution status data in a structured manner to form an execution feedback package. The execution feedback package includes at least instruction identifier, execution status, response time, and anomaly marker information, and is sent back to the evidence lifecycle management module to update the status, uncertainty parameters, or validity markers of relevant evidence objects, thereby realizing a closed-loop linkage between decision generation, execution feedback, and evidence update.

6. A predictive intelligent decision-making system for personnel transfer based on multi-hypothesis evidence deduction, characterized in that, The system includes an evidence construction and lifecycle management module, a multi-source evidence fusion and consistency constraint module, and a multi-hypothesis evidence deduction and decision output module. The evidence construction and lifecycle management module is used to obtain evidence input information corresponding to each spatial unit, taking spatial units as the basic decision objects. The evidence input information includes hazard classification, vulnerability classification, and personnel transfer knowledge input information. Based on the preset classification mapping rules, the classification information is transformed into initial transfer decision evidence related to personnel transfer. The evidence construction and lifecycle management module is also used to combine the prediction lead information to assign corresponding uncertainty characteristics to the initial transfer decision evidence and uniformly incorporate it into the evidence lifecycle management process. The multi-source evidence fusion and consistency constraint module is used to dynamically update and aggregate evidence from valid states of different spatial units to form a multi-source evidence set for personnel transfer decisions. The multi-source evidence fusion and consistency constraint module is also used to perform fusion calculation and consistency verification on the multi-source evidence set to generate corresponding predictive personnel transfer decision results. The multi-hypothesis evidence deduction and decision output module is used to construct a variety of reasonable evidence state hypotheses based on the multi-hypothesis evidence deduction mechanism, while keeping the current environmental constraints unchanged, to evaluate the consistency and stability of the predictive personnel transfer decision results under different evidence hypothesis conditions, and to confirm the final personnel transfer decision results based on the evaluation results. The multi-hypothesis evidence deduction and decision output module is also used to output corresponding personnel transfer decision instructions based on the final personnel transfer decision results. At the same time, it continuously monitors the decision execution effect and subsequent data changes, and re-incorporates relevant feedback information into the evidence lifecycle management process for subsequent evidence updates and decision optimization.

7. The intelligent decision-making system for predictive personnel transfer based on multi-hypothesis evidence deduction according to claim 6, characterized in that, The evidence construction and lifecycle management module is also used to convert the evidence input information into standardized evidence objects, and configure evidence identifier, source information, timestamp, uncertainty parameter, weight parameter and status flag field for each evidence object to realize the object-oriented encapsulation of evidence; Based on the pre-defined level mapping rules, the level information is mapped to the corresponding initial value of the transfer tendency. The multi-source evidence fusion and consistency constraint module is also used to assign and correct the uncertainty of the evidence object based on the initial value of the transfer tendency and the prediction lead, so as to form an evidence state with time-sensitive attributes. Perform core calculations and update operations on the evidence object, identify graded changes, invalid changes or feedback anomalies through the change detection mechanism, and trigger evidence updates, decay or reconstruction; It outputs the current set of valid evidence and its metadata index information, and generates a complete evidence version chain and state transition record to support the analysis of decision interpretability and subsequent audit traceability, providing basic support for the formation of a closed-loop decision-making mechanism driven by the evidence lifecycle.

8. The intelligent decision-making system for predictive personnel transfer based on multi-hypothesis evidence deduction according to claim 7, characterized in that, The multi-source evidence fusion and consistency constraint module is also used to obtain the effective evidence set, extract the transfer tendency, uncertainty, weight and version information of each evidence object, and perform consistent representation of evidence from different sources and scales through evidence normalization processing, while initially screening out missing fields, expired evidence or outlier evidence. Evidence objects are aggregated and grouped according to spatial units, time windows, or decision-making scenarios to form corresponding multi-source evidence subsets, and the diversity and redundancy of the sources of evidence within each subset are identified. The fusion result is generated based on the evidence support, uncertainty parameters and weight information, and the corresponding candidate transfer decision evidence is output. The fusion credibility index is calculated simultaneously. Consistency verification and conflict analysis are performed on the fusion results. High-conflict or low-consistency scenarios are identified by calculating the consistency or conflict degree between evidence, and fusion diagnostic information and corresponding predictive personnel transfer decision results are generated. The fusion diagnostic information includes at least fusion consistency, uncertainty level and conflict marker. Specifically, when consistency is acceptable, the fusion result is retained and the result is corrected by increasing uncertainty or adjusting weights; when consistency is insufficient or conflict is unacceptable, abnormal evidence is downweighted or removed, or the fusion result is marked as uncertain.

9. The intelligent decision-making system for predictive personnel transfer based on multi-hypothesis evidence deduction according to claim 8, characterized in that, The multi-hypothesis evidence deduction and decision output module is also used to predict the results of the predictive personnel transfer decision, and to determine whether the uncertainty exceeds the preset threshold and whether there is a conflict or anomaly in the fusion result, so as to determine whether it is necessary to enter the multi-hypothesis deduction path. When the uncertainty is low and the fusion results are consistent, skip the deduction process and go directly to the decision confirmation stage; when the uncertainty is high or there are conflicts, determine the number of assumptions and deduction dimensions required for the deduction, and construct a variety of reasonable evidence state assumptions under the premise of meeting business rules and constraints. Based on uncertainty disturbances, evidence deficiency simulation, and low-reliability evidence replacement methods, an evidence state set is formed, and the results of the predictive personnel transfer decision are evaluated under each evidence state to obtain the corresponding hypothetical decision results. The hypothetical decision-making results under various assumptions are compared and analyzed to calculate the decision consistency rate, decision change magnitude, and sensitivity to evidence disturbances. Based on this, the decision stability is classified into stable, substable, or unstable levels. For different stability levels, corresponding decision-making strategies are implemented: for sub-stable decisions, decision uncertainty is increased and risk warnings are attached; for unstable decisions, they are marked as requiring manual review or postponed; for stable decisions, the final personnel transfer decision result is confirmed, and decision explanation information including a summary of the deduction process, the number of assumptions, and the stability assessment conclusions is generated to provide a basis for decision issuance, execution, and auditing.

10. The intelligent decision-making system for predictive personnel transfer based on multi-hypothesis evidence deduction according to claim 9, characterized in that, It also includes a decision execution and feedback module, which is used to convert the final personnel transfer decision result into a final personnel transfer decision instruction. Based on the preset instruction parsing rules and the decision interpretation information, the final personnel transfer decision instruction is parsed into a standard execution instruction format and sent to the corresponding execution terminal. The system acquires the execution status information returned by the execution terminal, monitors whether the instruction was successfully issued, whether it was responded to, and the execution progress, and encapsulates the collected execution status data into an execution feedback package. The execution feedback package includes at least instruction identifier, execution status, response time, and anomaly marker information, and is sent back to the evidence lifecycle management module to update the status, uncertainty parameters, or validity markers of relevant evidence objects, thereby realizing a closed-loop linkage between decision generation, execution feedback, and evidence update.

Images