Safety leadership intelligent evaluation method and system based on knowledge graph

Abstract

The invention relates to the technical field of security management, and discloses a security leadership intelligent evaluation method and system based on a knowledge graph, and the method comprises the steps: constructing a security decision cognitive load knowledge graph, and building a structured cognitive load mapping network; multi-modal cognitive load monitoring is achieved, physiological and behavior data of a tested person are collected, and multi-modal data fusion analysis is carried out; performing an adaptive cognitive load test, generating an adversarial limit scene, dynamically adjusting task complexity, and determining a cognitive resource critical point; analyzing attention resource allocation efficiency, and identifying a resource allocation strategy mode; quantifying a cognitive efficiency index, and evaluating decision-making efficiency and multi-target balance capability in a high-pressure environment; safety leader evaluation is extended from surface behavior observation to deep cognitive process analysis, and the cognitive limit of a leader in an extreme safety situation is accurately identified.

Description

Technical Field

[0001] The present invention relates to the field of security management technology, and more specifically, to a security leadership intelligent assessment method and system based on knowledge graph. Background Art

[0002] Safety leadership is a key factor in organizational safety management and is of great significance for preventing major safety accidents. Traditional safety leadership assessment methods are mainly based on questionnaires, empirical observations or simple scenario simulations, and have three obvious shortcomings: First, the assessment content is superficial, mainly evaluating behavioral performance rather than deep cognitive ability, and it is impossible to accurately measure the changes in leaders' cognitive load in extreme safety situations; second, the assessment method is static, using fixed scenarios and difficulty settings, and it is impossible to locate the critical point of cognitive resource exhaustion; third, the assessment scope is limited, making it difficult to effectively simulate rare but high-risk safety incidents.

[0003] In the existing technology, some methods try to use static cognitive load assessment scales, but they cannot capture the dynamic changes of cognitive load in complex safety decision-making processes. Other methods propose to establish scenario simulation tests, but because the scenarios are fixed, they cannot adapt to the differences in cognitive abilities of different leaders. There are also methods that propose to use single physiological indicators such as eye tracking to assess cognitive status, but single-dimensional data is difficult to fully reflect complex cognitive processes.

[0004] Currently, there is no method that can organically integrate knowledge graphs, multimodal cognitive monitoring, adaptive testing, and cognitive performance analysis technologies to achieve in-depth, dynamic, and comprehensive evaluation of safety leadership. Therefore, there is an urgent need for an intelligent safety leadership evaluation method that can deeply measure the limits of leaders' cognitive processing capabilities. Summary of the invention

[0005] The present invention provides a knowledge graph-based intelligent safety leadership assessment method and system to solve the technical problems of superficial assessment content, static assessment method and limited assessment scope existing in the safety leadership assessment method in the related art.

[0006] The present invention provides a security leadership intelligent assessment method based on knowledge graph, comprising: Construct a knowledge graph of cognitive load for security decision-making, collect security incident data, analyze cognitive resource requirements for various security decision-making tasks, and establish a structured cognitive load mapping network; Realize multimodal cognitive load monitoring. Based on the knowledge graph of safety decision-making cognitive load, collect the subject's physiological and behavioral data through eye tracking, voice analysis and expression recognition, and conduct multimodal data fusion analysis; Perform adaptive cognitive load testing, use multimodal cognitive load monitoring results to generate adversarial extreme scenarios, dynamically adjust task complexity, and determine the critical point of cognitive resources; Analyze the effectiveness of attention resource allocation, measure the leader's ability to allocate resources between multiple security tasks based on the critical point data of step cognitive resources, calculate the attention conversion cost, and identify the resource allocation strategy pattern; Quantify the cognitive efficiency index, integrate the above contents, calculate the safety decision-making effect under unit cognitive resource input, and evaluate the decision-making efficiency and multi-objective balance ability under high-pressure environment.

[0007] Furthermore, the step of constructing a safety decision cognitive load knowledge graph includes: Collect historical event data in the security field to form an initial security event data set; Conduct cognitive load analysis on security event data to identify the type of cognitive resources and load level required for each type of security decision-making task; Construct knowledge graph entities and relationships to form a networked representation of the cognitive load of safety decisions; Set cognitive load threshold inference rules to determine critical thresholds for different cognitive resources in different security scenarios.

[0008] Furthermore, the step of implementing multimodal cognitive load monitoring includes: Configure an eye tracking data acquisition unit to capture the subject's gaze point distribution, gaze duration, pupil dilation degree, and blink frequency eye movement parameters; Construct a speech analysis and processing unit to extract the characteristic parameters of speech rate change, pause frequency, pitch fluctuation, and vocabulary complexity in the subject's speech; Implement the facial micro-reaction recognition unit to identify the micro-expression changes of the subject's face, including forehead wrinkles, eyebrow position, and mouth corner state characteristics; Perform multimodal data fusion analysis and calculate the comprehensive cognitive load index through multimodal fusion algorithm.

[0009] Furthermore, the expression for calculating the comprehensive cognitive load index by the multimodal fusion algorithm is: ; in represents the multimodal fusion function, , , Represents eye movement, voice and expression data respectively, is the weight coefficient of each mode, is the feature mapping function of each mode, is the eigenvector of the corresponding mode; It means to sum the three modes, from arrive , is the modal index.

[0010] Furthermore, the steps of performing the adaptive cognitive load test include: Generating an adversarial extreme scenario to create a challenging evaluation scenario, with the expression: ; where is the generated adversarial security decision scenario, is the basic security scenario, is the perturbation intensity, is the gradient of the scenario loss function, is the scenario evaluation function, is the target difficulty level; is the sign function; Constructing a task complexity dynamic adjustment algorithm to control the change of task complexity, with the expression: ; where is the task complexity at the next moment, is the task complexity at the current moment, is the target cognitive load level, is the currently measured cognitive load, is the adjustment rate parameter, is the time index; Implementing the determination of the cognitive resource critical point to determine the cognitive resource critical point of the tested person, with the expression: ; where represents the cognitive load critical point, represents the th cognitive load level, represents the performance at the cognitive load under, represents the performance threshold; represents the maximum value function, used to select the maximum cognitive load level that meets the conditions; represents the index of the cognitive load level; Introducing a multi-objective conflict test to test the leader's goal balance ability under high cognitive load.

[0011] Furthermore, the steps of analyzing the allocation efficiency of attention resources include: Constructing a multi-task parallel test environment, requiring the tested person to simultaneously pay attention to and process multiple security issues; Calculating the attention switching cost index to quantify the switching efficiency between different tasks; Analyze the resource allocation strategy model to identify efficient and inefficient resource allocation strategies; Determine the individual resource management characteristics, including dimensions such as task switching flexibility, resource allocation balance, and priority judgment accuracy.

[0012] Furthermore, the expression for calculating the attention switching cost metric is: ; where represents the attention switching cost from task to task , is the reaction time from task to task , is the baseline reaction time, represents the source task index, represents the target task index.

[0013] Furthermore, the steps for quantifying the cognitive efficiency metric include: Construct a decision performance evaluation model and calculate the comprehensive performance of the decision , with the expression: ; where represents the comprehensive performance of the decision , is the weight of the th performance metric, is the score of the decision on the th metric, represents the summation of all performance metrics, from to , represents the total number of performance metrics, represents the performance metric index; Calculate the cognitive efficiency ratio to quantify the decision-making effect generated by unit cognitive resource input, with the expression: ; where is the decision performance score, is the corresponding cognitive load level; represents the cognitive efficiency ratio, quantifying the decision-making effect generated by unit cognitive resource input, and generating a cognitive efficiency curve; Analyze the efficiency change under high-pressure conditions and calculate the efficiency change trend at different pressure levels; Evaluate the multi-objective balance ability and quantify the balance ability of the leader in a multi-objective conflict situation.

[0014] Further, the multi-objective balance ability expression is as follows: ; where and are the standard deviation and mean of the performance of each objective respectively, represents the balance degree of the objective set , represents the objective set, represents the performance of each objective.

[0015] A knowledge graph-based intelligent evaluation system for safety leadership is used to execute the above-mentioned knowledge graph-based intelligent evaluation method for safety leadership, and includes: A safety decision-making cognitive load knowledge graph construction module for establishing the mapping relationship between different safety tasks and the cognitive resource requirements; A multi-modal cognitive load monitoring module for real-time monitoring of the cognitive state through eye movement tracking, speech analysis, and facial expression recognition technologies; An adaptive cognitive load testing module for dynamically adjusting the task complexity and determining the cognitive resource critical point; An attention resource allocation efficiency analysis module for measuring the multi-task processing ability and resource allocation strategy; A cognitive efficiency index quantification module for evaluating the decision-making efficiency and multi-objective balance ability.

[0016] The beneficial effects of the present invention are as follows: extending the safety leadership evaluation from surface behavior observation to in-depth cognitive process analysis, accurately identifying the cognitive limit of leaders in extreme safety situations, improving the accuracy of determining the cognitive critical point, and effectively preventing major safety decision-making mistakes caused by cognitive overload; Realizing the effective simulation evaluation of rare safety events, creating a challenging test environment through an adversarial extreme scenario generation algorithm, and evaluating the ability of leaders to cope with unconventional safety challenges; Introducing a cognitive efficiency index to quantitatively measure the decision-making efficiency and multi-objective balance ability of leaders under different stress levels, providing a new evaluation dimension for the organization to select and cultivate efficient safety leaders. Brief Description of the Drawings

[0017] Figure 1 is a flowchart of a knowledge graph-based intelligent evaluation method for safety leadership in the present invention; Figure 2 is a flowchart of step 1 of the present invention; Figure 3 is a flowchart of step 2 of the present invention; Figure 4 is a flowchart of step 3 of the present invention; Figure 5 It is the flowchart of step 4 of the present invention; Figure 6 It is the flowchart of step 5 of the present invention. Detailed implementation manners

[0018] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that discussing these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein, and the functions and arrangements of the elements discussed can be changed without departing from the scope of protection of the content of this specification. Each example can omit, substitute, or add various processes or components as needed. Additionally, the features described in some examples can also be combined in other examples.

[0019] At least one embodiment of the present invention discloses a method for intelligent evaluation of safety leadership based on a knowledge graph, as Figures 1 to 6 shown, including: Step 1, construct a knowledge graph of safety decision-making cognitive load, collect safety event data, analyze the cognitive resource requirements of various safety decision-making tasks, and establish a structured cognitive load mapping network; In this step, by collecting and organizing historical major safety accidents and near-miss event cases, the mapping relationship between different safety tasks and cognitive resource requirements is established to form a structured knowledge network.

[0020] Step 1.1, collect historical event data in the safety field; Collect data on major safety accidents and near-miss events from different safety fields (such as industrial safety, public safety, network security, etc.), including information such as event descriptions, environmental conditions, personnel roles, decision-making processes, and result impacts, to form an initial safety event dataset.

[0021] Step 1.2, analyze the cognitive load of safety events; Perform a cognitive load analysis on the collected safety event data, identify the types of cognitive resources (such as working memory, attention resources, reasoning ability, etc.) and load levels required for each type of safety decision-making task, and use a cognitive load assessment model to calculate the cognitive load value of the task: ; where represents the cognitive load value of task , is the weight of the th type of cognitive resource, is the demand for this type of cognitive resource by task , is the total number of types of cognitive resources, is the index of the type of cognitive resource; represents the sum of all cognitive resource types, from arrive .

[0022] Step 1.3, construct knowledge graph entities and relationships; The analysis results are constructed into a knowledge graph, in which entities include security scenarios (Scenario), task types (TaskType), cognitive resources (CognitiveResource), load levels (LoadLevel), etc. The relationships between entities include "needs" (Requires), "influences" (Affects), "belongs to" (BelongsTo), etc., forming a networked representation of the cognitive load of security decisions.

[0023] Step 1.4, cognitive load threshold inference rule setting; The cognitive load threshold inference rule is set in the knowledge graph to determine the critical threshold for different cognitive resources in different security scenarios. The expression is: ; in Cognitive Resources The basic threshold value, is the situation adjustment factor, Representing cognitive resources In a safe situation The critical threshold below is used for subsequent cognitive limit determination.

[0024] The outputs of step 1 include: a safety decision cognitive load knowledge graph, which contains a structured representation of safety scenarios, task types, cognitive resource types and their relationships; a cognitive load assessment model, which is used to quantify the degree of demand for various cognitive resources for different tasks; and a cognitive load threshold rule set, which defines the critical values ​​of cognitive resources in different situations.

[0025] These data will form the basic framework and reference standard for cognitive load monitoring and evaluation in subsequent steps.

[0026] Step 2: Implement multimodal cognitive load monitoring. Based on the safety decision-making cognitive load knowledge graph, collect the subject's physiological and behavioral data through eye tracking, voice analysis and expression recognition, and perform multimodal data fusion analysis; This step is based on the safety decision-making cognitive load knowledge graph and its threshold rule set constructed in step 1. By integrating multiple physiological and behavioral data collection technologies, it monitors and evaluates the changes in the leader's cognitive state during the safety decision-making process in real time. The cognitive load model established in step 1 provides a theoretical basis to guide the interpretation of physiological data and the quantitative evaluation of cognitive load in this step.

[0027] Step 2.1, configure the eye movement tracking data acquisition unit; Configure a high-precision eye movement tracking device to capture in real time the eye movement parameters of the subject, such as the distribution of fixation points, fixation duration, pupil dilation degree, blink frequency, etc., analyze the allocation of the leader's visual attention resources and the cognitive processing intensity, and generate a time series data set of eye movement parameters.

[0028] Step 2.2, construct a speech analysis and processing unit; Construct a natural language processing model to extract characteristic parameters such as speech rate change, pause frequency, pitch fluctuation, and vocabulary complexity in the subject's speech. These parameters reflect the speech cognitive load of the leader during the safety decision-making process, and output a speech feature vector: ; where represents the speech feature vector, , , respectively represent the 1st, 2nd, and th speech feature parameters, represents the total number of speech feature parameters.

[0029] Step 2.3, implement the micro-expression recognition unit; Apply computer vision algorithms to recognize the micro-expression changes on the subject's face, including subtle expression features such as forehead wrinkles, eyebrow positions, and mouth corner states. These features can reflect the leader's emotional state and cognitive stress level, and generate an expression feature sequence.

[0030] Step 2.4, perform multi-modal data fusion analysis; Synchronize the eye movement data, speech features, and expression data in time and fuse the features, and calculate the comprehensive cognitive load index through a multi-modal fusion algorithm: ; where represents the multi-modal fusion function, , , respectively represent eye movement, speech, and expression data, is the weight coefficient of each modality, is the feature mapping function of each modality, is the feature vector of the corresponding modality; represents the summation of the 3 modalities, from to , is the modality index.

[0031] This multi-modal fusion algorithm solves the problem that a single physiological index is difficult to comprehensively reflect the cognitive state, and realizes a more accurate cognitive load assessment by weighted combination of feature mappings of different modalities. The algorithm first performs time alignment and feature extraction on the data of each modality, and then through a specific mapping function converts the features of different modalities into a unified cognitive load representation space, and finally assigns weights based on the reliability of each modality under the current task type for weighted fusion. This method can adaptively adjust the contribution degrees of different modalities. When a certain modality is affected by interference or noise, the system will automatically reduce its weight to ensure the robustness and accuracy of the evaluation results.

[0032] The outputs of Step 2 include: real-time cognitive load indicators, providing dynamic change data of the cognitive load of the tested person during the safety decision-making process; multi-modal physiological data sets, containing physiological parameters related to cognitive activities such as eye movement, speech, and expression; cognitive state evaluation reports, based on the knowledge graph threshold to compare the relationship between the current cognitive load level and the safety decision-making critical value.

[0033] These outputs will provide real-time feedback signals and adjustment bases for the next adaptive cognitive load test.

[0034] Step 3, perform an adaptive cognitive load test, utilize the multi-modal cognitive load monitoring results, generate adversarial extreme scenarios, dynamically adjust the task complexity, and determine the cognitive resource critical point; This step relies on the safety decision-making cognitive load knowledge graph constructed in Step 1 as the theoretical basis, utilizes the multi-modal cognitive load monitoring results in Step 2 as real-time feedback signals, dynamically adjusts the complexity and stress level of the safety decision-making task according to the real-time reactions of the tested person, determines the cognitive resource critical point, and creates a challenging evaluation scenario; the real-time cognitive load indicator provided by Step 2 is the core input parameter of the task complexity dynamic adjustment algorithm in this step.

[0035] Step 3.1, generate adversarial extreme scenarios; Apply the adversarial extreme scenario generation algorithm to create a challenging evaluation scenario, and the expression is: ; where is the generated adversarial safety decision-making scenario, is the basic safety scenario, is the perturbation intensity, is the gradient of the scenario loss function, is the scenario evaluation function, is the target difficulty level; is the sign function.

[0036] Inspired by the idea of adversarial example generation, this algorithm searches for tiny changes in the security scenario space that can maximize the cognitive load, making the scenario more challenging by adding directional perturbations. First, the algorithm calculates the gradient direction of the scenario evaluation function with respect to the current scenario parameters, and then adds controlled perturbations along this direction to generate new scenario variants.

[0037] In practical applications, security scenario parameters can include multiple dimensions such as time pressure, information uncertainty, and decision-making complexity. Through this algorithm, security decision-making scenarios under boundary conditions and extreme conditions can be automatically generated to effectively test the response capabilities of leaders in the face of unconventional security challenges.

[0038] Step 3.2, construct an algorithm for dynamically adjusting task complexity; Based on the real-time cognitive load monitoring results in Step 2, construct an algorithm for dynamically adjusting task complexity to control the change of task complexity, and the expression is: ; where is the task complexity at the next moment, is the task complexity at the current moment, is the target cognitive load level, is the currently measured cognitive load, is the adjustment rate parameter, is the time index.

[0039] This algorithm adopts the proportional regulation principle and dynamically adjusts the task complexity according to the gap between the current cognitive load and the target load. When the measured cognitive load is lower than the target level, the complexity is increased; otherwise, it is decreased, ensuring that the tested person is always in a state of cognitive challenge but not completely collapsing.

[0040] In the evaluation of security leadership, this algorithm can automatically adjust the task difficulty to the range most suitable for the ability level of the tested person, effectively avoiding the problems that the test is too simple to distinguish ability differences or too difficult to make the test data invalid.

[0041] Step 3.3, implement the determination of the cognitive resource critical point; Through the binary search or gradient ascent method, systematically adjust the task difficulty to find the cognitive load level that makes the performance index exactly reach the threshold , and determine the cognitive resource critical point of the tested person, that is: ; where represents the cognitive load critical point, represents the th cognitive load level, represents at the cognitive load The performance under represents the performance threshold; represents the maximum value function, which is used to select the maximum cognitive load level that meets the conditions; represents the index of the cognitive load level.

[0042] Step 3.4, introduce multi-objective conflict testing; Create a decision-making scenario with multiple objective conflicts such as safety and efficiency, short-term and long-term, quantify the degree of objective conflict, test the leader's ability to balance objectives under high cognitive load, and generate a conflict handling ability score. The expression is: ; where represents the objective and the conflict metric between, and are the priority weights of the objectives and respectively, represents the absolute value of the difference in the priority weights of the two objectives.

[0043] The outputs of Step 3 include: cognitive resource critical point data, which identifies the upper limit of the cognitive load that the tested person can bear in various safety decision-making tasks; an adversarial scenario library, which contains a series of safety decision-making scenarios from easy to difficult and their corresponding cognitive load requirements; task adaptability data, which records the adaptation curve of the tested person to tasks of different complexities; and a target conflict handling ability score, which measures the leader's multi-objective balance ability under limited cognitive resources. These outputs provide basic data and a test environment for subsequent analysis of the leader's attention resource allocation efficiency.

[0044] Step 4, analyze the attention resource allocation efficiency, based on the cognitive resource critical point data, measure the leader's resource allocation and coordination ability among multiple safety tasks, calculate the attention switching cost, and identify the resource allocation strategy pattern; This step is based on the safety decision-making cognitive load knowledge graph framework established in Step 1, combines the multi-modal cognitive load monitoring technology in Step 2 and the cognitive resource critical point data determined in Step 3, measures the leader's resource allocation and coordination ability among multiple safety tasks, and identifies the optimal strategies and individual differences in cognitive resource management. The adversarial scenario library and the target conflict handling ability score in Step 3 provide important references for constructing a multi-task parallel test environment in this step.

[0045] Step 4.1, construct a multi-task parallel test environment; Construct a test environment that includes multiple parallel security tasks, each with different priorities and resource requirements. Require the tested person to pay attention to and handle multiple security issues simultaneously, and record the task processing sequence and resource allocation behavior data.

[0046] Step 4.2, calculate the attention switching cost metric; Based on the task switching time and accuracy loss, calculate the attention switching cost metric: ; where represents the attention switching cost from task to task , is the reaction time from task to task , is the baseline reaction time, represents the source task index, represents the target task index.

[0047] Step 4.3, analyze the resource allocation strategy pattern; Use a clustering algorithm to analyze the resource allocation patterns of the tested person under different stress levels. The expression is: ; where represents the resource allocation pattern , represents the proportion of resources allocated to task , , represents the th task, represents the proportion of resources allocated to the th task, represents the task index, and its value range is from 1 to , represents the total number of tasks.

[0048] Step 4.4, determine the individual resource management characteristics; By comparing the differences between the resource allocation behaviors of the tested person in different test scenarios and the optimal strategy, determine their resource management characteristics, including dimensions such as task switching flexibility, resource allocation balance, and priority judgment accuracy, and form an individual resource management characteristics profile.

[0049] The outputs of Step 4 include: an attention switching cost matrix, which shows the cognitive costs of switching between different security tasks; a resource allocation strategy library, which records various typical resource allocation patterns and their effectiveness evaluations; and an individual resource management profile, which contains the resource allocation characteristics of the tested person and suggestions for the best working mode.

[0050] These outputs provide multi-dimensional data support for the final quantified leader cognitive efficiency index, evaluating safety leadership from the perspective of resource allocation.

[0051] Step 5: Quantify the cognitive efficiency index, integrate the above content, calculate the safety decision-making effect under the input of unit cognitive resources, and evaluate the decision-making efficiency and multi-objective balance ability under high-pressure environments; This step integrates all the results of the previous four steps: using the cognitive load knowledge graph in Step 1 as the theoretical framework, combining the multi-modal cognitive load data in Step 2 as the load evaluation basis, referring to the cognitive resource critical point data in Step 3, and conducting a comprehensive evaluation based on the analysis results of the attention resource allocation efficiency in Step 4. By calculating the safety decision-making effect under the input of unit cognitive resources, evaluate the decision-making efficiency and multi-objective balance ability of leaders in high-pressure environments.

[0052] Step 5.1: Construct a decision performance evaluation model; Construct a decision performance evaluation model and calculate the comprehensive performance of the decision The expression is: ; where represents the comprehensive performance of the decision , is the weight of the th performance indicator, is the score of the decision on the th indicator, represents the sum of all performance indicators, from to , represents the total number of performance indicators, represents the performance indicator index.

[0053] Step 5.2: Calculate the cognitive efficiency ratio; Based on the results of Step 2 and Step 5.1, calculate the cognitive efficiency ratio: ; where is the decision performance score, is the corresponding cognitive load level; represents the cognitive efficiency ratio, quantifying the decision-making effect generated by the input of unit cognitive resources and generating a cognitive efficiency curve.

[0054] Step 5.3: Analyze the efficiency change under high-pressure conditions; By comparing the cognitive efficiency ratios under different pressure levels, analyze the trend of efficiency change of leaders under high-pressure conditions and calculate the pressure level The efficiency change under is expressed as: ; where is the cognitive efficiency at this pressure level, is the baseline efficiency, represents the pressure level the efficiency change under, represents the pressure level.

[0055] Step 5.4, evaluate the multi-objective balance ability; In a multi-objective conflict situation, evaluate the leader's ability to balance each objective, and quantify the balance degree of the objective set The expression is: ; where and are the standard deviation and mean of the performance of each objective respectively, represents the balance degree of the objective set represents the objective set, represents the performance of each objective.

[0056] The final output of Step 5 includes: a cognitive efficiency curve showing the decision-making efficiency change under different cognitive load levels; a stress tolerance assessment report analyzing the leader's efficiency maintenance ability in a high-pressure environment; a multi-objective balance ability score quantifying the decision-making balance of the safety leader in a complex situation; and a comprehensive safety leadership profile integrating the evaluation results of all the above dimensions to form a complete safety leadership evaluation.

[0057] Through the organic connection and gradual progress of these five steps, this method constructs a complete safety leadership evaluation system. From the construction of the knowledge graph to the multi-modal data collection, then to the adaptive test, resource allocation analysis, and efficiency evaluation, clear data streams and logical chains are formed between the steps. The output of each step becomes an important input and basis for the subsequent steps, ensuring the systematicness of the evaluation process and the comprehensiveness of the evaluation results. The final evaluation results not only reflect the performance level of the safety leader but also deeply analyze its cognitive mechanism, providing a scientific basis for the selection and cultivation of safety management talents.

[0058] A knowledge graph-based intelligent safety leadership evaluation system for implementing the above-mentioned knowledge graph-based intelligent safety leadership evaluation method, including: A safety decision-making cognitive load knowledge graph construction module for establishing the mapping relationship between different safety tasks and the cognitive resource requirements; ​A multi-modal cognitive load monitoring module for real-time monitoring of the cognitive state through eye movement tracking, speech analysis, and facial expression recognition technologies; An adaptive cognitive load testing module for dynamically adjusting task complexity and determining the critical point of cognitive resources; An attention resource allocation efficiency analysis module for measuring multitasking ability and resource allocation strategies; A cognitive efficiency index quantification module for evaluating decision-making efficiency and multi-objective balance ability.

[0059] Here, the present invention provides an implementation example: This implementation was applied in the safety leadership evaluation of a large petrochemical enterprise. In recent years, several safety accidents have occurred in this enterprise, and some of these accidents are related to the decision-making mistakes of leaders in high-pressure environments. The enterprise's safety management department hopes to identify and cultivate safety leaders who can maintain efficient decision-making in extreme safety situations through scientific evaluation methods. The evaluation objects are 32 middle and senior managers in the enterprise, with an age distribution between 35 and 55 years old and 3 to 20 years of safety management experience.

[0060] Data on 150 safety incidents and near misses in the past 10 years of this enterprise were collected and analyzed, including 28 major accident cases and 122 minor or potential accident cases. Cognitive load analysis was carried out on these incidents to identify 5 main types of cognitive resources (working memory, attention allocation, situation awareness, decision-making reasoning, executive control) and the corresponding load level requirements.

[0061] The mapping relationship between some safety task types and cognitive resource requirements is shown in Table 1: Table 1: Mapping Table of Safety Task Types and Cognitive Resource Requirements

[0062] Subsequently, a safety decision-making cognitive load knowledge graph containing 1,250 entity nodes and 3,800 relationship edges was constructed. The entity types include safety scenarios (180), task types (42), cognitive resources (5 categories and 22 subcategories), load levels (4 levels), etc. The relationship types include 12 relationship types such as "require", "affect", "belong to", etc.

[0063] Eye movement tracking devices, voice collection, and facial expression recognition systems were configured for the 32 managers to real-time monitor their physiological and behavioral data in safety decision-making simulation tasks. The eye movement data collection frequency is 120Hz, voice features are extracted every 500ms, and facial expressions are collected at a speed of 30 frames per second. The comprehensive cognitive load index is calculated through a multi-modal fusion algorithm. The contribution weights of different modalities were calibrated in the pre-experiment, and the finally determined weight coefficients are: eye movement data , voice features , facial expression .

[0064] The results of multi-modal data fusion of a certain test subject under tasks of different difficulties are shown in Table 2: Table 2: Example of multi-modal cognitive load monitoring data (a certain test subject)

[0065] According to the knowledge graph and multi-modal monitoring results, the system generated a sequence of safety decision-making scenarios from easy to difficult, including types such as equipment anomalies, process deviations, personnel violations, environmental hazards, and multi-system cascading failures. Through the adversarial extreme scenario generation algorithm, challenging evaluation scenarios were created, and the parameter setting was the perturbation intensity , and each test subject received 10 - 18 rounds of tests of different complexities until reaching the cognitive resource critical point.

[0066] The adaptive test process of a typical test subject is shown in Table 3: Table 3: Example of adaptive cognitive load test process

[0067] After completing the above steps, the system comprehensively analyzed the attention resource allocation efficiency and cognitive efficiency of 32 management personnel. The cognitive resource critical points and key evaluation indicators of some test subjects are shown in Table 4: Table 4: Comparison of key indicators for safety leadership cognitive evaluation

[0068] Accurately identify cognitive limits; Through the adaptive cognitive load test technology, this method accurately determined the cognitive resource critical points of safety leaders. To verify its accuracy, the adaptive test results were compared with the traditional fixed-difficulty test method. By adding 9 professional cognitive psychologists as evaluators in the experiment, the ROC curve analysis evaluation method was used to evaluate the accuracy. The results are shown in Table 5: Table 5: Comparison results of cognitive critical point determination methods

[0069] Predict safety management performance; After completing the evaluation, the actual safety management performance of the test management personnel in the next 6 months was tracked and recorded. The results of this evaluation method were analyzed for correlation with the actual performance indicators, and the predictive power of different evaluation dimensions was compared. The results are shown in Table 6: Table 6: Correlation analysis between evaluation results and actual safety management performance

[0070] Note: * indicates p < 0.01, with strong correlation; The results show that the two indicators of the cognitive efficiency ratio and the cognitive resource critical point in this method have the highest correlation with the actual safety management performance, and the correlation coefficients reach 0.83 and 0.79 respectively, which are higher than those of the traditional evaluation method (0.39); this proves the advantage of this method in predicting the actual performance of safety leadership.

[0071] The embodiments of the present invention have been described above, but these embodiments are not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative rather than restrictive. Under the inspiration of this embodiment, those of ordinary skill in the art can also make more equivalent embodiments in various forms, all of which fall within the protection scope of this embodiment.

Claims

1. A security leadership intelligent assessment method based on knowledge graph, characterized in that: include: Construct a knowledge graph of cognitive load for security decision-making, collect security incident data, analyze cognitive resource requirements for various security decision-making tasks, and establish a structured cognitive load mapping network; Realize multimodal cognitive load monitoring. Based on the knowledge graph of safety decision-making cognitive load, collect the subject's physiological and behavioral data through eye tracking, voice analysis and expression recognition, and conduct multimodal data fusion analysis; Perform adaptive cognitive load testing, use multimodal cognitive load monitoring results to generate adversarial extreme scenarios, dynamically adjust task complexity, and determine the critical point of cognitive resources; Analyze the effectiveness of attention resource allocation, measure the leader's ability to allocate resources between multiple security tasks based on the critical point data of step cognitive resources, calculate the attention conversion cost, and identify the resource allocation strategy pattern; Quantify the cognitive efficiency index, integrate the above contents, calculate the safety decision-making effect under unit cognitive resource input, and evaluate the decision-making efficiency and multi-objective balance ability under high-pressure environment.

2. According to claim 1, a security leadership intelligent assessment method based on knowledge graph is characterized in that: The steps of constructing the safety decision cognitive load knowledge graph include: Collect historical event data in the security field to form an initial security event data set; Conduct cognitive load analysis on security event data to identify the type of cognitive resources and load level required for each type of security decision-making task; Construct knowledge graph entities and relationships to form a networked representation of the cognitive load of safety decisions; Set cognitive load threshold inference rules to determine critical thresholds for different cognitive resources in different security scenarios.

3. According to the knowledge graph-based security leadership intelligent assessment method of claim 1, it is characterized in that: The steps of implementing multimodal cognitive load monitoring include: Configure an eye tracking data acquisition unit to capture the subject's gaze point distribution, gaze duration, pupil dilation degree, and blink frequency eye movement parameters; Construct a speech analysis and processing unit to extract the characteristic parameters of speech rate change, pause frequency, pitch fluctuation, and vocabulary complexity in the subject's speech; Implement the facial micro-reaction recognition unit to identify the micro-expression changes of the subject's face, including forehead wrinkles, eyebrow position, and mouth corner state characteristics; Perform multimodal data fusion analysis and calculate the comprehensive cognitive load index through multimodal fusion algorithm.

4. According to claim 3, a security leadership intelligent assessment method based on knowledge graph is characterized in that: The expression for calculating the comprehensive cognitive load index by the multimodal fusion algorithm is: ； in represents the multimodal fusion function, , , Represents eye movement, voice and expression data respectively, is the weight coefficient of each mode, is the feature mapping function of each mode, is the eigenvector of the corresponding mode; It means to sum the three modes, from arrive , is the modal index.

5. According to claim 1, a security leadership intelligent assessment method based on knowledge graph is characterized in that: The steps of performing the adaptive cognitive load test include: Generate adversarial extreme scenarios to create challenging evaluation scenarios, expressed as: ； in For the generated adversarial security decision scenario, As a basic security scenario, is the disturbance intensity, is the gradient of the scene loss function, is the scene evaluation function, is the target difficulty level; is a symbolic function; Construct a dynamic adjustment algorithm for task complexity to control changes in task complexity. The expression is: ； in is the task complexity at the next moment, is the complexity of the task at the current moment, is the target cognitive load level, is the cognitive load currently measured, To adjust the rate parameter, is the time index; Implement the cognitive resource critical point measurement to determine the cognitive resource critical point of the subject, which is expressed as: ； in represents the critical point of cognitive load, Indicates cognitive load level, Cognitive load The performance under Indicates performance threshold; represents the maximum function, which is used to select the maximum cognitive load level that meets the conditions; An index representing the level of cognitive load; Introducing a multi-goal conflict test to test leaders’ ability to balance goals under high cognitive load.

6. According to the knowledge graph-based security leadership intelligent assessment method of claim 1, it is characterized in that: The step of analyzing the efficiency of attention resource allocation includes: Construct a multi-task parallel testing environment, requiring the testee to focus on and handle multiple security issues at the same time; Calculate the attention switching cost index to quantify the switching efficiency between different tasks; Analyze resource allocation strategy patterns and identify efficient and inefficient resource allocation strategies; Determine individual resource management characteristics, including the dimensions of task switching flexibility, resource allocation balance, and priority judgment accuracy.

7. The method for intelligent assessment of security leadership based on knowledge graph according to claim 6 is characterized in that: The expression for calculating the attention conversion cost index is: ； in Indicates that from the task Switch to Task The attention switching cost, For the task Switch to Task The reaction time, is the baseline reaction time, represents the source task index, Indicates the target task index.

8. According to claim 1, a security leadership intelligent assessment method based on knowledge graph is characterized in that: The step of quantifying the cognitive efficiency index comprises: Build a decision performance evaluation model to calculate the decision The comprehensive performance of is expressed as: ； in Indicates decision The comprehensive performance of For the The weight of the performance indicator, For decision making In the The scores on the indicators, Indicates the sum of all performance indicators, from arrive , Indicates the total number of performance indicators, Represents the performance indicator index; Calculate the cognitive efficiency ratio to quantify the decision-making effect of unit cognitive resource input. The expression is: ； in is the decision performance score, is the corresponding cognitive load level; It represents the cognitive efficiency ratio, quantifies the decision-making effect produced by unit cognitive resource input, and generates a cognitive efficiency curve; Analyze the efficiency changes under high pressure conditions and calculate the efficiency change trends under different pressure levels; Evaluate the ability to balance multiple goals and quantify the leader's ability to balance in multi-goal conflict situations.

9. The method for intelligent assessment of security leadership based on knowledge graph according to claim 8 is characterized in that: The multi-objective balance ability expression is: ； in and are the standard deviation and mean of each target performance, Represents the target set The balance of represents the target set, Indicates the performance of each target.

10. A security leadership intelligent assessment system based on knowledge graph, characterized in that: A method for intelligent assessment of security leadership based on knowledge graphs for executing any one of claims 1 to 9, comprising: Safety decision-making cognitive load knowledge graph construction module, which is used to establish the mapping relationship between different safety tasks and cognitive resource requirements; A multimodal cognitive load monitoring module for real-time monitoring of cognitive status using eye tracking, speech analysis, and expression recognition technologies; Adaptive cognitive load test module, used to dynamically adjust task complexity and determine the critical point of cognitive resources; Attention resource allocation efficiency analysis module, used to measure multi-tasking ability and resource allocation strategy; The cognitive efficiency index quantification module is used to evaluate decision-making efficiency and multi-objective balancing ability.

Images