An artificial intelligence-based enterprise human resource management system

Abstract

The present application relates to the technical field of human resource management, and discloses an enterprise human resource management system based on artificial intelligence; an image construction module is used to construct a periodically updated digital twin image; a scene simulation module is used to simulate an enterprise virtual scene; a feature simulation module is used to simulate internal potential features; an employee screening module is used to screen potential employees from employees; a potential prediction module is used to identify training indicators and develop training management information; the present application can dynamically generate a virtual adjustable test simulation environment, ensure that employees can conveniently perform simulation tests in a virtual environment, avoid the subjectivity and one-sidedness of employee potential assessment in traditional human resource management, reduce the workload of the human resource department in employee assessment, potential screening and training planning, improve the work efficiency of human resource management, and achieve the effects of intelligentization, precision and high efficiency of human resource management.

Description

Technical Field

[0001] This invention relates to the field of human resource management technology, and more specifically, to an enterprise human resource management system based on artificial intelligence. Background Technology

[0002] As companies expand and market competition intensifies, human resource management has become an important component of a company's core competitiveness. Human resource management systems based on Excel spreadsheets or simple databases are no longer sufficient to meet the company's needs for accurate identification, scientific evaluation, and efficient training of potential talent. Therefore, it is necessary to leverage artificial intelligence to accurately identify potential talent within the company, thereby achieving efficient human resource management.

[0003] The patent application with publication number CN115204849A discloses an artificial intelligence-based enterprise human resource management method and system, which includes the following steps: receiving self-evaluation information on job rotations input by all new employees; receiving resource allocation information; obtaining new employee job assignment information based on the job matching degree of each new employee's job rotations and the number of employees required for each job; and breaking down the total job tasks for each job based on the number of employees required for each job and the matching degree of the specific job tasks of the corresponding new employees to obtain job task assignment information. Existing human resource management systems typically build employee profiles based on historical static data when identifying and evaluating potential employees. This results in the inability to dynamically correlate employee profiles with real-time changes in employee performance, leading to a lag in the data upon which subsequent human resource decisions rely. Furthermore, the use of subjective human experience in the evaluation process fails to comprehensively and accurately simulate employees' real performance in complex and ever-changing corporate management scenarios. It is difficult to scientifically and comprehensively discover employees' deep-seated competencies in key situations, making the evaluation results highly subjective and limited, thereby reducing the effectiveness of human resource management.

[0004] In view of this, the present invention proposes an artificial intelligence-based enterprise human resource management system to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: an enterprise human resource management system based on artificial intelligence, comprising: The profile building module is used to dynamically match the multimodal data of employees with profile positions to build a digital twin profile with update nodes and regular updates. The scene simulation module is used to extract the portrait tags of the digital twin profile, formulate scene requirements based on the portrait tags as the core of the scene, and simulate a virtual enterprise scene that matches the scene requirements. The feature simulation module is used to import digital twin profiles into the enterprise virtual scene and dynamically simulate the enterprise virtual scene under the simulation mechanism to simulate the inherent potential features. The employee screening module is used to convert intrinsic potential characteristics into feature scores, calculate comprehensive potential scores based on feature scores, and screen potential employees from among the existing employees. The potential prediction module is used to combine multimodal data and intrinsic potential characteristics into comprehensive identification data, identify training indicators through a potential quantification identification model, and formulate training management information for potential employees based on the training indicators.

[0006] Furthermore, multimodal data includes hardware performance characteristics, software behavior characteristics, and test feedback characteristics; Hardware performance characteristics include performance evaluation values, pre-assessment rate, project participation level, project completion rate, and business skills curve; Software behavioral characteristics include effective communication rate, cooperation tendency rate, decision accuracy rate, conflict interaction rate, and stress resistance level; Test feedback characteristics include problem-solving approaches, strategic risk appetite, and learning agility.

[0007] Furthermore, the method for constructing a digital twin profile is as follows: Remove duplicate and erroneous data from the multimodal data and convert the multimodal data into a uniform format to obtain formatted data; A basic profile is simulated using digital twin technology. Three label layers, one inside and one outside, are established in the basic profile and are named the outer layer, the inner layer, and the potential layer in order from the outside to the inside. A circularly distributed image positions are set in the external layer, internal layer and potential layer respectively. With one feature corresponding to one image position as the standard, hardware performance features, software behavior features and test feedback features are imported into the image positions of the external layer, internal layer and potential layer one by one. Remove blank image positions and build a bidirectional transmission channel between two adjacent image positions, which is called the image channel; Update nodes with update cycles are configured on all portrait channels to facilitate the conversion of basic portraits into digital twin portraits.

[0008] Furthermore, the profile tags include cross-departmental resource conflicts, sudden public relations crises, disagreements on new product strategies, and low team morale; The method for extracting image tags is as follows: Based on an update cycle, the multimodal data in the external, internal, and potential layers of the digital twin profile are recorded as index data to obtain the first index data, the second index data, and the third index data. The first index data, the second index data, and the third index data are compared with the standard tags in the tag database for consistency, and the standard tags that are basically consistent or completely consistent with the comparison results are recorded as valid tags. All valid labels are compared repeatedly, and the valid labels that are repeated are recorded as portrait labels, resulting in B portrait labels.

[0009] Furthermore, the simulation method for enterprise virtual scenarios is as follows: The system retrieves employee identity and work information from the human resources database, and then binds and summarizes the identity and work information to generate basic information. Using the image tags as the core of the scene and the basic information as the scene outline, the core of the scene and the scene outline are combined into scene requirements to obtain B scene requirements. The B scenario requirements are imported one by one into the intelligent scenario engine. The scenario permissions and scenario logic of the intelligent scenario engine are set to simulate the enterprise virtual scenario corresponding to the B scenario requirements.

[0010] Furthermore, the simulation mechanism is as follows: a dynamic simulation ends when all the intrinsic potential features are simulated, and all dynamic simulations end when the overlap between any two adjacent intrinsic potential features is greater than the calibrated overlap threshold.

[0011] Furthermore, inherent potential characteristics include business decision-making characteristics, communication characteristics, resource allocation characteristics, administrative management characteristics, and risk prevention and control characteristics; The simulation method for intrinsic potential characteristics is as follows: A10: Arrange the B enterprise virtual scenarios into a scenario queue, and import the digital twin profile into the enterprise virtual scenario that is first in the scenario queue; A11: Initialize simulation parameters in the enterprise virtual scene, send dynamic simulation commands to the enterprise virtual scene, simulate the inherent potential characteristics, and perform integrity analysis on the inherent potential characteristics; A12: When the intrinsic potential characteristic is not in a complete state, repeat A11; when the intrinsic potential characteristic is in a complete state, execute A13. A13; Remove the simulated enterprise virtual scenarios from the scenario queue, import the digital twin profile into the remaining enterprise virtual scenario in the first position of the scenario queue, and repeat A11-A13 until the overlap of any two adjacent intrinsic potential features is greater than the calibrated overlap threshold, and then obtain the intrinsic potential features.

[0012] Furthermore, the feature scores include business decision-making scores, communication scores, resource allocation scores, administrative management scores, and risk prevention and control scores; The screening method for potential employees is as follows: Query the professional attributes of employees in the enterprise, and use the professional attributes as the index base to index the corresponding set of coefficients from the database. The weighted coefficients in the coefficient set are assigned to the business decision-making score, communication score, resource allocation score, administrative management score, and risk prevention score, respectively, and then the weighted sum is calculated to obtain the comprehensive potential score. Analyze the overall potential scores of employees and identify those whose overall potential scores exceed a preset potential score threshold as potential employees.

[0013] Furthermore, the training indicators include professional direction, management level, development cycle, and benefit value; Specializations include human resources organization, technology development, and financial management; Management levels include senior management, middle management, and junior management; The molding cycle includes one month, six months, and twelve months.

[0014] Furthermore, the cultivation management information is divided into three sections: the first section is the historical section, the second section is the testing section, and the third section is the cultivation section. Multimodal data is imported into the historical section, intrinsic potential characteristics are imported into the testing section, and cultivation indicators are imported into the cultivation section to generate cultivation management information.

[0015] The technical effects of this invention's artificial intelligence-based enterprise human resource management system are as follows: (1): This invention uses multimodal data to construct regularly updated digital twin profiles, which can dynamically link employees’ work performance in the enterprise in historical periods and current periods, ensuring the real-time and accuracy of relevant data information in the digital twin profiles, effectively avoiding the problems of static and unchanging employee profiles and data lag in traditional human resource management systems, and thus providing real and comprehensive data support for subsequent human resource management.

[0016] (2): This invention dynamically simulates the inherent potential characteristics in a virtual enterprise scenario and calculates the comprehensive potential score accordingly. This not only dynamically generates a virtual and adjustable test simulation environment, ensuring that employees can conduct convenient simulation tests in the virtual environment, but also truly reflects the performance of employees in actual management scenarios. This avoids the subjectivity and one-sidedness of employee potential assessment in traditional human resource management. At the same time, it reduces the workload of human resource departments in employee assessment, potential screening and training planning, improves the efficiency of human resource management, and achieves the effects of intelligent, precise and efficient human resource management. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of a module of an artificial intelligence-based enterprise human resource management system provided in Embodiment 1 of the present invention; Figure 2 This is a flowchart illustrating an artificial intelligence-based enterprise human resource management method provided in Embodiment 2 of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: Please refer to Figure 1 As shown in this embodiment, an artificial intelligence-based enterprise human resource management system includes: The profile building module collects multimodal data of employees, dynamically matches the multimodal data with profile positions, and builds a digital twin profile of the employee. Multimodal data is used to analyze the multidimensional data on employees' personal development and growth trajectory within an enterprise over a historical period. In order to represent employees from different types and perspectives, multimodal data needs to include data on employees' external hardware aspects, internal software aspects, and evaluation and testing aspects. Specifically, multimodal data includes hardware performance characteristics, software behavior characteristics, and test feedback characteristics.

[0020] Hardware performance characteristics are features used to represent an employee's business capabilities and growth over a historical period, serving as the basis for screening employees at the external hardware level. In this embodiment, hardware performance characteristics include, but are not limited to, performance evaluation values, pre-evaluation rate, project participation level, project completion rate, and business skill curve.

[0021] It should be noted that the performance appraisal value refers to the specific performance of an employee in each performance appraisal, the assessment lead rate refers to the percentage of employees who are in the lead area in the performance appraisal, the project participation level refers to the specific level of each enterprise project that an employee participates in, the project completion rate refers to the completion progress of each enterprise project that an employee participates in, and the business skills curve refers to the distribution curve formed by the results of each business skills appraisal of an employee.

[0022] Software behavioral characteristics are features used to represent employees’ behavior and communication style over a historical period, serving as the basis for screening employees at the internal software level. In this embodiment, software behavioral characteristics include, but are not limited to, effective communication rate, cooperation tendency rate, decision accuracy rate, conflict interaction rate, and stress resistance level.

[0023] It should be noted that the effective communication rate refers to the percentage of time that employees communicate in a correct and positive state when conducting work communication; the cooperation tendency rate refers to the probability that employees cooperate with other colleagues; the decision accuracy rate refers to the probability that the results of the decisions and measures that employees participate in and formulate are correct; the conflict interaction rate refers to the probability that conflicts may occur when employees work with other colleagues; and the stress resistance level refers to the level of employees' resistance to stress in a high-pressure work environment.

[0024] Test feedback features are characteristics used to represent employees' performance in tests and assessments over a historical period, serving as the basis for employee selection at the test assessment level. In this embodiment, the test feedback characteristics include, but are not limited to, problem-solving approaches, strategy risk preferences, and learning agility.

[0025] It should be noted that problem-solving approach refers to the approach employees take when dealing with test problems during internal testing and evaluation. Strategy risk appetite refers to the level of risk associated with the strategies and measures provided by employees. Learning agility refers to how quickly employees can correctly learn the assessment content during the testing and evaluation process.

[0026] In this embodiment, multimodal data typically comes from internal enterprise data such as OA, project management, CRM, training platforms, anonymous surveys, and compliance-compliant anonymized data.

[0027] Once multimodal data is obtained, a digital twin profile that matches the specific performance of employees in historical periods can be constructed based on the multimodal data. In this embodiment, the digital twin profile is a digital twin model constructed based on digital twin technology and combined with the multimodal data of employees, and serves as the basis for subsequent evaluation and analysis of whether employees have training value and management potential.

[0028] When constructing a digital twin profile, the digital twin profile usually contains multiple relatively independent and dynamically related profile bits. Each profile bit can correspond to a piece of data in a certain type, so as to achieve a one-to-one correspondence between multimodal data and digital twin profile. The method for constructing a digital twin profile is as follows: The multimodal data is cleaned to remove duplicate and erroneous data, and the cleaned multimodal data is converted into a unified format to obtain formatted data. A basic profile is simulated using digital twin technology. Three label layers, one inside and one outside, are established in the basic profile and are named the outer layer, the inner layer, and the potential layer in order from the outside to the inside. A circularly distributed image positions are set in the external layer, internal layer and potential layer respectively. With one feature corresponding to one image position as the standard, hardware performance features, software behavior features and test feedback features are imported into the image positions of the external layer, internal layer and potential layer one by one. Remove the image bits that do not have hardware performance characteristics, software behavior characteristics, and test feedback characteristics imported, and build a bidirectional transmission channel between two adjacent image bits, which is called the image channel. By building the image channel between image bits, the data in adjacent image bits can be transmitted and interacted bidirectionally, ensuring the dynamic management effect of data in different locations in the digital twin image and avoiding the situation where the data is too discrete to be integrated and interacted. Update nodes with update cycles are configured on all portrait channels, and the portrait positions are dynamically updated after an update cycle, so as to convert the basic portrait into a digital twin portrait.

[0029] It should be noted that the update cycle is the time frame used to update the multimodal data in the digital twin profile, to ensure that the digital twin profile can maintain a dynamic correlation with the actual performance of employees in the enterprise, and to avoid the data lag that exists in static digital twin profiles.

[0030] The scene simulation module extracts the portrait tags of the digital twin profile and uses the portrait tags as the core of the scene to simulate a virtual enterprise scene that matches the portrait tags. A digital twin profile is a holistic representation of an employee's actual performance within a company. Profile tags represent the true meaning and direct content of the digital twin profile and serve as a reflection of the employee's true performance within the company. Specifically, the number of image tags contained in each digital twin profile is not fixed; it may be one or more. Therefore, it is necessary to accurately and comprehensively identify and extract the image tags of the digital twin profile.

[0031] In this embodiment, the profile tags of the digital twin profile include, but are not limited to, cross-departmental resource conflicts, sudden public relations crises, disagreements on new product strategies, and low team morale.

[0032] The method for extracting image tags is as follows: Based on an update cycle, the multimodal data in the external, internal, and potential layers of the digital twin profile are recorded as index data to obtain the first index data, the second index data, and the third index data. The first index data, the second index data, and the third index data are compared with the standard labels in the label database for consistency. Standard labels that are basically consistent or completely consistent with the comparison results are recorded as valid labels. The comparison results are used to represent the specific results of the consistency comparison between the index data and the standard labels. Specifically, the comparison results include basically consistent, completely consistent, and inconsistent. All valid labels are compared repeatedly, and the valid labels that are repeated are recorded as portrait labels, resulting in B portrait labels.

[0033] After extracting the profile tags, it is necessary to use the profile tags as the core of the scene to simulate and construct a virtual enterprise scene that matches each profile tag, so that the virtual enterprise scene can serve as a virtual environment for simulating tests on employees. In this embodiment, the number of enterprise virtual scenes needs to be consistent with the number of profile tags, so that one profile tag corresponds to one enterprise virtual scene, thereby enabling independent virtual simulation of each profile tag.

[0034] The simulation method for enterprise virtual scenarios is as follows: The system retrieves employee identity and work information from the human resources database, and then binds and summarizes the identity and work information to generate basic information. Identity information refers to information about an employee's personal identity, including but not limited to name, gender, department, and date of employment. Work information refers to information about an employee's job, including but not limited to job level and professional skills. Using the image tags as the core of the scene and the basic information as the scene outline, the core of the scene and the scene outline are combined into scene requirements to obtain B scene requirements. Each of the B scenario requirements is imported into the intelligent scenario engine. Scenario permissions and logic are then set within the engine to simulate enterprise virtual scenarios corresponding to these requirements. Scenario permissions and logic serve as the basis for setting the permissions and logic for simulating the enterprise virtual scenarios, ensuring that the simulated scenarios maintain reasonable operational permissions and orderly control logic.

[0035] It should be noted that the intelligent scene engine is not a general simulation tool, but an AI system that embeds organizational behavior, leadership mechanics, and specific business logic. It can dynamically generate high-fidelity management challenge environments. It can learn the correspondence between scene requirements and enterprise virtual scenes by taking different employees, different profile tags, different scene requirements, and corresponding enterprise virtual scenes as learning targets. Thus, given the scene requirements, it can automatically simulate the required enterprise virtual scenes by combining the set scene permissions and scene logic.

[0036] The feature simulation module imports the digital twin profile into the enterprise's virtual scene and dynamically simulates the enterprise's virtual scene under the simulation mechanism to simulate its inherent potential features; Once the enterprise's virtual scene is obtained, the digital twin profiles of employees can be imported into the enterprise's virtual scene to dynamically simulate the digital twin profiles, thereby simulating their inherent potential characteristics.

[0037] When dynamically simulating digital twin profiles in a virtual enterprise environment, a simulation mechanism needs to be set up during the dynamic simulation to ensure the accuracy and reliability of the simulation process and the reliability of the results. Specifically, the simulation mechanism is as follows: a dynamic simulation ends when all the intrinsic potential features are simulated, and all dynamic simulations end when the overlap of the intrinsic potential features is greater than the calibration overlap threshold.

[0038] It should be noted that the calibration overlap threshold refers to the minimum value at which the number of intrinsic potential features obtained in two consecutive simulations remain in the same state. This provides a numerical basis for the consistency of the results of dynamic simulation of enterprise virtual scenarios, thereby obtaining relatively accurate employee simulation test results.

[0039] After dynamic simulation in a virtual enterprise scenario, corresponding intrinsic potential characteristics are obtained, which can be used as simulation test results automatically generated by employees when conducting test simulations, and can reflect the decisions and measures made by employees in different types of simulation tests. Since intrinsic potential characteristics represent employees' corresponding results in test simulations, intrinsic potential characteristics need to include multiple different dimensions; Specifically, the inherent potential characteristics include business decision-making characteristics, communication characteristics, resource allocation characteristics, administrative management characteristics, and risk prevention and control characteristics.

[0040] Business decision characteristics refer to the results of simulated tests conducted by employees in a virtual enterprise setting regarding their business skills and decision-making abilities.

[0041] Communication characteristics refer to the results of simulated tests conducted on employees in virtual corporate scenarios regarding the exchange of ideas and hierarchical communication.

[0042] Resource allocation characteristics refer to the simulation test results of employees in a virtual enterprise scenario regarding material allocation and resource control.

[0043] Administrative management characteristics refer to the results of simulated tests conducted by employees in a virtual enterprise setting regarding administrative personnel and enterprise management.

[0044] Risk control characteristics refer to the results of simulated tests conducted by employees in a virtual enterprise setting regarding location risks and security control.

[0045] The simulation method for intrinsic potential characteristics is as follows: A10: Arrange the B enterprise virtual scenarios into a scenario queue, and import the digital twin profile into the enterprise virtual scenario that is first in the scenario queue; A11: Initialize simulation parameters in the enterprise virtual scene, send dynamic simulation commands to the enterprise virtual scene, simulate the inherent potential characteristics, and perform integrity analysis on the inherent potential characteristics; A12: When the intrinsic potential characteristic is not in a complete state, repeat A11; when the intrinsic potential characteristic is in a complete state, execute A13. A13; Remove the simulated enterprise virtual scenarios from the scenario queue, import the digital twin profile into the remaining enterprise virtual scenario in the scenario queue that is ranked first, and repeat A11-A13 until all enterprise virtual scenarios are dynamically simulated and the overlap of any two adjacent intrinsic potential features is greater than the calibrated overlap threshold, and then obtain the intrinsic potential features.

[0046] It should be noted that simulation parameters are multi-dimensional parameters used to limit the dynamic simulation operation of the enterprise virtual scene, ensuring that the enterprise virtual scene can be dynamically simulated normally and stably. Specifically, simulation parameters include simulation start time, simulation end time, simulation permission range, and simulation logic order. They can limit the start and end time of the enterprise virtual scene during dynamic simulation, as well as limit the permission range and the simulation order of the inherent potential features during dynamic simulation.

[0047] The employee screening module converts intrinsic potential characteristics into feature scores, calculates the comprehensive potential score of employees based on these feature scores, and then screens out potential employees from among the existing workforce. After obtaining the intrinsic potential features, these features cannot be converted to numerical values ​​at this point. Therefore, feature scoring conversion is required for the intrinsic potential features to facilitate subsequent calculations. In this embodiment, the feature scores include business decision-making scores, communication scores, resource allocation scores, administrative management scores, and risk prevention and control scores.

[0048] Specifically, when converting business decision-making characteristics, communication characteristics, resource allocation characteristics, administrative management characteristics, and risk prevention and control characteristics, it is necessary to compare these characteristics with the standard characteristics corresponding to different score levels. The score corresponding to the score level that matches the standard characteristics is recorded as the score of that inherent potential characteristic, thereby obtaining the business decision-making score, communication score, resource allocation score, administrative management score, and risk prevention and control score.

[0049] It should be noted that when converting business decision-making scores, communication scores, resource allocation scores, administrative management scores, and risk control scores, the specific values ​​of all scores need to be standardized to avoid some scores being too large or too small. For example, the scores for business decision-making, communication, resource allocation, administration, and risk control are all set to 1, 2, 3, 4, and 5, respectively.

[0050] The comprehensive potential score is used to give employees specific scores across a range of dimensions and levels in a mock test. It serves as the numerical basis for judging whether an employee's performance in this mock test is satisfactory and is also a direct basis for screening potential employees.

[0051] Potential employees are those who performed relatively well in this simulation test and have the potential and qualifications to be cultivated as future leaders and managers in the company. The screening method for potential employees is as follows: Query the professional attributes of employees in the enterprise, and use the professional attributes as the index base to index the corresponding set of coefficients from the database. The weighted coefficients in the coefficient set are assigned to the business decision-making score, communication score, resource allocation score, administrative management score, and risk prevention score, respectively, and then the weighted sum is calculated to obtain the comprehensive potential score. The formula for calculating the overall potential score is: ; In the formula, To score overall potential, Scoring based on business decisions, To score points for communication and exchange, Score for resource allocation. For administrative management scores, To score for risk prevention and control, , , , , These are the weighting coefficients for the scores in business decision-making, communication, resource allocation, administration, and risk control, respectively. , , , , The sum of is 1; Analyze the overall potential scores of employees and identify those whose overall potential scores exceed a preset potential score threshold as potential employees.

[0052] It should be noted that the preset potential score threshold refers to the minimum value of the comprehensive potential score corresponding to being recorded as a potential employee, so as to ensure that potential employees participating in subsequent training and development can maintain a high level of competence and quality at the comprehensive level; and the number of potential employees in this embodiment is not fixed, and can be one or more.

[0053] The potential prediction module inputs the multimodal data and intrinsic potential characteristics of potential employees into the potential quantification and identification model to identify the training indicators for potential employees and formulate training management information. The potential quantification identification model is an artificial intelligence model based on deep learning technology. It takes multimodal data and intrinsic potential characteristics of potential employees as input data and training indicators as output data. This enables the potential quantification identification model to perform comprehensive deep learning on the contained and hidden information in multimodal data and intrinsic potential characteristics, and automatically identify the corresponding training indicators, thereby avoiding the inefficiency caused by manual learning and identification operations.

[0054] Training indicators refer to the metrics used to develop potential employees into leaders and managers within an organization. These indicators can provide a basis for the human resources department to develop relevant training plans and achieve the goal of precise management of potential employees by the human resources department. Specifically, the training indicators include professional direction, management level, development cycle, and benefit value.

[0055] Specialization refers to the specific area of ​​professional skills that potential employees will acquire during their subsequent training within the company, including but not limited to human resources organization, technology development, and financial management.

[0056] Management level refers to the specific level of the management position corresponding to a potential employee in the subsequent development of the company, including but not limited to senior management, middle management, and junior management.

[0057] The development cycle refers to the time it takes for a potential employee to reach the training goals during subsequent training within the company, including but not limited to one month, six months, twelve months, etc.

[0058] Benefit value refers to the benefits and added value that potential employees can bring to the company during subsequent training. In this embodiment, benefit value is a specific numerical value.

[0059] In this embodiment, the potential quantification identification model is developed through continuous training and optimization. When training the potential quantification identification model, it is necessary to collect a large amount of multimodal data, intrinsic potential characteristics, and training indicators of employees with different potentials in advance. The multimodal data and intrinsic potential characteristics are then summarized into comprehensive identification data. The comprehensive identification data is used as input data, and the corresponding training indicators are used as output data. The deep learning model is continuously iteratively optimized and trained until the accuracy of the training indicators output by the deep learning model reaches the preset accuracy threshold. At this point, the deep learning model will be upgraded to the potential quantification identification model.

[0060] After summarizing the multimodal data and intrinsic potential characteristics of potential employees into comprehensive identification data, it can be input into the potential quantification identification model to identify the corresponding training indicators.

[0061] After obtaining the training indicators, it is necessary to formulate corresponding training management information based on the training indicators to ensure that potential employees can be quickly and accurately trained into the talents needed by the company. This will enable the company to discover and cultivate employees with potential and value, and achieve the goal of targeted, precise and efficient human resource management.

[0062] In this embodiment, when formulating training management information, the training management information is divided into three sections: the first section is the historical section of potential employees, the second section is the testing section of potential employees, and the third section is the training section of potential employees. Specifically, multimodal data is imported into the first section, intrinsic potential characteristics are imported into the second section, and training indicators are imported into the third section, so as to realize the complete formulation of training management information.

[0063] Example 2: Please refer to Figure 2 As shown, parts not described in detail in this embodiment are described in Embodiment 1. This embodiment provides an artificial intelligence-based enterprise human resource management method, implemented through an artificial intelligence-based enterprise human resource management system, including: S01: Dynamically match employee multimodal data with profile positions to construct a digital twin profile with update nodes and regular updates; S02: Extract the portrait tags of the digital twin profile, formulate scene requirements based on the portrait tags as the core of the scene, and simulate a virtual enterprise scene that matches the scene requirements. S03: Import the digital twin profile into the enterprise virtual scene, and dynamically simulate the enterprise virtual scene under the simulation mechanism to simulate its inherent potential characteristics; S04: Convert intrinsic potential characteristics into characteristic scores, calculate comprehensive potential scores using characteristic scores, and select potential employees from among the employees; S05: Combine multimodal data and intrinsic potential characteristics to form comprehensive identification data, identify training indicators through a potential quantification identification model, and formulate training management information for potential employees based on the training indicators.

[0064] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. An enterprise human resource management system based on artificial intelligence, characterized in that, include: The profile building module is used to dynamically match the multimodal data of employees with profile positions to build a digital twin profile with update nodes and regular updates. The scene simulation module is used to extract the portrait tags of the digital twin profile, formulate scene requirements based on the portrait tags as the core of the scene, and simulate a virtual enterprise scene that matches the scene requirements. The feature simulation module is used to import digital twin profiles into the enterprise virtual scene and dynamically simulate the enterprise virtual scene under the simulation mechanism to simulate the inherent potential features; The employee screening module is used to convert intrinsic potential characteristics into feature scores, calculate comprehensive potential scores based on feature scores, and screen potential employees from among the existing employees. The potential prediction module is used to combine multimodal data and intrinsic potential characteristics into comprehensive identification data, identify training indicators through a potential quantification identification model, and formulate training management information for potential employees based on the training indicators.

2. The enterprise human resource management system based on artificial intelligence according to claim 1, characterized in that, Multimodal data includes hardware performance characteristics, software behavior characteristics, and test feedback characteristics; Hardware performance characteristics include performance evaluation values, pre-assessment rate, project participation level, project completion rate, and business skills curve; Software behavioral characteristics include effective communication rate, cooperation tendency rate, decision accuracy rate, conflict interaction rate, and stress resistance level; Test feedback characteristics include problem-solving approaches, strategic risk appetite, and learning agility.

3. The enterprise human resource management system based on artificial intelligence according to claim 2, characterized in that, The method for constructing a digital twin profile is as follows: Remove duplicate and erroneous data from the multimodal data and convert the multimodal data into a uniform format to obtain formatted data; A basic profile is simulated using digital twin technology. Three label layers, one inside and one outside, are established in the basic profile and are named the outer layer, the inner layer, and the potential layer in order from the outside to the inside. A circularly distributed image positions are set in the external layer, internal layer and potential layer respectively. With one feature corresponding to one image position as the standard, hardware performance features, software behavior features and test feedback features are imported into the image positions of the external layer, internal layer and potential layer one by one. Remove blank image positions and build a bidirectional transmission channel between two adjacent image positions, which is called the image channel; Update nodes with update cycles are configured on all portrait channels to facilitate the conversion of basic portraits into digital twin portraits.

4. The enterprise human resource management system based on artificial intelligence according to claim 3, characterized in that, Profile tags include cross-departmental resource conflicts, sudden public relations crises, disagreements on new product strategies, and low team morale; The method for extracting image tags is as follows: Based on an update cycle, the multimodal data in the external, internal, and potential layers of the digital twin profile are recorded as index data to obtain the first index data, the second index data, and the third index data. The first index data, the second index data, and the third index data are compared with the standard tags in the tag database for consistency, and the standard tags that are basically consistent or completely consistent with the comparison results are recorded as valid tags. All valid labels are compared repeatedly, and the valid labels that are repeated are recorded as portrait labels, resulting in B portrait labels.

5. The enterprise human resource management system based on artificial intelligence according to claim 4, characterized in that, The simulation method for enterprise virtual scenarios is as follows: The system retrieves employee identity and work information from the human resources database, and then binds and summarizes the identity and work information to generate basic information. Using the image tags as the core of the scene and the basic information as the scene outline, the core of the scene and the scene outline are combined into scene requirements to obtain B scene requirements. The B scenario requirements are imported one by one into the intelligent scenario engine. The scenario permissions and scenario logic of the intelligent scenario engine are set to simulate the enterprise virtual scenario corresponding to the B scenario requirements.

6. The enterprise human resource management system based on artificial intelligence according to claim 5, characterized in that, The simulation mechanism is as follows: a dynamic simulation ends when all the intrinsic potential features are simulated, and all dynamic simulations end when the overlap between any two adjacent intrinsic potential features is greater than the calibrated overlap threshold.

7. The enterprise human resource management system based on artificial intelligence according to claim 6, characterized in that, The inherent potential characteristics include business decision-making characteristics, communication characteristics, resource allocation characteristics, administrative management characteristics, and risk prevention and control characteristics; The simulation method for intrinsic potential characteristics is as follows: A10: Arrange the B enterprise virtual scenarios into a scenario queue, and import the digital twin profile into the enterprise virtual scenario that is first in the scenario queue; A11: Initialize simulation parameters in the enterprise virtual scene, send dynamic simulation commands to the enterprise virtual scene, simulate the inherent potential characteristics, and perform integrity analysis on the inherent potential characteristics; A12: When the intrinsic potential characteristic is not in a complete state, repeat A11; when the intrinsic potential characteristic is in a complete state, execute A13. A13; Remove the simulated enterprise virtual scenarios from the scenario queue, import the digital twin profile into the remaining enterprise virtual scenario in the first position of the scenario queue, and repeat A11-A13 until the overlap of any two adjacent intrinsic potential features is greater than the calibrated overlap threshold, and then obtain the intrinsic potential features.

8. The enterprise human resource management system based on artificial intelligence according to claim 7, characterized in that, The feature scores include business decision-making scores, communication scores, resource allocation scores, administrative management scores, and risk prevention and control scores; The screening method for potential employees is as follows: Query the professional attributes of employees in the enterprise, and use the professional attributes as the index base to index the corresponding set of coefficients from the database. The weighted coefficients in the coefficient set are assigned to the business decision-making score, communication score, resource allocation score, administrative management score, and risk prevention score, respectively, and then the weighted sum is calculated to obtain the comprehensive potential score. Analyze the overall potential scores of employees and identify those whose overall potential scores exceed a preset potential score threshold as potential employees.

9. The enterprise human resource management system based on artificial intelligence according to claim 8, characterized in that, The training indicators include professional direction, management level, development cycle, and benefit value; Specializations include human resources organization, technology development, and financial management; Management levels include senior management, middle management, and junior management; The molding cycle includes one month, six months, and twelve months.

10. The enterprise human resource management system based on artificial intelligence according to claim 9, characterized in that, The training management information is divided into three sections: the first section is the historical section, the second section is the testing section, and the third section is the training section. Multimodal data is imported into the historical section, intrinsic potential characteristics are imported into the testing section, and training indicators are imported into the training section to generate training management information.

Images