A digital performance evaluation method and system based on artificial intelligence

Abstract

The present invention provides a digital performance evaluation method and system based on artificial intelligence, which belongs to the field of performance evaluation technology and includes the following steps: S101: obtaining HR assessment record data, employee task log data and employee real-time behavior data; S102: establishing a performance evaluation ID that matches the data, and performing data cleaning operations; S103: establishing a unified performance evaluation data map and difficulty-contribution association network based on a cross-modal data alignment operation; S104: performing a map performance evaluation operation and a secondary performance evaluation operation to obtain a map evaluation value and a network evaluation value; S105: obtaining an evaluation threshold and performing a performance level matching operation; S106: performing a subsequent work improvement evaluation operation based on a graph neural network model and a time series prediction model and generating an employee performance report; the method can improve the objectivity and fairness of employee performance evaluation, and can provide employees with related subsequent work improvement suggestions.

Description

Technical Field

[0001] The present invention belongs to the technical field of performance evaluation, and more specifically, relates to a digital performance evaluation method and system based on artificial intelligence. Background Art

[0002] In today's highly competitive business environment, efficient business operations and development rely on a scientific and rational performance evaluation system. Traditional performance evaluation methods are often influenced by subjective factors during the evaluation process. Standards and metrics used by different evaluators are difficult to standardize, resulting in a lack of objectivity and fairness in the evaluation results. Furthermore, traditional performance evaluation methods often fail to provide employees with relevant follow-up work improvement suggestions. Summary of the Invention

[0003] In order to solve the above technical problems, the present invention provides a digital performance evaluation method and system based on artificial intelligence to solve the technical problems in the existing technology that traditional performance evaluation methods often cannot provide employees with relevant follow-up work improvement suggestions and the performance evaluation has low objectivity and fairness.

[0004] The purpose and effectiveness of the digital performance evaluation method and system based on artificial intelligence of the present invention are achieved by the following specific technical means:

[0005] A digital performance evaluation method based on artificial intelligence includes the following steps:

[0006] S101: Obtain an employee evaluation database, which is used to record and store data related to employees' daily work. Establish an API interface, establish a link with the employee evaluation database through the API interface, and call relevant data in the employee evaluation database. Based on the employee evaluation database, obtain HR assessment record data, employee task log data, and real-time employee behavior data;

[0007] HR assessment record data is used to evaluate employees' daily work status, employee task log data is used to evaluate employees' work results, and employee real-time behavior data is used to evaluate employees' daily work activity.

[0008] S102: Create a performance evaluation ID. The performance evaluation ID represents the unique authentication ID of each employee and the data that matches the employee. Match and connect HR assessment record data, employee task log data, and employee real-time behavior data with the performance evaluation ID. Perform data cleansing on the data related to the performance evaluation ID, and records the performance evaluation ID that contains abnormal data during the data cleansing operation.

[0009] S103: Perform a cross-modal data alignment operation on the performance evaluation IDs that do not contain abnormal data. Based on the cross-modal data alignment operation, a unified performance evaluation data map is established for each performance evaluation ID. The unified performance evaluation data map represents the comprehensive work performance of each performance evaluation ID. A difficulty-contribution association network is established. The difficulty-contribution association network is used to comprehensively consider the workload and difficulty of each performance evaluation ID.

[0010] S104: Based on the unified performance evaluation data graph, a graph performance evaluation operation is performed on the performance evaluation ID without abnormal data. After the evaluation is completed, a graph evaluation value is generated. Based on the difficulty-contribution value association network, a secondary performance evaluation operation is performed on the performance evaluation ID without abnormal data. After the evaluation is completed, a network evaluation value is generated. The graph evaluation value and the network evaluation value are matched and connected with the performance evaluation ID;

[0011] S105: Obtain an evaluation threshold, which is used to match the performance evaluation ID to a performance level. The evaluation threshold is regularly changed as the market environment changes. The graph evaluation value and the network evaluation value are matched to the evaluation threshold for performance level matching.

[0012] Performance level matching operations include:

[0013] S1051: If both the graph evaluation value and the network evaluation value are lower than the evaluation threshold, the performance evaluation ID is marked as unqualified;

[0014] S1052: If both the graph evaluation value and the network evaluation value are within a range close to the evaluation threshold, the performance evaluation ID within the range is marked as medium performance;

[0015] S1053: If a single value of the graph evaluation value and the network evaluation value is lower than the evaluation threshold, the performance evaluation ID is marked as pending review. After the marking is completed, the performance evaluation ID to be reviewed is imported into the expert review mechanism;

[0016] S1054: If both the graph evaluation value and the network evaluation value are higher than the close interval of the evaluation threshold, the performance evaluation ID is marked as having good performance;

[0017] S106: After the performance level matching operation is completed, the marked performance evaluation ID is imported into the graph neural network model and the time series prediction model for subsequent work improvement evaluation operation. The graph neural network model is used to perform employee capability analysis operation on the performance evaluation ID, and the time series prediction model is used to evaluate the growth trend operation of the performance evaluation ID. Based on the graph neural network model and the time series prediction model, an employee performance report matching the performance evaluation ID is generated. The employee performance report includes the performance level and subsequent work improvement suggestions. The employee performance report is forwarded to the employee matching the performance evaluation ID and the enterprise management;

[0018] Regularly perform S101-S106 to conduct performance evaluation of corporate employees.

[0019] As a further solution of the present invention, the expert review mechanism includes:

[0020] Establish a manual expert review panel to manually review the performance evaluation IDs to be reviewed. After the review is completed, remove the pending review mark from the performance evaluation IDs and mark the performance evaluation IDs as good, average, or unqualified. After the marking is completed, re-import the performance evaluation IDs into S106;

[0021] The specific steps include:

[0022] A1: Establish a human expert review panel. The panel will include multiple expert groups, including HR managers, department heads, and data analysis experts from different departments.

[0023] A2: Assign the performance evaluation IDs to be reviewed by different departments to the corresponding expert groups, obtain relevant work data based on the performance evaluation IDs to be reviewed, and forward the relevant work data to the corresponding expert groups;

[0024] A3: An expert panel conducts an independent review of the performance evaluation IDs to be reviewed, generates HR scores, supervisor scores, and data analysis scores, calculates the difference between the HR scores, supervisor scores, and data analysis scores, and screens the performance evaluation IDs to be reviewed based on the difference calculation.

[0025] A4: Calculate the average of the HR score, supervisor score, and data analysis score for the performance evaluation ID that is retained for review, and generate the review score based on the average calculation.

[0026] A5: Remove the pending review mark and mark the pending performance evaluation ID as good, average, or unqualified based on the review score. Re-import the marked performance evaluation ID into step S106.

[0027] As a further solution of the present invention, a difference calculation is performed on the human resources score, the supervisor score and the data analysis score, and the performance evaluation ID to be reviewed is screened based on the difference calculation, including:

[0028] Get three different sets of differences. The calculation formula for the three sets of differences is:

[0029] ;

[0030] in, 、 and Represent three different sets of differences, if 、 and If the absolute value of any group of differences is too large, it is considered an abnormal review situation and the performance evaluation ID is screened. After the screening, the expert group will start a review meeting for the performance evaluation ID and conduct a second performance review of the performance evaluation ID to generate a unified review score. The performance evaluation ID after the review score is generated is imported into step A5.

[0031] As a further solution of the present invention, a second performance review is conducted on the performance evaluation ID to generate a unified review score, including:

[0032] If the expert group is unable to generate a unified review score during the review meeting, the performance evaluation of the performance evaluation ID will be temporarily suspended, and then an observation period will be set for the performance evaluation ID, and the performance evaluation ID will be marked for the observation period. After the observation period ends, the relevant work data of the performance evaluation ID will be updated, and the performance evaluation ID will be re-imported into step S104.

[0033] As a further solution of the present invention, a difficulty-contribution correlation network is established, including:

[0034] Model different levels of work difficulty and different levels of work contribution;

[0035] After the modeling is completed, the network nodes are set. The network nodes include task nodes and contribution nodes. Task nodes represent the difficulty of different work tasks, and contribution nodes represent the incidental contribution of different work tasks. Based on the connection between task nodes and contribution nodes, a task-contribution connection is generated. Based on the task-contribution connection and the network nodes, a task-contribution bidirectional graph structure is established. In the task-contribution bidirectional graph structure, the task nodes and contribution nodes are connected to each other.

[0036] A difficulty-contribution association network is established based on network nodes, task-contribution connections, and the task-contribution bidirectional graph structure. After the establishment is completed, the features in the task-contribution bidirectional graph structure are extracted based on the graph convolutional neural network, and the difficulty-contribution association network is optimized. A network evaluation value is generated based on the difficulty-contribution association network.

[0037] As a further solution of the present invention, the network evaluation value can be expressed as:

[0038] ;

[0039] in, Expressed as the network evaluation value, Indicates the total frequency of employee task submissions. Represented as employee task completion rate data, Indicates that the employee Frequency data of employee task submissions submitted times, Indicates that the employee The difficulty of the work task submitted Indicates that the employee The incidental contribution of the work task submitted.

[0040] As a further aspect of the present invention, the evaluation threshold is periodically changed as the market environment changes, including:

[0041] Obtain market factors and internal enterprise factors. Market factors include industry economic data and macroeconomic indicator data, which are used to assess the impact of the external economy on the enterprise. Internal enterprise factors include enterprise profit margin data and enterprise revenue data, which are used to assess the overall operation of the enterprise.

[0042] The threshold change formula is constructed based on external market factors and internal enterprise factors. The threshold change formula can be expressed as: ;

[0043] in, Denotes the evaluation threshold, Expressed as the average of the company's employees' historical performance, Expressed as the market impact factor, It is expressed as the business operation impact factor, and Expressed as weight;

[0044] Based on the threshold change formula, the evaluation threshold that changes with the market environment is regularly obtained.

[0045] As a further solution of the present invention, HR assessment record data, employee task log data and employee real-time behavior data are obtained based on the employee evaluation database, including:

[0046] HR assessment record data includes employee attendance data, employee periodic assessment results data, and employee periodic training results data;

[0047] Employee task log data includes employee task completion rate data and employee task submission frequency data;

[0048] Employees' real-time behavior data includes meeting attendance frequency data and working hours data.

[0049] As a further solution of the present invention, a data cleaning operation is performed on the data related to the performance evaluation ID, and the performance evaluation ID with abnormal data in the data cleaning operation is recorded, including:

[0050] After the data cleaning operation is completed, a defect list is created to record the performance evaluation IDs of abnormal data. The defect list is sent to the operator in real time in the form of a reminder pop-up box to remind the operator to supplement the abnormal data.

[0051] If the data supplementation is completed, the data cleaning operation is performed again on the performance evaluation IDs for which the data supplementation is completed, and the performance evaluation IDs for which no abnormal data appears in the data supplementation are re-imported into S103.

[0052] A digital performance evaluation system based on artificial intelligence, comprising:

[0053] Data module, which includes a data acquisition module and a data processing module;

[0054] The data acquisition module is used to establish an API interface to obtain the required relevant data;

[0055] The data processing module is used to clean the data related to the performance evaluation ID, perform cross-modal data alignment on the performance evaluation IDs without abnormal data, and move the performance evaluation IDs with abnormal data to the defect list;

[0056] A threshold module is used to obtain an evaluation threshold and adjust the evaluation threshold based on a threshold change formula, so that the evaluation threshold can be changed regularly as the market environment changes;

[0057] The authentication module is used to establish a performance evaluation ID and match the relevant data with each employee's performance evaluation ID. In addition, the performance evaluation ID that is undergoing performance evaluation can be marked.

[0058] An evaluation module is used to establish a performance evaluation data map and a difficulty-contribution correlation network, and obtain network evaluation values ​​and map evaluation values ​​based on the performance evaluation data map and the difficulty-contribution correlation network. The network evaluation values ​​and the map evaluation values ​​are matched with each other based on the evaluation threshold. After the performance level matching operation is completed, the performance evaluation ID is evaluated for subsequent work improvements based on the graph neural network model and the time series prediction model. An employee performance report matching the performance evaluation ID is generated. The employee performance report includes the performance level and subsequent work improvement suggestions. The employee performance report is forwarded to the employee matching the performance evaluation ID and the enterprise management.

[0059] An execution module, which regularly executes S101-S106 of the method to perform performance evaluation on enterprise employees;

[0060] The review module is used to implement the expert review mechanism.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] First, obtain the employee evaluation database, establish a connection with the employee evaluation database through the API interface, obtain HR assessment record data, employee task log data and employee real-time behavior data based on the employee evaluation database, and after the acquisition is completed, establish a performance evaluation ID, match the HR assessment record data, employee task log data and employee real-time behavior data with the performance evaluation ID and establish a connection, perform data cleaning operations, and record the performance evaluation ID of abnormal data to prevent data errors during the evaluation process and improve the accuracy of subsequent performance evaluation operations. Then, based on the cross-modal data alignment operation, establish a unified performance evaluation data map that matches each performance evaluation ID, and establish a difficulty-contribution value association network. After the establishment is completed, perform map performance evaluation operations and secondary performance evaluation operations to generate map evaluation values ​​and network evaluation values, obtain evaluation thresholds, and perform performance level matching operations based on the evaluation thresholds. In addition, introduce experts when performing performance level matching operations. The review mechanism, after the performance level matching operation is completed, performs subsequent work improvement evaluation operations based on the graph neural network model and the time series prediction model to generate an employee performance report that matches the performance evaluation ID. The employee performance report includes the performance level and subsequent work improvement suggestions. Finally, the employee performance report is forwarded to the employee and the corporate management that match the performance evaluation ID. When conducting performance evaluation, this method can automatically read the relevant data required for employee performance evaluation and evaluate employee performance by introducing models and associated networks, thereby improving the objectivity of employee performance evaluation and preventing the standards and scales of different evaluators from being inconsistent in manual evaluation. When conducting employee performance evaluation, an expert review mechanism is introduced to conduct a secondary evaluation of the performance of employees with medium and low performance through an expert review panel, thereby improving fairness and being more humane. At the same time, this method can also generate employee performance reports based on the employee's relevant work data, and the report includes performance levels and subsequent work improvement suggestions. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a flowchart of the steps of a digital performance evaluation method based on artificial intelligence of the present invention;

[0064] Figure 2 This is a flowchart of the performance level matching operation in step S105 of a digital performance evaluation method based on artificial intelligence of the present invention;

[0065] Figure 3 This is a flowchart of the steps of the expert review mechanism in the digital performance evaluation method based on artificial intelligence of the present invention. DETAILED DESCRIPTION

[0066] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following examples are used to illustrate the technical solutions of the present invention, but are not intended to limit the scope of protection of the present invention.

[0067] Example 1: As shown in the attached Figure 1 、 Figure 2 、 Figure 3 As shown:

[0068] The present invention provides a digital performance evaluation method based on artificial intelligence, which is applicable to performance evaluation and includes the following steps:

[0069] S101: Obtain an employee evaluation database, which is used to record and store data related to employees' daily work. Establish an API interface, establish a link with the employee evaluation database through the API interface, and call relevant data in the employee evaluation database. Based on the employee evaluation database, obtain HR assessment record data, employee task log data, and employee real-time behavior data.

[0070] Among them, HR assessment record data is used to evaluate employees' daily work status, employee task log data is used to evaluate employees' work results, and employee real-time behavior data is used to evaluate employees' daily work activity.

[0071] Furthermore, HR assessment record data includes employee attendance data, employee periodic assessment result data, and employee periodic training result data. Through the selection of HR assessment record data, attendance data and assessment scores are based on objective records, reducing human subjective intervention. Periodic assessments can reflect employees' long-term work performance rather than just short-term fluctuations. Training data can measure whether employees are actively improving their own abilities and whether they are adapting to job development.

[0072] Employee task log data includes employee task completion rate data and employee task submission frequency data. Through the selection of employee task log data, task completion rate is the most intuitive KPI, which can quantify employee contributions. Task submission frequency can reflect the employee's work rhythm and whether there are procrastination problems. Task logs are automatically recorded by the system, avoiding human tampering or misjudgment.

[0073] Employee real-time behavior data includes meeting attendance frequency data and working hours data. Through the selection of employee real-time behavior data, the frequency of meeting participation can reflect whether employees are willing to communicate and collaborate, and working hours data can reveal whether employees are working efficiently or working overtime inefficiently. If the task completion rate is high and the working hours are moderate, it means that employees are working efficiently; otherwise, there may be a problem of inefficient work.

[0074] If only one type of data is used, it may lead to one-sided conclusions. Therefore, combining the three types of data can more accurately assess the true working status of employees. This data-driven performance evaluation method can help companies optimize the performance appraisal system and improve the accuracy and fairness of employee management.

[0075] S102: Establish a performance evaluation ID. The performance evaluation ID represents the unique authentication ID of each employee and the data that matches the employee. Match the HR assessment record data, employee task log data, and employee real-time behavior data with the performance evaluation ID and establish a connection. Perform data cleaning operations on the relevant data of the performance evaluation ID, and record the performance evaluation ID that has abnormal data during the data cleaning operation.

[0076] It is understandable that the performance evaluation ID is the unique identity of each employee in the performance appraisal system. It is mainly used to match and integrate data from different sources (task logs, behavioral data, etc.). Through the setting of a uniquely certified performance evaluation ID, all performance-related data are bound to the ID to avoid data loss or duplicate calculations. It can also avoid the use of sensitive information such as real names for data processing, reduce the risk of privacy leakage, and ensure data consistency during cross-departmental analysis.

[0077] Specifically, data cleaning operations can be used to remove erroneous, redundant or abnormal data to ensure the accuracy of performance evaluation. When selecting the specific algorithm to use for the data cleaning operation, algorithms such as anomaly detection algorithms can be selected. If the user needs to prevent data tampering, anti-tampering related encryption storage and permission control can be set to ensure objectivity and fairness when this method is used to automatically evaluate employee performance.

[0078] Furthermore, after the data cleaning operation is completed, a defect list is established. The defect list is used to record the performance evaluation IDs with abnormal data, and the defect list is sent to the operator in real time in the form of a reminder pop-up box to remind the operator to supplement the abnormal data. After receiving the reminder pop-up box, the operator can promptly remind the employees whose performance evaluation IDs match the defect list, thereby supplementing the missing data. The data cleaning operation can prevent data errors during the evaluation process and improve the accuracy of subsequent performance evaluation operations; if the data supplementation is completed, the data cleaning operation is performed again on the performance evaluation ID that has completed data supplementation. The data cleaning operation is performed again on the performance evaluation ID that has completed supplementation, which can prevent data errors again. For example, when an employee's average task completion rate is greater than 100% or the working hours exceed 24 hours, the erroneous data will be cleaned to avoid sudden abnormalities. If the performance evaluation ID still has abnormal data, repeat the above steps and re-import the performance evaluation ID that has completed data supplementation and has no abnormal data into S103.

[0079] S103: Perform a cross-modal data alignment operation on the performance evaluation IDs without abnormal data, and establish a unified performance evaluation data map that matches each performance evaluation ID based on the cross-modal data alignment operation. The unified performance evaluation data map represents the comprehensive work performance of each performance evaluation ID, and a difficulty-contribution association network is established. The difficulty-contribution association network is used to comprehensively consider the workload and difficulty of each performance evaluation ID.

[0080] Understandably, in the performance evaluation process, data sources are diverse, including HR assessment record data, employee task log data, employee real-time behavior data, etc. These data usually belong to different modalities, that is, the data types, formats, and sources are different. For example, attendance data is structured tabular data, while employee behavior data may come from unstructured log records, which makes it difficult to directly integrate these data for analysis. Therefore, cross-modal data alignment is required to unify data from these different sources. When establishing a unified performance evaluation data graph, a knowledge graph structure can be used for construction. At the same time, when obtaining graph evaluation values, the graph evaluation values ​​are obtained based on the knowledge graph structure.

[0081] Furthermore, when establishing the difficulty-contribution correlation network:

[0082] Model different levels of work difficulty and different levels of work contribution;

[0083] After the modeling is completed, the network nodes are set. The network nodes include task nodes and contribution nodes. Task nodes represent the difficulty of different work tasks, and contribution nodes represent the incidental contribution of different work tasks. Based on the connection between task nodes and contribution nodes, a task-contribution connection is generated. Based on the task-contribution connection and the network nodes, a task-contribution bidirectional graph structure is established. In the task-contribution bidirectional graph structure, the task nodes and contribution nodes are connected to each other.

[0084] A difficulty-contribution association network is established based on network nodes, task-contribution connections, and the task-contribution bidirectional graph structure. After the establishment is completed, the features in the task-contribution bidirectional graph structure are extracted based on the graph convolutional neural network, and the difficulty-contribution association network is optimized. Based on the difficulty-contribution association network, a network evaluation value is generated.

[0085] S104: Based on the unified performance evaluation data graph, a graph performance evaluation operation is performed on the performance evaluation ID without abnormal data. After the evaluation is completed, a graph evaluation value is generated. Based on the difficulty-contribution value association network, a secondary performance evaluation operation is performed on the performance evaluation ID without abnormal data. After the evaluation is completed, a network evaluation value is generated. The graph evaluation value and the network evaluation value are matched and connected with the performance evaluation ID;

[0086] Among them, the network evaluation value can be expressed as:

[0087] ;

[0088] in, Expressed as the network evaluation value, Indicates the total frequency of employee task submissions. Represented as employee task completion rate data, Indicates that the employee Frequency data of employee task submissions submitted times, Indicates that the employee The difficulty of the work task submitted Indicates that the employee The incidental contribution of the work task submitted.

[0089] S105: Obtain an evaluation threshold. The evaluation threshold is used to match the performance evaluation ID to a performance level. The evaluation threshold is changed regularly as the market environment changes. The graph evaluation value and the network evaluation value are matched with the evaluation threshold to perform a performance level matching operation.

[0090] Specifically, the assessment thresholds are regularly changed as the market environment changes, including:

[0091] Obtain market factors and internal enterprise factors. Market factors include industry economic data and macroeconomic indicator data. Market factors are used to evaluate the impact of the external economy on the enterprise. Internal enterprise factors include enterprise profit margin data and enterprise revenue data. Internal enterprise factors are used to evaluate the overall situation of the enterprise's operations. The thresholds in traditional performance evaluation methods are often static and cannot respond to fluctuations in the external environment in a timely manner. The performance standards of the enterprise should be dynamically adjusted as the market environment changes.

[0092] It is understandable that market factors include industry economic data and macroeconomic indicator data. Industry economic data evaluates industry trends through industry growth rate, market demand changes, competitor data, etc., while macroeconomic indicator data include GDP growth rate, CPI (Consumer Price Index), PPI (Producer Price Index), interest rate changes and other factors. Internal factors of the enterprise include enterprise profit margin data and enterprise revenue data. Enterprise profit margin and enterprise revenue data can reflect the overall situation of the enterprise's operations.

[0093] The threshold change formula is constructed based on external market factors and internal enterprise factors. The threshold change formula can be expressed as:

[0094] ;

[0095] in, Denotes the evaluation threshold, Expressed as the average of the company's employees' historical performance, Expressed as the market impact factor, It is expressed as the business operation impact factor, and Expressed as weights, the evaluation threshold that changes with the market environment is regularly obtained based on the threshold change formula.

[0096] Furthermore, the performance level matching operation includes:

[0097] S1051: If both the graph evaluation value and the network evaluation value are lower than the evaluation threshold, the performance evaluation ID is marked as unqualified.

[0098] S1052: If both the graph evaluation value and the network evaluation value are in a range close to the evaluation threshold, the performance evaluation ID in the range is marked as medium performance.

[0099] S1053: If a single value of the graph evaluation value and the network evaluation value is lower than the evaluation threshold, the performance evaluation ID is marked as pending review. After the marking is completed, the performance evaluation ID to be reviewed is imported into the expert review mechanism.

[0100] Specifically, the expert review mechanism includes:

[0101] Establish a manual expert review panel to manually review the performance evaluation IDs to be reviewed. After the review is completed, remove the pending review mark from the performance evaluation IDs and mark the performance evaluation IDs as good, average, or unqualified. After the marking is completed, re-import the performance evaluation IDs into S106;

[0102] The specific steps include:

[0103] A1: Establish a human expert review panel. The panel will include multiple expert groups, including HR managers, department heads, and data analysis experts from different departments.

[0104] A2: Assign the performance evaluation IDs to be reviewed by different departments to the corresponding expert groups, obtain relevant work data based on the performance evaluation IDs to be reviewed, and forward the relevant work data to the corresponding expert groups;

[0105] A3: An expert panel conducts an independent review of the performance evaluation IDs to be reviewed, generates HR scores, supervisor scores, and data analysis scores, calculates the difference between the HR scores, supervisor scores, and data analysis scores, and screens the performance evaluation IDs to be reviewed based on the difference calculation.

[0106] A4: Calculate the average of the HR score, supervisor score, and data analysis score for the performance evaluation ID that is retained for review, and generate the review score based on the average calculation.

[0107] A5: Remove the pending review mark and mark the pending performance evaluation ID as good, average, or unqualified based on the review score. Re-import the marked performance evaluation ID into step S106.

[0108] It is understandable that although artificial intelligence can provide efficient, data-driven performance evaluation, in some complex or borderline situations, relying entirely on algorithms may lead to misjudgments. Therefore, the introduction of an expert review mechanism can effectively improve fairness. For example, the use of an expert review mechanism can take special circumstances into account and avoid a one-size-fits-all approach. For the phenomenon that traditional performance evaluation methods all use artificial intelligence, the use of an expert review mechanism can consider and conduct secondary reviews of individual special circumstances, improve fairness, and reduce employee dissatisfaction and turnover rates.

[0109] For example, an employee's short-term performance may decline due to participation in a high-difficulty project, but his long-term value is high. One of the values ​​given by AI may be lower than the evaluation threshold, and the expert review mechanism can adjust the evaluation results based on actual conditions. For employees who have just joined the company or have been temporarily transferred, the algorithm may evaluate the value to be low due to insufficient data. Experts can provide additional judgment from a business perspective.

[0110] Furthermore, three different sets of differences are obtained, and the calculation formula for the three sets of differences is:

[0111] ;

[0112] in, 、 and Represent three different sets of differences, if 、 and If the absolute value of any group of differences is too large, it is considered an abnormal review situation and the performance evaluation ID is screened. After the screening, the expert group will start a review meeting for the performance evaluation ID and conduct a second performance review of the performance evaluation ID to generate a unified review score. The performance evaluation ID after the review score is generated is imported into step A5.

[0113] Furthermore, if the expert group is unable to generate a unified review score during the review meeting, the performance evaluation of the performance evaluation ID will be temporarily suspended, and then an observation period will be set for the performance evaluation ID. The observation period can be set to one to two months. By setting the observation period, employees can be given the opportunity to re-evaluate, and it is avoided that the expert group cannot reach a unified review result and adopts an unreasonable evaluation method. The performance evaluation ID is marked as an observation period. After the observation period ends, the relevant work data of the performance evaluation ID is updated, and the performance evaluation ID is re-imported into step S104.

[0114] S1054: If both the graph evaluation value and the network evaluation value are higher than the approximate interval of the evaluation threshold, the performance evaluation ID is marked as having good performance.

[0115] S106: After the performance level matching operation is completed, the marked performance evaluation ID is imported into the graph neural network model and the time series prediction model for subsequent work improvement evaluation operations. The graph neural network model is used to perform employee capability analysis operations on the performance evaluation ID, and the time series prediction model is used to evaluate the growth trend of the performance evaluation ID. Based on the graph neural network model and the time series prediction model, an employee performance report matching the performance evaluation ID is generated. The employee performance report includes performance levels and subsequent work improvement suggestions, and the employee performance report is forwarded to the employee matching the performance evaluation ID and the corporate management.

[0116] Specifically, the graph neural network model and the time series prediction model can be combined with the results to generate personalized follow-up work improvement suggestions, rather than simple scoring, so that employees can improve themselves more clearly. For example, the follow-up work improvement suggestions for technical talents may suggest that they participate in professional skills training to improve specific technical capabilities, while for management talents, they may be advised to enhance team communication and leadership, and improve management skills through training or actual tasks. Management can also optimize talents based on employee performance reports. For example, some employees can be promoted, unsuitable employees can be transferred to other positions or departments, or employees can be optimized.

[0117] Regularly perform S101-S106 to conduct performance evaluation of corporate employees.

[0118] A digital performance evaluation system based on artificial intelligence, comprising:

[0119] Data module, which includes a data acquisition module and a data processing module;

[0120] The data acquisition module is used to establish an API interface to obtain the required relevant data;

[0121] The data processing module is used to clean the data related to the performance evaluation ID, perform cross-modal data alignment on the performance evaluation IDs without abnormal data, and move the performance evaluation IDs with abnormal data to the defect list;

[0122] A threshold module is used to obtain an evaluation threshold and adjust the evaluation threshold based on a threshold change formula, so that the evaluation threshold can be changed regularly as the market environment changes;

[0123] The authentication module is used to establish a performance evaluation ID and match the relevant data with each employee's performance evaluation ID. In addition, the performance evaluation ID that is undergoing performance evaluation can be marked.

[0124] An evaluation module is used to establish a performance evaluation data map and a difficulty-contribution correlation network, and obtain network evaluation values ​​and map evaluation values ​​based on the performance evaluation data map and the difficulty-contribution correlation network. The network evaluation values ​​and the map evaluation values ​​are matched with each other based on the evaluation threshold. After the performance level matching operation is completed, the performance evaluation ID is evaluated for subsequent work improvements based on the graph neural network model and the time series prediction model. An employee performance report matching the performance evaluation ID is generated. The employee performance report includes the performance level and subsequent work improvement suggestions. The employee performance report is forwarded to the employee matching the performance evaluation ID and the enterprise management.

[0125] An execution module, which regularly executes S101-S106 of the method to perform performance evaluation on enterprise employees;

[0126] The review module is used to implement the expert review mechanism.

[0127] The specific usage and function of the first embodiment of this invention are as follows: First, obtain the employee evaluation database, establish a connection with the employee evaluation database through the API interface, obtain HR assessment record data, employee task log data and employee real-time behavior data based on the employee evaluation database, and after the acquisition is completed, establish a performance evaluation ID, match the HR assessment record data, employee task log data and employee real-time behavior data with the performance evaluation ID and establish a connection, perform data cleaning operations, and record the performance evaluation ID of abnormal data to prevent data errors during the evaluation process and improve the accuracy of subsequent performance evaluation operations. Then, based on the cross-modal data alignment operation, establish a unified performance evaluation data map for each performance evaluation ID, and establish a difficulty-contribution value association network. After the establishment is completed, perform map performance evaluation operations and secondary performance evaluation operations to generate map evaluation values ​​and network evaluation values, obtain an evaluation threshold, perform a performance level matching operation based on the evaluation threshold, and perform a performance level matching operation. An expert review mechanism is introduced during the matching operation. After the performance level matching operation is completed, a subsequent work improvement evaluation operation is performed based on the graph neural network and time series prediction model to generate an employee performance report that matches the performance evaluation ID. The employee performance report includes the performance level and subsequent work improvement suggestions. Finally, the employee performance report is forwarded to the employee and the corporate management that match the performance evaluation ID. When conducting performance evaluation, this method can automatically read the relevant data required for employee performance evaluation and evaluate employee performance by introducing models and association networks, thereby improving the objectivity of employee performance evaluation and preventing the standards and scales of different evaluators from being inconsistent in manual evaluation. When conducting employee performance evaluation, an expert review mechanism is introduced to conduct a secondary evaluation of the performance of employees with medium and low performance through an expert review panel, thereby improving fairness and being more humane. At the same time, this method can also generate an employee performance report based on the employee's relevant work data, which includes performance levels and subsequent work improvement suggestions.

[0128] An electronic device, comprising:

[0129] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method proposed in the first embodiment of the present invention.

[0130] The following is a detailed introduction to the various components of electronic equipment:

[0131] The term "processor" is the control center of an electronic device and can be a single processor or a collective term for multiple processing elements. For example, the processor can be one or more central processing units (CPUs), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the first embodiment of the present invention, such as one or more digital signal processors (DSPs) or one or more field programmable gate arrays (FPGAs).

[0132] The processor can execute various functions of the electronic device by running or executing software programs stored in the memory and calling data stored in the memory.

[0133] The memory is used to store the software program for executing the solution of the present invention, and the execution is controlled by the processor. The specific implementation method can refer to the above method embodiment and will not be repeated here.

[0134] The memory may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, a random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only (CD-ROM), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. The memory may be integrated with the processor or exist independently and be coupled to the processor via an interface circuit of the electronic device, and this is not specifically limited in the embodiments of the present invention.

[0135] The above embodiments can be implemented in whole or in part via software, hardware (e.g., circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When loaded or executed on a computer, the processes or functions described in accordance with the embodiments of the present invention are fully or partially performed. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wireless communication (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium accessible by a computer, or a data storage device such as a server or data center that contains a collection of one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0136] It should be understood that the term "and / or" as used herein simply describes an association between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent the existence of A alone, the existence of both A and B, or the existence of B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the related objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0137] It should be understood that in the embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0138] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A digital performance evaluation method based on artificial intelligence, characterized in that: The following steps are included: S101: Obtain the employee evaluation database and call the HR assessment record data, employee task log data, and employee real-time behavior data in the database through the API interface; S102: Create a performance evaluation ID, match HR assessment record data, employee task log data, and employee real-time behavior data with the performance evaluation ID and establish a connection, perform data cleaning operations on the relevant data connected to the performance evaluation ID and record abnormal data; S103: Perform cross-modal data alignment on performance evaluation IDs that do not contain abnormal data, construct a unified performance evaluation data map to represent the overall performance of employees, and establish a difficulty-contribution correlation network to comprehensively consider the workload and difficulty of each performance evaluation ID; S104: Generate a graph evaluation value based on the unified performance evaluation data graph, perform a secondary performance evaluation operation on the performance evaluation ID without abnormal data based on the difficulty-contribution value association network, generate a network evaluation value after the evaluation is completed, and match the graph evaluation value with the network evaluation value and the performance evaluation ID; Build a difficulty-contribution correlation network, including: Model different levels of work difficulty and different levels of work contribution; After the modeling is completed, the network nodes are set. The network nodes include task nodes and contribution nodes. Task nodes represent the difficulty of different work tasks, and contribution nodes represent the incidental contribution of different work tasks. Based on the connection between task nodes and contribution nodes, a task-contribution connection is generated. Based on the task-contribution connection and the network nodes, a task-contribution bidirectional graph structure is established. In the task-contribution bidirectional graph structure, the task nodes and contribution nodes are connected to each other. A difficulty-contribution association network is established based on network nodes, task-contribution connections, and the task-contribution bidirectional graph structure. After the establishment is completed, features in the task-contribution bidirectional graph structure are extracted using a graph convolutional neural network, and the difficulty-contribution association network is optimized. A network evaluation value is generated based on the difficulty-contribution association network; The network evaluation value can be expressed as: ； in, Expressed as the network evaluation value, Indicates the total frequency of employee task submissions. Represented as employee task completion rate data, Indicates that the employee Frequency data of employee task submissions submitted times, Indicates that the employee The difficulty of the work task submitted Indicates that the employee The incidental contribution of the work tasks submitted; S105: Obtain an evaluation threshold, which is used to match the performance evaluation ID to a performance level. The evaluation threshold is regularly changed as the market environment changes. The graph evaluation value and the network evaluation value are matched to the evaluation threshold for performance level matching. S106: Input the performance evaluation ID into the graph neural network model and the time series prediction model to generate an employee performance report containing performance grades and improvement suggestions, and distribute it to employees and management; Regularly perform S101-S106 to conduct performance evaluation of corporate employees.

2. The digital performance evaluation method based on artificial intelligence according to claim 1 is characterized in that: Performance level matching operations include: S1051: If both the graph evaluation value and the network evaluation value are lower than the evaluation threshold, the performance evaluation ID is marked as unqualified; S1052: If both the graph evaluation value and the network evaluation value are within a range close to the evaluation threshold, the performance evaluation ID within the range is marked as medium performance; S1053: If a single value of the graph evaluation value and the network evaluation value is lower than the evaluation threshold, the performance evaluation ID is marked as pending review. After the marking is completed, the performance evaluation ID to be reviewed is imported into the expert review mechanism; S1054: If both the graph evaluation value and the network evaluation value are higher than the close interval of the evaluation threshold, the performance evaluation ID is marked as having good performance; The expert review mechanism includes: Establish a manual expert review panel to manually review the performance evaluation IDs to be reviewed. After the review is completed, remove the pending review mark from the performance evaluation IDs and mark the performance evaluation IDs as good, average, or unqualified. After the marking is completed, re-import the performance evaluation IDs into S106; The specific steps include: A1: Establish a human expert review panel. The panel will include multiple expert groups, including HR managers, department heads, and data analysis experts from different departments. A2: Assign the performance evaluation IDs to be reviewed by different departments to the corresponding expert groups, obtain relevant work data based on the performance evaluation IDs to be reviewed, and forward the relevant work data to the corresponding expert groups; A3: An expert panel conducts an independent review of the performance evaluation IDs to be reviewed, generates HR scores, supervisor scores, and data analysis scores, calculates the difference between the HR scores, supervisor scores, and data analysis scores, and screens the performance evaluation IDs to be reviewed based on the difference calculation. A4: Calculate the average of the HR score, supervisor score, and data analysis score for the performance evaluation ID that is retained for review, and generate the review score based on the average calculation. A5: Remove the pending review mark and mark the pending performance evaluation ID as good, average, or unqualified based on the review score. Re-import the marked performance evaluation ID into step S106.

3. The digital performance evaluation method based on artificial intelligence according to claim 2 is characterized in that: Calculate the difference between the human resources score, supervisor score, and data analysis score, and filter the performance evaluation IDs to be reviewed based on the difference calculation, including: Get three different sets of differences. The calculation formula for the three sets of differences is: ； in, 、 and Represent three different sets of differences, if 、 and If the absolute value of any group of differences is too large, it is considered an abnormal review situation and the performance evaluation ID is screened. After the screening, the expert group will start a review meeting for the performance evaluation ID and conduct a second performance review of the performance evaluation ID to generate a unified review score. The performance evaluation ID after the review score is generated is imported into step A5.

4. The digital performance evaluation method based on artificial intelligence according to claim 3 is characterized in that: Conduct a second performance review on the performance evaluation ID to generate a unified review score, including: If the expert group is unable to generate a unified review score during the review meeting, the performance evaluation of the performance evaluation ID will be temporarily suspended, and then an observation period will be set for the performance evaluation ID, and the performance evaluation ID will be marked for the observation period. After the observation period ends, the relevant work data of the performance evaluation ID will be updated, and the performance evaluation ID will be re-imported into step S104.

5. The digital performance evaluation method based on artificial intelligence according to claim 1 is characterized in that: Assessment thresholds are regularly updated based on market conditions, including: Obtain market factors and internal enterprise factors. Market factors include industry economic data and macroeconomic indicator data, which are used to assess the impact of the external economy on the enterprise. Internal enterprise factors include enterprise profit margin data and enterprise revenue data, which are used to assess the overall operation of the enterprise. The threshold change formula is constructed based on external market factors and internal enterprise factors. The threshold change formula can be expressed as: ； in, Denotes the evaluation threshold, Expressed as the average of the historical performance of the company's employees, Expressed as the market impact factor, It is expressed as the business operation impact factor, and Expressed as weight; Based on the threshold change formula, the evaluation threshold that changes with the market environment is regularly obtained.

6. The digital performance evaluation method based on artificial intelligence according to claim 1 is characterized in that: HR assessment record data includes employee attendance data, employee periodic assessment results data, and employee periodic training results data; employee task log data includes employee task completion rate data and employee task submission frequency data; employee real-time behavior data includes meeting attendance frequency data and working hours data.

7. The digital performance evaluation method based on artificial intelligence according to claim 1 is characterized in that: After the data cleaning operation is completed, a defect list is created to record the performance evaluation IDs of abnormal data. The defect list is sent to the operator in real time in the form of a reminder pop-up box to remind the operator to supplement the abnormal data. If the data supplementation is completed, the data cleaning operation is performed again on the performance evaluation IDs for which the data supplementation is completed, and the performance evaluation IDs for which no abnormal data appears in the data supplementation are re-imported into S103.

8. A digital performance evaluation system based on artificial intelligence, used to implement the method described in claim 1, characterized in that: include: Data module, which includes a data acquisition module and a data processing module; The data acquisition module calls HR assessment record data, employee task log data, and employee real-time behavior data through the API interface; The data processing module is used to clean the data related to the performance evaluation ID and record abnormal data; The authentication module is used to establish a performance evaluation ID and match and connect HR assessment record data, employee task log data, and employee real-time behavior data with the performance evaluation ID; An evaluation module is used to establish a performance evaluation data map and a difficulty-contribution correlation network, and obtain network evaluation values ​​and map evaluation values ​​based on the performance evaluation data map and the difficulty-contribution correlation network. The network evaluation values ​​and the map evaluation values ​​are matched with each other based on the evaluation threshold. After the performance level matching operation is completed, the performance evaluation ID is evaluated for subsequent work improvements based on the graph neural network model and the time series prediction model. An employee performance report matching the performance evaluation ID is generated. The employee performance report includes the performance level and subsequent work improvement suggestions. The employee performance report is forwarded to the employee matching the performance evaluation ID and the enterprise management. The execution module regularly executes S101-S106 to conduct performance evaluation operations on enterprise employees.

Images