A personnel system-based employee profile intelligent maintenance method and system

Abstract

The present application relates to the technical field of data processing, and more particularly to a kind of employee archives intelligent maintenance method and system based on personnel system, comprising: obtaining operation log by listening to business system, using natural language processing technology to parse the triple containing employee identification, skill label and timestamp, combining time interval, individual forgetting and technical iteration decay factor, calculating current mastery score through double-factor decay model, comparing score with threshold to update to skilled usable, skill warning or forgotten archive state, when in warning state, difference learning data is pushed to terminal according to version difference, and after learning, asymptotic formula is used to compensate calculation to restore skilled state.The present application can quantify the real usable level of employee under the change of time and technology, effectively avoid project misassignment risk, at the same time, directional knowledge compensation for skill recession and dynamic closed-loop maintenance of archive data are realized, and the accuracy of man-job matching is improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for intelligent maintenance of employee files based on a human resources system. Background Technology

[0002] As enterprises deepen their digital transformation, human resource management systems have become a core infrastructure for internal talent assessment and project assignment. In high-tech R&D or high-end manufacturing companies with matrix management structures, employees often need to be deployed across business lines, and accurate competency profiles are a prerequisite for efficient person-job matching. Employees' professional skills have an objective attribute of fluctuating over time. Not only will skills not used for a long time gradually become rusty according to individual memory patterns, but the value of skills themselves also objectively depreciates with industry updates due to the frequent iterations of modern technology stacks and equipment procedures. Therefore, modern HR systems must have the ability to dynamically track and accurately quantify employees' tacit skills.

[0003] Existing HR systems generally adopt a one-time manual entry or a static storage model that relies on the validity period of certificates. Once a skill tag is acquired, it becomes fixed. This model makes it difficult to objectively reflect the true usability of employees under the dual influence of the passage of time and technological changes. This leads to a disconnect between the proficiency status presented in the file and the actual ability, which can easily result in the wrong assignment of employees with insufficient actual competence when allocating business projects, thereby causing the risk of project delays. Summary of the Invention

[0004] To overcome the above shortcomings, this invention provides an intelligent employee file maintenance method and system based on a human resources system, aiming to improve the problem that existing static skill files are difficult to reflect the true usability of employees, thus leading to mismatch between people and positions and misassignment of projects.

[0005] In a first aspect, the present invention provides the following technical solution: an intelligent maintenance method for employee files based on a personnel system, comprising:

[0006] S1. Obtain the operation log data of the target employee by monitoring the preset business system, and use natural language processing technology to parse out the skill trigger triplet from it. The skill trigger triplet includes the employee identifier, skill tag and trigger timestamp.

[0007] S2. Calculate the time interval between the skill tag and the current time based on the trigger timestamp, and calculate the current mastery score of the target employee for the skill tag by combining the individual forgetting factor and the technology iteration decay factor through a preset two-factor decay model.

[0008] S3. Compare the current mastery score with a preset status threshold, and update the file status of the skill tag of the target employee in the human resources system according to the comparison result. The file status includes proficient and usable status, skill warning status and forgotten status.

[0009] S4. When the file status is updated to the skill warning status, generate corresponding differential learning data based on the obtained version change difference information of the skill tag, and push the differential learning data to the learning terminal of the target employee.

[0010] S5. Monitor the completion status of the learning terminal on the differential learning data. When the completion status is determined to be completed, use a preset asymptotic compensation formula to compensate the current mastery score and update the file status to the proficient and usable status.

[0011] Preferably, in step S1, the step of parsing the skill trigger triples using natural language processing technology includes:

[0012] The operation log data is cleaned by text cleaning, and the cleaned log text is converted into a standardized word vector sequence.

[0013] The word vector sequence is used to extract features using a pre-defined named entity recognition model to identify skill-specific nouns representing technical objects;

[0014] Extract the trigger timestamp and employee identifier corresponding to the operation log data, and structurally bind them with the skill proper noun to construct the skill trigger triplet.

[0015] Preferably, in step S2, the step of combining the individual forgetting factor and the technology iteration decay factor includes:

[0016] Extract the historical trigger frequency characteristics of the target employee for the skill tag, and calculate the individual forgetting factor that characterizes the strength of individual memory retention.

[0017] Obtain the version change statistics of the skill tags in the knowledge graph within a preset time window, and generate the technology iteration decay factor that represents the objective depreciation rate of the technology accordingly.

[0018] The individual forgetting factor and the technology iteration decay factor are normalized to eliminate the dimensional differences between them, thereby forming a standardized feature factor for subsequent calculations.

[0019] Preferably, in step S2, the step of calculating the target employee's current mastery score for the skill tag includes:

[0020] Retrieve the baseline mastery score of the target employee when the skill tag was last triggered, and use it as the initial base for the calculation period;

[0021] Substitute the time interval and the individual forgetting factor into a preset exponential smoothing equation to calculate the individual forgetting loss value in the time dimension;

[0022] The individual forgetting loss value is subtracted from the initial base, and the result is multiplied by the technology iteration decay factor to obtain the calculated current mastery score.

[0023] Preferably, in step S3, the step of comparing the current mastery score with a preset state threshold includes:

[0024] The system retrieves preset high-level security thresholds and low-level early warning thresholds as the state thresholds;

[0025] The current mastery score is compared with the high-level safety threshold and the low-level warning threshold to determine the numerical range of the current mastery score.

[0026] Based on the determination result of the numerical range, the file status of the skill tag is updated to the proficient and usable status, the skill warning status, or the forgotten status.

[0027] Preferably, in step S4, the step of generating corresponding differential learning data based on the obtained version change difference information of the skill tags includes:

[0028] Compare the historical usage versions of the skill tags with the current mainstream versions, and extract the difference feature paths as the version iteration difference information;

[0029] Based on the difference feature path, feature retrieval is performed in the preset training material library to retrieve matching multimedia clips;

[0030] The multimedia segments are assembled and spliced ​​with a preset mini-test question bank to generate differential learning data used to compensate for version generation differences.

[0031] Preferably, in step S4, the step of pushing the differential learning data to the learning terminal of the target employee includes:

[0032] The system activity status of the target employee is detected, and the currently online device node is identified as the learning terminal.

[0033] The differential learning data is encapsulated into a front-end interactive component and sent to the locked learning terminal using a message push mechanism;

[0034] An event listener hook is injected into the front-end interaction component to capture the front-end interaction behavior of the target employee for subsequent monitoring.

[0035] Preferably, in step S5, the step of calculating compensation for the current mastery score using a preset asymptotic compensation formula includes:

[0036] The test accuracy of the target employee when completing the differential learning data is obtained and used as the driving coefficient for compensation calculation.

[0037] Calculate the difference between the preset theoretical maximum score and the currently mastered score to determine the repairable upper limit of this round of compensation;

[0038] The repair increment is calculated by multiplying the difference space by the test accuracy, and the repair increment is added to the current mastery score to complete the compensation calculation.

[0039] Secondly, the present invention provides the following technical solution: an intelligent employee file maintenance system based on a personnel system, comprising:

[0040] The log parsing module is used to obtain the operation log data of the target employee by monitoring the preset business system, and use natural language processing technology to parse out the skill trigger triplet from it. The skill trigger triplet includes the employee identifier, skill tag and trigger timestamp.

[0041] The decay calculation module is used to calculate the time interval between the skill tag and the current time based on the trigger timestamp, and to calculate the current mastery score of the target employee for the skill tag by combining the individual forgetting factor and the technology iteration decay factor through a preset two-factor decay model.

[0042] The status determination module is used to compare the current mastery score with a preset status threshold, and update the file status of the skill tag of the target employee in the human resources system according to the comparison result. The file status includes proficient and usable status, skill warning status, and forgotten status.

[0043] The data push module is used to generate corresponding differential learning data based on the obtained version change difference information of the skill tags when the file status is updated to the skill warning status, and push the differential learning data to the learning terminal of the target employee.

[0044] The compensation and repair module is used to monitor the completion status of the learning terminal on the differential learning data. When the completion status is determined to be completed, the module uses a preset asymptotic compensation formula to calculate the compensation for the current mastery score and updates the file status to the proficient and usable status.

[0045] The present invention has the following beneficial effects:

[0046] 1. In this invention, by constructing a dynamic decay model that combines individual forgetting and technological iteration as dual factors, the problem of static and fixed skill files in the personnel system is improved. Based on the frequency of historical operations and the pattern of version updates, the real-time actual ability score of employees is accurately calculated, which effectively avoids the risk of misassignment due to unfamiliarity with skills when transferring across projects and improves the accuracy of person-job matching.

[0047] 2. In this invention, the file state machine and the training system are linked in a deep automation. When the skill score falls into the warning range, the version difference is automatically extracted to generate differential learning data and pushed. After the employee completes the test, the ability value is automatically compensated and repaired using an asymptotic formula. This realizes targeted knowledge compensation for skill decay items and automated closed-loop update of file status.

[0048] 3. In this invention, a data collection mechanism that combines background log monitoring and natural language processing, which is independent of the main business flow, is adopted. Skill trigger triples are automatically parsed from daily business logs as the algorithm calculation benchmark. Under the premise of avoiding manual additional input, it effectively improves the subjective bias and data lag problems caused by the traditional manual evaluation mode, and ensures the objectivity and timeliness of the underlying data source of the intelligent maintenance system. Attached Figure Description

[0049] Figure 1 This is a flowchart of an intelligent employee file maintenance method based on a human resources system proposed in this invention;

[0050] Figure 2 This is a flowchart of the skill trigger triple extraction process based on natural language processing proposed in this invention;

[0051] Figure 3 This is a flowchart of the parallel calculation process for two-factor dynamic decay proposed in this invention;

[0052] Figure 4 This is a flowchart of the feature matching and differential learning data push process proposed in this invention;

[0053] Figure 5 This is a flowchart of the incremental compensation write-back process with cyclic monitoring proposed in this invention.

[0054] Figure 6 This is an architecture diagram of an intelligent employee file maintenance system based on a human resources system proposed in this invention. Detailed Implementation

[0055] The technical solutions in 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.

[0056] Example 1

[0057] In the first embodiment of the present invention, the present invention provides an intelligent maintenance method for employee files based on a personnel system, such as... Figure 1 As shown, it includes:

[0058] S1. Obtain the operation log data of the target employee by monitoring the preset business system, and use natural language processing technology to parse out the skill trigger triplet from it. The skill trigger triplet includes the employee identifier, skill tag and trigger timestamp.

[0059] Further, in step S1, the step of parsing out the skill trigger triples using natural language processing technology includes:

[0060] The operation log data is cleaned by text cleaning, and the cleaned log text is converted into a standardized word vector sequence.

[0061] The word vector sequence is used to extract features using a pre-defined named entity recognition model to identify skill-specific nouns representing technical objects;

[0062] Extract the trigger timestamp and employee identifier corresponding to the operation log data, and structurally bind them with the skill proper noun to construct the skill trigger triplet.

[0063] Specifically, in order to achieve efficient processing and structured feature extraction of unordered logs, the processing flow is as follows: Figure 2 As shown, the system bypasses the data bus to monitor pre-set business systems within the enterprise to obtain the target employee's operation log data. These business systems refer to code hosting platforms or task tracking systems. The operation log data captured in real-time typically includes unstructured text such as code commit comments, system error logs, and project workflow descriptions. After acquiring the data, the system performs text cleaning on the operation log data, using a pre-set filtering script to remove code snippets, garbled characters, and punctuation interference, and then uses a word segmentation tool for word segmentation. The cleaned log text is then input into a pre-trained word embedding layer, transforming discrete natural language words into fixed-dimensional real-valued vectors, thus converting them into a standardized word vector sequence.

[0064] This method utilizes a pre-defined named entity recognition model to extract features from standardized word vector sequences, thereby identifying skill-specific terms representing technical objects. The named entity recognition model employs an architecture combining a bidirectional long short-term memory network and a conditional random field, enabling sequence annotation based on the contextual information of employee business logs. Given a cleaned log text with a corresponding word vector sequence... ,in Represents the first in the operation log Feature vectors of each word This represents the length of the entire log sequence. The model's goal is to predict the corresponding state label sequence. ,in Indicates the first Does each term belong to a technical noun representing a skill in a programming language or software tool? The model calculates the true label sequence. The formula for the conditional probability distribution is:

[0065] ;

[0066] in the formula This indicates that in a given log word vector sequence Generate a specific state label sequence under the condition of The probability, This represents the pre-trained weight parameter matrix in a conditional random field. Represents the input word vector sequence With candidate label sequence The global joint feature mapping function between them This represents a probability normalization summation over all possible label sequence paths. The system uses the Viterbi algorithm to find the optimal label path that maximizes the above conditional probabilities, thereby extracting consecutive words whose labels are classified as target categories as skill-specific nouns.

[0067] After accurately extracting the skill-specific terms, the system uses a data parsing script to directly extract the corresponding system timestamp and account identification code from the metadata header information of the original operation log data. The system queries the enterprise organizational structure database, uniquely mapping the extracted account identification code to a standardized employee identifier, and simultaneously converting the system timestamp into an absolute time value in a standard time zone. Finally, the system structurally binds the extracted employee identifier, the skill tag representing the skill-specific term, and the trigger timestamp in the form of key-value pairs, constructing an immutable skill trigger triplet, and writes it into a distributed graph database for storage.

[0068] This step automatically transforms employees' disordered daily business operation records into highly structured skill interaction credentials, providing an objective and accurate data source guarantee for subsequent skill decay calculations.

[0069] S2. Calculate the time interval between the skill tag and the current time based on the trigger timestamp, and calculate the current mastery score of the target employee for the skill tag by combining the individual forgetting factor and the technology iteration decay factor through a preset two-factor decay model.

[0070] Further, in step S2, the step of combining the individual forgetting factor and the technology iteration decay factor includes:

[0071] Extract the historical trigger frequency characteristics of the target employee for the skill tag, and calculate the individual forgetting factor that characterizes the strength of individual memory retention.

[0072] Obtain the version change statistics of the skill tags in the knowledge graph within a preset time window, and generate the technology iteration decay factor that represents the objective depreciation rate of the technology accordingly.

[0073] The individual forgetting factor and the technology iteration decay factor are normalized to eliminate the dimensional differences between them, thereby forming a standardized feature factor for subsequent calculations.

[0074] Further, in step S2, the step of calculating the target employee's current mastery score for the skill tag includes:

[0075] Retrieve the baseline mastery score of the target employee when the skill tag was last triggered, and use it as the initial base for the calculation period;

[0076] Substitute the time interval and the individual forgetting factor into a preset exponential smoothing equation to calculate the individual forgetting loss value in the time dimension;

[0077] The individual forgetting loss value is subtracted from the initial base, and the result is multiplied by the technology iteration decay factor to obtain the calculated current mastery score.

[0078] Specifically, after obtaining an objective and accurate data source, the system will perform a two-dimensional capability quantification and conversion, with the attenuation calculation process as follows: Figure 3 As shown. The system obtains the current system timestamp and calculates the difference between it and the trigger timestamp in the skill trigger triplet to determine the time interval between the current time and the current time for that skill tag. The system retrieves the total number of historical triggers for that skill tag from the operation log database as the historical trigger frequency feature, and calculates the individual forgetting factor, which characterizes the strength of individual memory retention, by combining it with the memory decay law. The formula for calculating the individual forgetting factor is:

[0079] ;

[0080] in the formula Indicating the individual's forgetting factor, This indicates the historical trigger frequency characteristics of the target employee for this skill tag. This represents the preset memory retention coefficient for different technology stacks. In this solution, the memory retention coefficient... The optimal value is 0.05. The system extracts the number of major version updates that occurred for this skill tag within a preset time window by calling an external technology knowledge graph interface as a version change statistic, and generates a technology iteration decay factor that characterizes the objective depreciation rate of the technology. The formula for calculating the technology iteration decay factor is:

[0081] ;

[0082] in the formula Indicates the technology iteration decay factor. This indicates the version change statistics for this skill tag. This represents the industry technology agility weighting constant. In this solution, the industry technology agility weighting constant... The preferred value is 0.2. The system uses the max-min normalization algorithm to normalize the calculated individual forgetting factor and the technology iteration decay factor, eliminating the dimensional differences caused by different physical meanings, and compressing them into a unified real number space to form standardized feature factors.

[0083] After completing the two-factor calculation, the system retrieves the baseline mastery score stored in the system from the personnel system when the target employee last successfully triggered the skill tag, using it as the initial base for the calculation cycle. The system substitutes the time interval from the current date and the standardized individual forgetting factor into a preset exponential smoothing equation to deduce the individual forgetting loss value generated over time in the individual memory dimension. The formula for calculating the individual forgetting loss value is:

[0084] ;

[0085] in the formula Indicates the individual's forgetting loss value. This represents the baseline score used as the initial base. This indicates the time interval from the current point in time. This represents the individual forgetting factor after standardization. This represents the preset memory half-life time constant. In this scheme, it is assumed that the time interval... The memory half-life time constant is measured in days. The optimal value is 180. The system performs a difference operation to subtract the calculated individual forgetting loss value from the initial base, and then orthogonally multiplies the stage ability value after deducting the loss value with the standardized technology iteration decay factor. Through two consecutive steps of conversion, the system obtains the target employee's real-time current mastery score for the skill tag.

[0086] This step enables the dynamic quantification and conversion of employees' implicit skill levels across the dual physical dimensions of time and technological updates.

[0087] S3. Compare the current mastery score with a preset status threshold, and update the file status of the skill tag of the target employee in the human resources system according to the comparison result. The file status includes proficient and usable status, skill warning status and forgotten status.

[0088] Further, in step S3, the step of comparing the current mastery score with a preset state threshold includes:

[0089] The system retrieves preset high-level security thresholds and low-level early warning thresholds as the state thresholds;

[0090] The current mastery score is compared with the high-level safety threshold and the low-level warning threshold to determine the numerical range of the current mastery score.

[0091] Based on the determination result of the numerical range, the file status of the skill tag is updated to the proficient and usable status, the skill warning status, or the forgotten status.

[0092] Specifically, the system retrieves preset high-level safety thresholds and low-level warning thresholds from the parameter configuration table in the HR system backend and uses them as state thresholds for logical judgment. The system inputs the current mastery score, calculated by the attenuation model, into the judgment engine. A comparator program compares the current mastery score with the high-level safety threshold and the low-level warning threshold for boundary judgment, thereby accurately determining the specific numerical range of the current mastery score within the entire numerical space.

[0093] The system uses a segmented decision algorithm to convert continuous score values ​​into discrete business status labels. The specific logical formula of this status decision algorithm is as follows:

[0094] ;

[0095] in the formula This indicates the final output file status. This indicates the current mastery score of the input judgment engine. This indicates the high-level security threshold being retrieved. This indicates the low-level warning threshold being retrieved. Indicates a proficient and usable state. Indicates a skill warning status. This indicates a forgotten state. In this scheme, it is assumed that the maximum value of the score is 100 points, and the high-level safety threshold is... Preferably, the low-level early warning threshold is 80. The preferred value is 60.

[0096] Based on the calculation results of the above numerical range, the system automatically triggers an internal state machine change event. The system locks and updates the original status identifier field of the target employee for that skill tag in the database by calling the data overwrite interface of the personnel system, forcing the file status to be accurately updated to the target value in the proficient and usable state, skill warning state, or forgotten state.

[0097] This step completes the automatic mapping and conversion of skill levels from continuous decaying scores to discrete management labels.

[0098] S4. When the file status is updated to the skill warning status, generate corresponding differential learning data based on the obtained version change difference information of the skill tag, and push the differential learning data to the learning terminal of the target employee.

[0099] Further, in step S4, the step of generating corresponding differential learning data based on the obtained version change difference information of the skill tags includes:

[0100] Compare the historical usage versions of the skill tags with the current mainstream versions, and extract the difference feature paths as the version iteration difference information;

[0101] Based on the difference feature path, feature retrieval is performed in the preset training material library to retrieve matching multimedia clips;

[0102] The multimedia segments are assembled and spliced ​​with a preset mini-test question bank to generate differential learning data used to compensate for version generation differences.

[0103] Further, in step S4, the step of pushing the differential learning data to the learning terminal of the target employee includes:

[0104] The system activity status of the target employee is detected, and the currently online device node is identified as the learning terminal.

[0105] The differential learning data is encapsulated into a front-end interactive component and sent to the locked learning terminal using a message push mechanism;

[0106] An event listener hook is injected into the front-end interaction component to capture the front-end interaction behavior of the target employee for subsequent monitoring.

[0107] Specifically, when the file status changes, the system will automatically execute precise intervention strategies. The generation and push process at this stage is as follows: Figure 4As shown. When the system's backend service detects a change in the target employee's file status to a skill warning status, it automatically triggers the learning data generation process. The system retrieves the historical version node corresponding to the skill tag in the file record and the current mainstream version node registered in the enterprise resource library. It compares the attribute structures corresponding to the two version nodes and extracts the difference feature paths representing new features or code obsolescence changes, using this as version change difference information. The system uses a text embedding model to convert the difference feature paths into semantic feature vectors and performs vector retrieval operations in a preset training material library to retrieve matching multimedia clips. The system uses a cosine similarity algorithm to calculate the matching degree between the difference feature path vector and the candidate multimedia clip vector. The specific calculation formula is as follows:

[0108] ;

[0109] in the formula This represents the calculated feature matching score. This represents the semantic feature vector generated by the differential feature path transformation. This represents the pre-defined feature vector of candidate multimedia segments in the training material library. The system sorts the multimedia segments in descending order of matching scores and loads the highest-scoring segments from the beginning into memory. The system extracts in-class test questions that are mapped and bound to the above multimedia segments from a pre-set mini-test question bank. Through a data splicing script, the multimedia segments and test questions are structurally assembled and spliced ​​to generate small-volume differential learning data specifically used to compensate for version differences.

[0110] After data generation is complete, the system probes the target employee's current system activity status by calling the heartbeat detection interface of the enterprise's internal identity authentication gateway. The system parses the heartbeat data packets to identify the target employee's current network access type and accurately identifies the online office computer application or mobile office software as the learning terminal to receive push notifications.

[0111] The system uses an interface rendering template to encapsulate the differential learning data into a front-end interactive component containing a media player and answer controls. It then uses a long-connection-based message push mechanism to asynchronously distribute this front-end interactive component to the locked learning terminals via a network protocol. When generating the front-end interactive component, the system injects event listener hook code into its underlying Document Object Model (DOM) tree. This hook code is used to capture in real-time the target employee's front-end interactive behaviors on the learning terminal, such as video playback clicks, screen swipes, and final test answer submissions, and uploads the behavioral payloads to the server for real-time monitoring of subsequent processes.

[0112] This step enables precise knowledge supplementation and automated training interventions tailored to employees' skill weaknesses.

[0113] S5. Monitor the completion status of the learning terminal on the differential learning data. When the completion status is determined to be completed, use a preset asymptotic compensation formula to compensate the current mastery score and update the file status to the proficient and usable status.

[0114] Further, in step S5, the step of calculating compensation for the current mastered score using a preset asymptotic compensation formula includes:

[0115] The test accuracy of the target employee when completing the differential learning data is obtained and used as the driving coefficient for compensation calculation.

[0116] Calculate the difference between the preset theoretical maximum score and the currently mastered score to determine the repairable upper limit of this round of compensation;

[0117] The repair increment is calculated by multiplying the difference space by the test accuracy, and the repair increment is added to the current mastery score to complete the compensation calculation.

[0118] Specifically, after the front-end push is completed, the system will enter the closed-loop verification and value repair phase, and its monitoring and reset process is as follows: Figure 5 As shown, the system continuously monitors the business flow status of the learning terminal's differential learning data by parsing the interactive behavior payload stream returned by the front-end event listening hook in real time through the data receiving interface opened on the server side. When the system parses the control instruction feature code representing the completion of the test submission in the payload stream, it determines that the front-end task is completed. The system then calls the automatic marking engine to perform a Boolean comparison between the answer feature sequence submitted by the target employee and the standard answer sequence associated with the question bank, and calculates the percentage of correct questions out of the total number of questions. The system extracts this percentage as the test accuracy rate, which characterizes the effect of learning knowledge absorption and transformation, and sets it as the driving coefficient for subsequent score compensation calculations.

[0119] The system extracts the preset theoretical maximum score corresponding to the skill tag in the fully mastered state from the system configuration library. It then performs algebraic subtraction to subtract the currently retained mastery score from the theoretical maximum score, calculating the difference. This difference precisely defines the absolute repairable upper limit achievable in this round of skill awakening. The system uses a preset asymptotic compensation formula to fuse the driving coefficient with this repairable upper limit, obtaining the final repair increment and resetting the score. The specific asymptotic compensation calculation formula is as follows:

[0120] ;

[0121] in the formula This indicates the skill mastery score reset after compensation calculation. This represents the current mastery score recorded in the underlying database before the compensation calculation is performed. This represents the test accuracy rate calculated by the system after the target employees complete the test. This represents the preset theoretical maximum score. The overall representation is the calculated difference space. In this scheme, the preset theoretical maximum score... The preferred value is 100.

[0122] After completing the compensation and superposition calculation of the score, the system overwrites the newly generated skill mastery score as persistent data into the business database, replacing the original decay record. Based on the rebound signal of this score, the system triggers the reverse reset logic of the underlying state machine. By calling the data update bus of the human resources system, it forcibly updates the file status of the target employee for this skill tag to the proficient and usable state, removes the original warning restriction, and thus completes the closed-loop link of the entire intelligent maintenance of skill files.

[0123] This step enables quantitative compensation of employee skill levels in the system's underlying data and automatic restoration of their record status.

[0124] Example 2

[0125] Existing HR systems generally employ a one-time manual data entry or a static storage model relying on certificate validity periods. Once skill tags are acquired, they are fixed. This model fails to objectively reflect an employee's true usability under the combined effects of time and technological advancements, leading to a disconnect between the proficiency status presented in the personnel file and actual capabilities. This can easily result in the misassignment of employees with insufficient actual competence during project deployment, thereby increasing the risk of project delays. To address these issues, this invention provides an intelligent employee file maintenance system based on an HR system, the structure of which is as follows: Figure 6 As shown. The specific implementation process of this system is as follows:

[0126] The log parsing module is used to obtain the operation log data of the target employee by monitoring the preset business system, and use natural language processing technology to parse out the skill trigger triplet from it. The skill trigger triplet includes the employee identifier, skill tag and trigger timestamp.

[0127] The decay calculation module is used to calculate the time interval between the skill tag and the current time based on the trigger timestamp, and to calculate the current mastery score of the target employee for the skill tag by combining the individual forgetting factor and the technology iteration decay factor through a preset two-factor decay model.

[0128] The status determination module is used to compare the current mastery score with a preset status threshold, and update the file status of the skill tag of the target employee in the human resources system according to the comparison result. The file status includes proficient and usable status, skill warning status, and forgotten status.

[0129] The data push module is used to generate corresponding differential learning data based on the obtained version change difference information of the skill tags when the file status is updated to the skill warning status, and push the differential learning data to the learning terminal of the target employee.

[0130] The compensation and repair module is used to monitor the completion status of the learning terminal on the differential learning data. When the completion status is determined to be completed, the module uses a preset asymptotic compensation formula to calculate the compensation for the current mastery score and updates the file status to the proficient and usable status.

[0131] Specifically, the log parsing module first obtains unstructured operation logs by bypassing business systems such as code hosting platforms. After text cleaning and word vectorization, it extracts skill-specific terms representing technical objects using a named entity recognition model based on a bidirectional long short-term memory network and conditional random field architecture. It then extracts system timestamps and parsed account identifiers for structured binding, generating immutable skill trigger triples. Subsequently, the decay calculation module calculates the time interval of the skill tag, generating individual forgetting factors and technology iteration decay factors based on the target employee's historical trigger frequency and the overall version change statistics in the knowledge graph. After normalization, it deducts the time forgetting loss value derived from the exponential smoothing equation from the employee's initial baseline mastery score, and then orthogonally multiplies the result with the technology iteration decay factor to obtain the current mastery score. Next, the status determination module inputs this mastery score into a comparator engine, comparing it against preset high-level safety thresholds and low-level warning thresholds. Based on the determined numerical range, it calls the underlying data overwrite interface. The continuous scores are discretely mapped and updated to the corresponding proficient, skill warning, or forgotten states. Once the file status falls into the skill warning state, the data push module responds immediately. By comparing historical data with mainstream versions to extract the difference feature paths and performing vector retrieval in the training material library, highly matched multimedia segments and mini-test question banks are spliced ​​together to assemble differential learning data. At the same time, the activity status is checked to lock the employee's current online terminal, and the front-end interactive components injected with event listening hooks are asynchronously pushed. Finally, the compensation and repair module analyzes the interactive behavior payload flow returned by the front-end hooks in real time. After the test is completed, the automatic grading engine is started to obtain the test accuracy as the driving coefficient. The difference space between the theoretical highest score of the system and the current score is calculated. The driving coefficient and the difference space are fused using the asymptotic compensation formula to calculate the repair increment and complete the score reset. Then, based on the score recovery signal, the reverse reset logic of the state machine is triggered to force the employee's file status to be updated back to the proficient and usable state, thereby completing the dynamic compensation of the employee's implicit skill scores and the closed-loop maintenance of the file data.

[0132] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent maintenance of employee files based on a personnel system, characterized in that, include: S1. Obtain the operation log data of the target employee by monitoring the preset business system, and use natural language processing technology to parse out the skill trigger triplet from it. The skill trigger triplet includes the employee identifier, skill tag and trigger timestamp. S2. Calculate the time interval between the skill tag and the current time based on the trigger timestamp, and calculate the current mastery score of the target employee for the skill tag by combining the individual forgetting factor and the technology iteration decay factor through a preset two-factor decay model. S3. Compare the current mastery score with a preset status threshold, and update the file status of the skill tag of the target employee in the human resources system according to the comparison result. The file status includes proficient and usable status, skill warning status and forgotten status. S4. When the file status is updated to the skill warning status, generate corresponding differential learning data based on the obtained version change difference information of the skill tag, and push the differential learning data to the learning terminal of the target employee. S5. Monitor the completion status of the learning terminal on the differential learning data. When the completion status is determined to be completed, use a preset asymptotic compensation formula to compensate the current mastery score and update the file status to the proficient and usable status.

2. The intelligent employee file maintenance method based on a personnel system according to claim 1, characterized in that, In step S1, the step of parsing the skill trigger triples using natural language processing technology includes: The operation log data is cleaned by text cleaning, and the cleaned log text is converted into a standardized word vector sequence. The word vector sequence is used to extract features using a pre-defined named entity recognition model to identify skill-specific nouns representing technical objects; Extract the trigger timestamp and employee identifier corresponding to the operation log data, and structurally bind them with the skill proper noun to construct the skill trigger triplet.

3. The intelligent employee file maintenance method based on a personnel system according to claim 1, characterized in that, In step S2, the step of combining the individual forgetting factor and the technology iteration decay factor includes: Extract the historical trigger frequency characteristics of the target employee for the skill tag, and calculate the individual forgetting factor that characterizes the strength of individual memory retention. Obtain the version change statistics of the skill tags in the knowledge graph within a preset time window, and generate the technology iteration decay factor that represents the objective depreciation rate of the technology accordingly. The individual forgetting factor and the technology iteration decay factor are normalized to eliminate the dimensional differences between them, thereby forming a standardized feature factor for subsequent calculations.

4. The intelligent employee file maintenance method based on a personnel system according to claim 1, characterized in that, In step S2, the step of calculating the target employee's current mastery score for the skill tag includes: Retrieve the baseline mastery score of the target employee when the skill tag was last triggered, and use it as the initial base for the calculation period; Substitute the time interval and the individual forgetting factor into a preset exponential smoothing equation to calculate the individual forgetting loss value in the time dimension; The individual forgetting loss value is subtracted from the initial base, and the result is multiplied by the technology iteration decay factor to obtain the calculated current mastery score.

5. The intelligent employee file maintenance method based on a personnel system according to claim 1, characterized in that, Step S3, the step of comparing the current mastery score with a preset state threshold, includes: The system retrieves preset high-level security thresholds and low-level early warning thresholds as the state thresholds; The current mastery score is compared with the high-level safety threshold and the low-level warning threshold to determine the numerical range of the current mastery score. Based on the determination result of the numerical range, the file status of the skill tag is updated to the proficient and usable status, the skill warning status, or the forgotten status.

6. The intelligent employee file maintenance method based on a personnel system according to claim 1, characterized in that, Step S4, the step of generating corresponding differential learning data based on the obtained version change difference information of the skill tags, includes: Compare the historical usage versions of the skill tags with the current mainstream versions, and extract the difference feature paths as the version iteration difference information; Based on the difference feature path, feature retrieval is performed in the preset training material library to retrieve matching multimedia clips; The multimedia segments are assembled and spliced ​​with a preset mini-test question bank to generate differential learning data used to compensate for version generation differences.

7. The intelligent employee file maintenance method based on a personnel system according to claim 1, characterized in that, Step S4, the step of pushing the differential learning data to the learning terminal of the target employee, includes: The system activity status of the target employee is detected, and the currently online device node is identified as the learning terminal. The differential learning data is encapsulated into a front-end interactive component and sent to the locked learning terminal using a message push mechanism; An event listener hook is injected into the front-end interaction component to capture the front-end interaction behavior of the target employee for subsequent monitoring.

8. The intelligent maintenance method for employee files based on a personnel system according to claim 1, characterized in that, Step S5, the step of calculating compensation for the current mastered score using a preset asymptotic compensation formula, includes: The test accuracy of the target employee when completing the differential learning data is obtained and used as the driving coefficient for compensation calculation. Calculate the difference between the preset theoretical maximum score and the currently mastered score to determine the repairable upper limit of this round of compensation; The repair increment is calculated by multiplying the difference space by the test accuracy, and the repair increment is added to the current mastery score to complete the compensation calculation.

9. An intelligent employee file maintenance system based on a personnel system, characterized in that, A method for intelligent maintenance of employee files based on a personnel system, as described in any one of claims 1-8, wherein the system comprises: The log parsing module is used to obtain the operation log data of the target employee by monitoring the preset business system, and use natural language processing technology to parse out the skill trigger triplet from it. The skill trigger triplet includes the employee identifier, skill tag and trigger timestamp. The decay calculation module is used to calculate the time interval between the skill tag and the current time based on the trigger timestamp, and to calculate the current mastery score of the target employee for the skill tag by combining the individual forgetting factor and the technology iteration decay factor through a preset two-factor decay model. The status determination module is used to compare the current mastery score with a preset status threshold, and update the file status of the skill tag of the target employee in the human resources system according to the comparison result. The file status includes proficient and usable status, skill warning status, and forgotten status. The data push module is used to generate corresponding differential learning data based on the obtained version change difference information of the skill tags when the file status is updated to the skill warning status, and push the differential learning data to the learning terminal of the target employee. The compensation and repair module is used to monitor the completion status of the learning terminal on the differential learning data. When the completion status is determined to be completed, the module uses a preset asymptotic compensation formula to calculate the compensation for the current mastery score and updates the file status to the proficient and usable status.

Images