Matching threshold value adjusting method and system for post portrait dynamic updating

Abstract

The invention discloses a matching threshold adjustment method and system oriented to post portrait dynamic update, particularly relates to the technical field of intelligent recruitment, and is used for solving the problem of systematic offset caused by the fact that low-quality candidate data pollutes a model due to dynamic threshold adjustment of an existing recruitment platform. Candidate skill test scores and performance data corresponding to target post skill labels are obtained; calculating the KL divergence value of the skill test score distribution and the post reference curve; triggering a PID control algorithm based on a KL divergence threshold to generate a threshold correction coefficient; dynamically updating the matching threshold and generating a threshold range; adjusting a sampling weight strategy according to the deviation direction; and the model is retrained to output a recommendation list, collaborative optimization of recruitment efficiency and model stability is realized, and interference of data quality degradation on an algorithm benchmark is effectively inhibited.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent recruitment. More specifically, the present invention relates to a method and system for adjusting matching thresholds for dynamic updates of job portraits. Background Art

[0002] In the prior art, recruitment platforms respond to market supply and demand fluctuations by dynamically adjusting the matching thresholds between job requirements and candidates. When the skill requirements for a specific job surge or the talent supply is insufficient, the system usually expands the candidate pool by lowering the matching criteria to ensure the timely filling of job vacancies. It mainly relies on explicit indicators such as the number of resumes collected in real time and the distribution of skill tags for threshold adjustment, aiming to balance the recruitment efficiency of enterprises and the accuracy of talent screening.

[0003] However, dynamic threshold adjustment can lead to a systematic shift in the distribution of candidate capabilities. When the system relaxes the requirements for a specific skill dimension due to insufficient supply, a large number of low-quality candidates enter the recommendation list and are incorporated into the subsequent model training data, resulting in a continuous shift in the decision benchmark for the true ability standard by the algorithm. This shift will further exacerbate the error of threshold adjustment, forming a vicious cycle of low-quality data accumulation and degradation of matching criteria, ultimately damaging the long-term reliability of the recommendation system. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method and system for adjusting matching thresholds for dynamic updates of job portraits to solve the problems raised in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solutions: A method for adjusting matching thresholds for dynamic updates of job portraits, comprising the following steps: S1. Obtain the candidate skill test scores corresponding to the current skill tags of the target job and the performance data after employment; S2. Calculate the KL divergence value between the distribution curve of the candidate skill test scores within a preset period and the benchmark curve of the skill requirements of the target job; S3. When the KL divergence value exceeds the KL divergence threshold, generate a matching threshold correction coefficient based on the PID control algorithm; S4. Update the matching threshold of the current job skill tags according to the matching threshold correction coefficient to generate an updated threshold range; S5. Determine the sampling weight adjustment strategy for the candidate skill test scores based on the deviation direction between the KL divergence value and the updated threshold range; S6. Retrain the matching model according to the skill test score sampling weight adjustment strategy and output an updated candidate recommendation list.

[0006] In a preferred embodiment, the proportional coefficient of the PID control algorithm is dynamically adjusted according to the correlation coefficient between the candidate's skill test score and the performance data.

[0007] In a preferred embodiment, obtaining the candidate's skill test scores corresponding to the current skill tags of the target position and the performance data after employment, including: Obtaining the recruitment description text of the target position, and parsing the recruitment description text to generate the current skill tags of the target position; Obtaining the skill test scores of the candidate corresponding to the current skill tags of the target position from the online assessment system; Obtaining the performance data after employment of the candidate in the position corresponding to the current skill tags of the target position from the human resource management system.

[0008] In a preferred embodiment, calculating the KL divergence value between the distribution curve of the candidate's skill test scores within a preset period and the benchmark curve of the skill requirements of the target position, including: Performing binning processing on the candidate's skill test scores within a preset period according to the skill tag dimension to generate a discretized skill test score distribution; Dynamically adjusting the bin width based on the distribution form of the benchmark curve of the skill requirements of the target position to align the data interval of the discretized skill test score distribution with the benchmark curve; Recalculating the discretized skill test score distribution according to the adjusted bin width, and extracting the kurtosis parameter and skewness parameter of the benchmark curve; Based on the kurtosis parameter and skewness parameter, correcting the distribution function of the benchmark curve to generate a dynamically adapted benchmark curve of the skill requirements of the target position; Calculating the KL divergence value between the discretized skill test score distribution and the dynamically adapted benchmark curve of the skill requirements of the target position.

[0009] In a preferred embodiment, when the KL divergence value exceeds the KL divergence threshold, generating a matching threshold correction coefficient based on the PID control algorithm, including: Obtaining the correlation coefficient between the candidate's skill test score and the performance data, and the correlation coefficient is calculated by the Pearson product-moment correlation coefficient formula; Dynamically adjusting the proportional coefficient of the PID control algorithm based on the deviation degree between the correlation coefficient and the preset proportional reference value; Calculating the integral term according to the difference between the KL divergence value and the KL divergence threshold, and calculating the differential term based on the change rate of the KL divergence value; Superimposing the proportional coefficient, the integral term and the differential term according to the preset weights to generate a matching threshold correction coefficient.

[0010] In a preferred embodiment, updating the matching threshold of the current position skill tags according to the matching threshold correction coefficient to generate an updated threshold range, including: Obtain the original matching threshold of the current job skill label and the matching threshold correction coefficient generated in step S3; Linearly adjust the original matching threshold according to the matching threshold correction coefficient to generate a preliminary adjusted threshold; When the preliminary adjusted threshold exceeds the preset threshold range, perform truncation processing according to the boundary value of the preset threshold range to generate a compliant adjusted threshold; Based on the priority weights of the compliant adjusted threshold and the target job skill label, generate an updated threshold range; Write the updated threshold range into the job portrait database and trigger the recording of the threshold effective timestamp.

[0011] In a preferred embodiment, based on the deviation direction of the KL divergence value from the updated threshold range, determine the sampling weight adjustment strategy for the candidate skill test scores, including: Determine the deviation direction of the KL divergence value from the updated threshold range, and the deviation direction is divided into positive deviation from the threshold upper limit and negative deviation from the threshold lower limit; When the deviation direction is positive deviation from the threshold upper limit, reduce the sampling weight of the candidate skill test scores higher than the current threshold upper limit; When the deviation direction is negative deviation from the threshold lower limit, increase the sampling weight of the candidate skill test scores lower than the current threshold lower limit; Generate a dynamic sampling weight adjustment strategy according to the deviation direction type and deviation amplitude.

[0012] In a preferred embodiment, the sampling weight adjustment strategy includes a piecewise function mapping rule of a weight decay factor and a gain factor.

[0013] In a preferred embodiment, retrain the matching model according to the skill test score sampling weight adjustment strategy and output an updated candidate recommendation list, including: Extract the weight decay factor and gain factor mapping table from the sampling weight adjustment strategy; Based on the weight decay factor and gain factor, adjust the sampling weight distribution of the candidate skill test scores to generate a weighted training dataset; Retrain the matching model using the weighted training dataset, and the matching model updates the parameters by minimizing the prediction error of the weighted samples; Filter the output result of the matching model according to the updated threshold range, generate an updated candidate recommendation list and append the threshold effective timestamp.

[0014] On the other hand, the present invention provides a matching threshold adjustment system for dynamic update of job portraits, including the following modules: Data acquisition module: used to obtain the candidate skill test scores corresponding to the current skill tags of the target position and the performance data after employment; Divergence calculation module: used to calculate the KL divergence value between the distribution curve of the candidate skill test scores within a preset period and the benchmark curve of the skill requirements of the target position; Dynamic adjustment module: when the KL divergence value exceeds the KL divergence threshold, used to generate a matching threshold correction coefficient based on the PID control algorithm; Threshold update module: used to update the matching threshold of the current position skill tags according to the matching threshold correction coefficient and generate an updated threshold range; Strategy generation module: used to determine the sampling weight adjustment strategy of the candidate skill test scores based on the deviation direction between the KL divergence value and the updated threshold range; Model optimization module: used to retrain the matching model according to the sampling weight adjustment strategy of the skill test scores and output an updated candidate recommendation list.

[0015] Compared with the prior art, the present invention has the following beneficial effects:

[0016] 1. Through a closed-loop feedback mechanism, the dynamic optimization of the matching threshold and the two-way control of data quality are realized, effectively solving the model deviation problem caused by threshold adjustment in traditional recruitment systems; by introducing the KL divergence value to quantify the deviation degree between the candidate ability distribution and the position benchmark in real time, and combining the PID control algorithm to dynamically correct the threshold parameters, the matching standard can be adaptively adjusted according to the data quality feedback, avoiding the error accumulation caused by the fluctuation of a single index; embedding the correlation between skill-performance data into the proportional coefficient adjustment logic makes the threshold correction amplitude strongly correlated with the actual performance of candidates, ensuring that the threshold adjustment not only responds to market supply and demand changes but also takes into account the accuracy of talent screening; this closed-loop control mechanism based on mathematical metrics significantly improves the rationality of threshold adjustment and the anti-interference ability of the system.

[0017] 2. Through the asymmetric weight adjustment strategy and the time sequence synchronization mechanism, the negative impact of low-quality samples on the model is suppressed from the data source; the sampling weights of high / low-score candidates are differentially adjusted according to the KL divergence deviation direction, actively blocking the data deterioration path; at the same time, by strictly binding the threshold effective timestamp and the model version, the operation time sequence consistency of threshold update, model training, and recommendation result output is ensured; the synergistic effect of each technical feature maintains the stable determination of the algorithm for the real ability standard while ensuring the recruitment efficiency, ultimately realizing the dual improvement of talent matching accuracy and the long-term reliability of the system, and forming a technical closed-loop of dynamic balance between data quality and matching standards. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flowchart of the matching threshold adjustment method for dynamic update of the position portrait of the present invention; Figure 2 This is a schematic structural diagram of a matching threshold adjustment system for dynamic update of job portraits according to the present invention. Specific implementation mode

[0019] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0020] Embodiment 1: Figure 1 A matching threshold adjustment method for dynamic update of job portraits according to the present invention is given, which includes the following steps: S1. Obtain the candidate skill test scores corresponding to the current skill tags of the target job and the performance data after employment; S2. Calculate the KL divergence value between the distribution curve of the candidate skill test scores within a preset period and the benchmark curve of the skill requirements of the target job; S3. When the KL divergence value exceeds the KL divergence threshold, generate a matching threshold correction coefficient based on the PID control algorithm, and the proportional coefficient of the PID control algorithm is dynamically adjusted according to the correlation coefficient between the candidate skill test scores and the performance data; S4. Update the matching threshold of the current job skill tags according to the matching threshold correction coefficient to generate an updated threshold range; S5. Determine the sampling weight adjustment strategy for the candidate skill test scores based on the deviation direction between the KL divergence value and the updated threshold range; S6. Retrain the matching model according to the skill test score sampling weight adjustment strategy and output an updated candidate recommendation list.

[0021] S1. Obtain the candidate skill test scores corresponding to the current skill tags of the target job and the performance data after employment. The specific implementation is as follows: The specific implementation manner of obtaining the candidate skill test scores corresponding to the current skill tags of the target job and the performance data after employment is achieved through the following operations: The original recruitment information document of the target position is retrieved from the enterprise human resources management platform. The recruitment information document contains text content of job description, job qualification requirements, and technical ability description written in natural language. Perform structured analysis on the text content. The specific processing process includes the following steps: identify the start paragraph identifier and the end paragraph identifier of the "Job Requirements" chapter in the text, and extract the complete sentence set in the chapter; perform word segmentation on the sentence set, and use a dictionary-based word segmentation algorithm to split the sentence into independent vocabulary sequences; filter stop words in the vocabulary sequence, and retain nouns and verb-object phrases containing technical ability descriptions as candidate keyword sets; semantically match the candidate keyword set with the predefined standardized skill vocabulary. Each skill label in the skill vocabulary is associated with at least three synonym expressions. When "proficient in Java language development" appears in the original text, it is normalized to the "Java skill" label by matching the "Java development" label in the vocabulary, and the current skill label set of the target position is generated. The skill vocabulary is constructed by integrating the industry standard skill classification framework, which contains the common technology stack labels and corresponding synonym mapping rules in the Internet industry.

[0022] The specific implementation process of obtaining the skill test scores corresponding to the current skill tags of the candidate and the target position from the online assessment system is as follows: according to the generated current skill tag set of the target position, a skill assessment request is sent to the assessment system interface, and the request contains the skill tag name and the corresponding assessment level parameter. The online assessment system selects theoretical questions and practical tasks from the preset question bank according to the skill tag name, such as calling multi-threaded programming questions and JVM tuning scenario tasks for the "Java skill" tag. The difficulty level of the theoretical questions matches the assessment level parameters, and the design standards of the practical tasks are set in reference to the actual engineering requirements of the target position. After the candidate completes the assessment, the online assessment system calculates the skill test scores under each skill tag according to the preset scoring rules. The scoring rules include the dynamic proportional allocation of the theoretical question accuracy weight, the code execution efficiency weight, and the scenario task completion weight. The score results are stored in the intermediate database in the structured format of "candidate ID-skill tag-skill test score", and the database index mechanism is used to ensure the data relevance with the subsequent steps.

[0023] The specific operation steps for obtaining post - employment performance data from the human resource management system are as follows: Based on the current skill tag set of the target position, perform a joint query in the employee file database of the human resource management system. The query conditions include that the employee's current on - the - job position name exactly matches the target position name, the employee's skill file record contains the current target skill tag, and the employee's employment time exceeds the preset assessment cycle threshold. The performance data includes quantitative indicators automatically collected by the system and qualitative scores manually evaluated by the direct supervisor. The quantitative indicators are the code submission defect rate, the project delivery on - time rate, and the average task completion time. The qualitative scores are the innovative score of the technical solution and the teamwork ability score. When extracting data, associate the employee identity identifier with the skill tag to ensure that each performance data entry corresponds to the assessment result under a specific skill tag. The data update mechanism is synchronized with the enterprise's performance assessment cycle. After each performance assessment is completed, trigger the data synchronization interface and write the latest performance data into the skill - performance association database in the format of "employee ID - skill tag - performance data".

[0024] The implementation method for the associated storage of skill test scores and performance data is as follows: Establish a skill - performance mapping table in the relational database. The mapping table contains fields for the candidate's unique identifier, skill tag, skill test score, and performance data. The associated storage process is completed through the following steps: Combine the skill test score file exported from the online assessment system into a composite primary key based on the candidate ID and skill tag; Match the performance data file exported from the human resource management system with the same composite primary key; For the successfully matched records, merge the skill test scores and performance data and write them into the mapping table; For the unmatched records, perform exception logging and trigger the manual review process. Set a foreign key constraint for the skill tag field in the database table structure. The foreign key constraint points to the current skill tag set of the target position to ensure the consistency of the skill tags. When a skill tag mismatch or data loss is detected during the writing process, trigger the exception handling process and generate a log record. The administrator manually reviews the data source and re - executes the associated storage operation.

[0025] S2. Calculate the KL divergence value between the distribution curve of the candidate's skill test scores within the preset period and the benchmark curve of the target position's skill requirements. The specific implementation is as follows: The specific process of binning the candidate skill test scores by skill tag dimension within a preset period to generate a discretized skill test score distribution is as follows: Extract the candidate skill test score data corresponding to the current skill tags of the target position from the associated storage database in step S1, and classify and organize it into independent data sets according to the skill tag names. Perform data preprocessing on the skill test scores under each skill tag. The preprocessing includes removing data items outside the preset valid range and filling in missing values. The preset valid range is set according to the historical data distribution characteristics of the skill tag. For example, the upper limit of the skill test score is 1.2 times the theoretical maximum score, and the lower limit is 0.8 times the theoretical minimum score. After completing the preprocessing, use the equal-width binning method to discretize the skill test scores. The initial bin boundaries are dynamically set according to the historical score distribution characteristics of the skill tag. When the skill test scores are normally distributed, uniform binning is used. When the score distribution has obvious skewness, the number of bins is preferentially increased in the dense interval. After binning, count the proportion of the number of candidates in each interval to generate a discretized skill test score distribution table containing bin intervals, frequencies, and ratios. The number of bin intervals does not exceed the preset upper limit value, and the upper limit value is set according to the scoring granularity requirements of the skill tag. For example, when the theoretical maximum score is 100, the upper limit value is 10 bins.

[0026] The specific process of dynamically adjusting the bin width based on the distribution form of the target position skill requirement benchmark curve to align the discretized skill test score distribution with the data interval of the benchmark curve is as follows: Obtain the initial discretized skill test score distribution table generated in step S2 and the preset target position skill requirement benchmark curve. Analyze the distribution form characteristics of the benchmark curve. When the benchmark curve shows a high-density distribution in a specific score interval, reduce the bin width of the corresponding interval to 50% of the original width to increase the number of bins; when the benchmark curve shows a low-density distribution in a specific interval, expand the bin width of the corresponding interval to 150% of the original width to reduce the number of bins. The adjusted bin boundaries are strictly aligned with the data density distribution trend of the benchmark curve to ensure that the numerical range of each bin interval completely matches the key data segments of the benchmark curve. The number of bins is retained not to exceed the preset upper limit value during the dynamic adjustment process, and the upper limit value is set according to the scoring granularity requirements of the skill tag. For example, when the theoretical maximum score is 100, the upper limit value is 10 bins. The adjusted bin interval boundaries are ensured to be synchronized with the benchmark curve through the database transaction lock mechanism to avoid bin misalignment caused by data competition.

[0027] The specific process of recalculating the discretized skill test score distribution according to the adjusted bin width and extracting the kurtosis parameter and skewness parameter of the benchmark curve is as follows: Re-divide the skill test score intervals according to the dynamically adjusted bin width, count the proportion of the number of candidates in each new interval, and generate an updated discretized skill test score distribution table. Extract the statistical distribution parameters of the benchmark curve for the skill requirements of the target position. The kurtosis parameter is used to quantify the sharpness of the distribution shape, and its calculation method is based on the numerical result of the standardized fourth-order central moment; the skewness parameter is used to quantify the symmetry of the distribution shape, and its calculation method is based on the numerical result of the standardized third-order central moment. The calculation process of the kurtosis parameter and the skewness parameter performs local feature extraction by traversing the data points of the benchmark curve through the sliding window method. The window width is dynamically adjusted according to the total amount of data of the benchmark curve, and the window width setting rule is to round the square root of the total amount of data. The calculation results are stored in the feature parameter database classified by skill tags for subsequent steps to call.

[0028] The specific process of generating the benchmark curve for the skill requirements of the target position after dynamic adaptation by correcting the distribution function based on the kurtosis parameter and the skewness parameter is as follows: Compare the kurtosis parameter of the benchmark curve with the preset kurtosis threshold for the skill standard of the position. When the kurtosis parameter exceeds the threshold, perform a kurtosis attenuation operation on the distribution function of the benchmark curve. The kurtosis attenuation operation is achieved by reducing the probability density in the peak region of the distribution function; Compare the skewness parameter of the benchmark curve with the preset skewness threshold for the skill standard of the position. When the skewness parameter shows left or right skewness beyond the threshold range, perform a skewness correction operation on the distribution function. The correction operation is achieved by translating the center point position of the distribution function or adjusting the slope of the probability density function. The corrected distribution function is re-fitted to the skill test score distribution of historical successful candidates to generate the benchmark curve for the skill requirements of the target position after dynamic adaptation. The morphological characteristics of this curve reflect both the theoretical distribution characteristics of the skill requirements of the position and the dynamic change trend of the actual distribution of current candidates. The corrected benchmark curve records the historical change trajectory through the version control mechanism to support data backtracking and error analysis.

[0029] The specific process of calculating the KL divergence value between the discretized skill test score distribution and the benchmark curve for the skill requirements of the target position after dynamic adaptation is as follows: Convert the frequency proportion data in the updated discretized skill test score distribution table into a discrete probability distribution. The conversion rule is to divide the frequency proportion of each bin interval by the interval width to eliminate the bin scale difference.

[0030] Discretize the target job skill requirement benchmark curve after dynamic adaptation into a benchmark probability distribution with the same bin width, ensuring that the numerical range of the bin intervals is exactly the same as the skill test score distribution. Calculate the logarithm of the ratio of the candidate distribution probability to the benchmark curve probability for each bin, and sum the calculation results of each bin weighted by the candidate distribution probability to generate the final KL divergence value. Ignore the bin intervals with zero probability during the calculation process to avoid mathematical calculation failures. Store the calculation results of the KL divergence value in the analysis database classified by skill tags for subsequent steps to call. Generate a distribution comparison graph through a visualization tool to assist in manually reviewing the data rationality.

[0031] S3. When the KL divergence value exceeds the KL divergence threshold, generate a matching threshold correction coefficient based on the PID control algorithm. The proportional coefficient of the PID control algorithm is dynamically adjusted according to the correlation coefficient between the candidate skill test score and the performance data. The specific implementation is as follows: Obtain the correlation coefficient between the candidate skill test score and the performance data. The specific process of calculating the correlation coefficient through the Pearson product-moment correlation coefficient formula is as follows: Extract the candidate skill test scores corresponding to the current skill tags of the target job and the performance data after employment from the associated storage database in step S1, and classify and organize them into independent data sets according to the skill tag names. Perform data alignment operations on the skill test scores and performance data under each skill tag. The data alignment operations include removing the invalid candidate records that do not have both skill test scores and performance data, and sorting the skill test scores and performance data of the remaining records by timestamp to ensure the time series consistency of the data.

[0032] Calculate the covariance between the skill test score and the performance data. The calculation method of the covariance is the mean of the products of the corresponding items of the two sets of data minus the product of the means of the two sets of data; calculate the standard deviations of the skill test score and the performance data respectively. The calculation method of the standard deviation is the square root of the mean of the squares of the differences between the data items and the mean; divide the covariance by the product of the standard deviation of the skill test score and the standard deviation of the performance data to generate the standardized Pearson product-moment correlation coefficient. Store the calculation results in the analysis database classified by skill tags for subsequent steps to call.

[0033] The specific process of dynamically adjusting the proportional coefficient of the PID control algorithm based on the deviation degree between the correlation coefficient and the preset proportional reference value is as follows: Obtain the correlation coefficient calculated in step S3 and the preset proportional reference value. The proportional reference value is set according to the long-term correlation level between the skill test score and the performance data in the historical successful candidate data of the target job. The specific setting rule is to select the moving average of the correlation coefficients in the historical data as the reference value.

[0034] When the correlation coefficient is greater than the proportional reference value, the proportional coefficient is incremented according to a preset linear gain rule. The increment is the difference between the correlation coefficient and the reference value multiplied by a preset gain factor, and the gain factor is set according to the adjustment sensitivity requirement of the target position. When the correlation coefficient is less than the reference value, the proportional coefficient is decremented according to a preset exponential decay rule. The decrement is the difference between the reference value and the correlation coefficient multiplied by a decay factor. The dynamic adjustment range of the proportional coefficient is limited within a preset interval. The lower limit of the interval is set according to the minimum skill requirement of the position, and the upper limit of the interval is set according to the maximum skill tolerance of the position. During the adjustment process, it is detected in real time whether the proportional coefficient exceeds the interval range and boundary truncation processing is performed.

[0035] The specific process of calculating the integral term based on the difference between the KL divergence value and the KL divergence threshold and calculating the differential term based on the change rate of the KL divergence value is as follows: Obtain the KL divergence value calculated in step S2 and the preset KL divergence threshold, and calculate the difference between the current KL divergence value and the threshold as the input quantity of the integral term. The calculation method of the integral term is to accumulate the weighted sum of the historical difference sequence. The weight assignment of the weighted sum adopts a time decay strategy. The weight of the difference in the most recent calculation period is 1, and the weight of the previous period is the product of the decay factor and the reciprocal of the time interval. The decay factor is set according to the skill stability requirement of the position. The differential term is calculated based on the change rate of the KL divergence value within a preset time window. The width of the time window is set according to the adjustment response speed of the target position. The change rate calculation method is the difference between the current KL divergence value and the KL divergence value at the start time of the window divided by the length of the time interval. The length of the time interval is accurate to one decimal place in days. The calculation results of the integral term and the differential term are stored in the control parameter database classified by skill tags.

[0036] The specific process of generating a matching threshold correction coefficient by superimposing the proportional coefficient, the integral term, and the differential term according to preset weights is as follows: Extract the proportional coefficient, the integral term, and the differential term of the current skill tag from the control parameter database, and perform linear superposition according to the preset weight distribution ratio. The preset weights are set according to the skill dimension priority of the target position. The weight distribution of the core skill tag favors the proportional term and the differential term, and the weight distribution of the auxiliary skill tag favors the integral term and the differential term.

[0037] Before the superposition calculation, the integral term and the differential term are normalized. The integral term is mapped to the interval from 0 to 1, and the differential term is mapped to the interval from -1 to 1. The result of the linear superposition is smoothed by the Sigmoid function to generate a matching threshold correction coefficient. The value range of the correction coefficient is limited between 0.5 and 2.0. When it exceeds the range, it is truncated according to the boundary value. The correction coefficient is stored in the output database classified by skill tags for step S4 to call and write to the log file for subsequent audit tracking.

[0038] S4. Update the matching threshold of the current job skill label according to the matching threshold correction coefficient to generate an updated threshold range. The specific implementation is as follows: The specific process of obtaining the original matching threshold of the current job skill label and the matching threshold correction coefficient generated in step S3 is as follows: Read the original matching threshold of the current skill label of the target job from the matching threshold configuration table in the job portrait database. The original matching threshold is set as follows: According to the skill test score distribution of successful candidates in the historical recruitment data of the target job, select the 70th percentile value as the initial threshold benchmark.

[0039] For example, when the skill test scores of historical successful candidates under a certain skill label are arranged in ascending order, the score corresponding to the 70th percentile is 78 points, then the original matching threshold is set to 78 points. The percentile calculation method uses linear interpolation to ensure the continuity of the score distribution. Extract the matching threshold correction coefficient that exactly matches the current skill label name from the output database in step S3. The correction coefficient is generated through the linear superposition calculation in step S3 and is bound and stored with the unique encoding of the skill label. When extracting data, perform a consistency check on the skill label name. The check rule is: When the characters of the skill label name are exactly the same (including case and spaces) in the threshold configuration table and the correction coefficient table, it is considered a match; otherwise, trigger an exception alarm and abort the process.

[0040] Perform the following operations on the original matching threshold according to the matching threshold correction coefficient. When the original threshold is 78.00 points and the correction coefficient is 1.15, the preliminary adjusted threshold is 89.70 points; when the correction coefficient is 0.92, the preliminary adjusted threshold is 71.76 points. The calculation results are classified and stored in the temporary threshold table in the intermediate database according to the skill label name. The structure of the temporary threshold table includes fields for skill label name, original threshold, correction coefficient, and preliminary adjusted threshold. When writing data, enable the database row-level lock mechanism to prevent data overwriting or loss caused by concurrent writing.

[0041] When the initially adjusted threshold exceeds the preset threshold range, it is truncated according to the boundary values of the preset threshold range. The specific process of generating the compliant adjusted threshold is as follows: Obtain the preset threshold range boundary values from the threshold configuration policy table. The setting logic of the boundary values is as follows: According to the rigidity degree of the skill requirements of the target position, the skill tags are divided into two categories: core skills and auxiliary skills. The lower limit of the threshold range for core skills is set to the 30th percentile value of the skill test scores of historical successful candidates, and the upper limit is the 90th percentile value; the lower limit of auxiliary skills is the 20th percentile value, and the upper limit is the 80th percentile value. For example, if the 30th percentile score of historical successful candidates for the core skill tag is 65 points and the 90th percentile is 92 points, the threshold range is set to 65 - 92 points. When the initially adjusted threshold exceeds the upper limit value, the threshold is set to the upper limit value; when it is lower than the lower limit value, the threshold is set to the lower limit value. For example, when the initially adjusted threshold is 95.00 points and the upper limit is 92.00 points, the compliant adjusted threshold is truncated to 92.00 points; when the initially adjusted threshold is 58.00 points and the lower limit is 65.00 points, it is adjusted to 65.00 points. The compliant adjusted threshold after truncation processing is written into the audit queue and waits for the priority weight fusion operation.

[0042] The specific process of generating the updated threshold range based on the compliant adjusted threshold and the priority weights of the target position skill tags is as follows: Extract the priority weights of the target position skill tags from the weight configuration table in the position portrait database. The setting basis of the weights is the importance level annotation of the skill items in the position description. The weights of the core skill tags are dynamically calculated according to the annotation level: the weight of the annotation "required" is 0.75, and the weight of the annotation "preferred" is 0.60; the weights of the auxiliary skill tags are uniformly set to 0.30. The compliant adjusted threshold and the priority weights are weighted and fused to generate the updated threshold range. The weighted fusion rule is: Taking the compliant adjusted threshold as the reference value, the floating interval range is negatively correlated with the weight value. The calculation formula is floating interval = reference value × (1 - weight) × adjustment factor, and the adjustment factor is set according to the urgency of the position recruitment. Under normal circumstances, the adjustment factor is 0.10. For example, when the compliant adjusted threshold for core skills is 85.00 points and the weight is 0.75, the floating interval = 85.00 × (1 - 0.75) × 0.10 = 2.13 points, and the updated threshold range is 82.87 - 87.13 points; when the compliant adjusted threshold for auxiliary skills is 70.00 points and the weight is 0.30, the floating interval = 70.00 × (1 - 0.30) × 0.10 = 4.90 points, and the threshold range is 65.10 - 74.90 points.

[0043] The specific process of writing the updated threshold range to the job profile database and triggering the timestamp record of the threshold effectiveness is as follows: Update the generated updated threshold range to the matching threshold configuration table of the job profile database according to the skill tag name, overwriting the original threshold data. Create a database transaction snapshot before the update operation is executed. The snapshot content includes the original threshold data, the operation time, and the operation session ID to ensure that in case of an exception, it can be rolled back to the previous version through the snapshot. Insert a timestamp record of the threshold effectiveness into the transaction log table. The record fields include the operation time (accurate to milliseconds), the operator identifier (a unique identification code automatically generated by the system), the skill tag name, the old threshold range, and the new threshold range. The generation of the timestamp record and the threshold update are ensured to be consistent through a database atomic transaction. The implementation method of the atomic transaction is: Lock all write operations of the relevant data tables before the transaction is committed, and release the lock after both the log record and the threshold data are successfully written. The effective threshold range is automatically enabled at zero o'clock on the next natural day. When enabling, verify whether the new threshold range is compatible with the skill tag status of the current job profile. If there is a conflict (such as the lower limit of the threshold being higher than the highest score of the current candidate), trigger a manual review process. Migrate the old threshold version to the historical version library. Mark the invalidation timestamp and the operator identifier during migration. The historical version data is retained for five years for compliance auditing and data analysis purposes.

[0044] S5. Based on the deviation direction of the KL divergence value from the updated threshold range, determine the sampling weight adjustment strategy for the candidate's skill test score. The specific implementation is as follows: The process of determining the deviation direction of the KL divergence value from the updated threshold range, where the deviation direction is divided into positive deviation from the upper threshold and negative deviation from the lower threshold, is as follows: Extract the KL divergence value of the current skill tag from the calculation result of the KL divergence value in step S2 and compare it with the updated threshold range generated in step S4. The comparison process is implemented through a sign function operation: When the KL divergence value is greater than the upper limit value of the updated threshold range, it is marked as positive deviation from the upper threshold; when the KL divergence value is less than the lower limit value of the updated threshold range, it is marked as negative deviation from the lower threshold. The input of the sign function is the difference between the KL divergence value and the upper and lower limits of the threshold range, and the output is the deviation direction identifier. A positive number represents positive deviation, and a negative number represents negative deviation. The determination result of the deviation direction is stored in the deviation record table of the analysis database according to the skill tag name. The fields in the table include the skill tag name, the KL divergence value, the upper and lower limits of the threshold range, the deviation direction identifier, and the timestamp.

[0045] When the deviation direction is a positive deviation from the upper threshold limit, the specific process of reducing the sampling weight of the candidate skill test scores higher than the current upper threshold limit is as follows: Obtain the upper limit value of the updated threshold range and the candidate skill test score data, and filter out the candidate records with skill test scores higher than the current upper threshold limit. Perform an attenuation operation on the sampling weights of this part of the records. The rule of the attenuation operation is: Set the weight attenuation factor according to the amplitude exceeding the upper limit. The deviation amplitude is divided into three intervals: 0 - 5 points, 5 - 10 points, and above 10 points, corresponding to attenuation factors of 0.9, 0.7, and 0.5 respectively. For example, if a candidate's skill test score is 95 points and the upper threshold limit is 90 points, with an exceeding amplitude of 5 points, then the attenuation factor is 0.7, and the original weight of 1.0 is adjusted to 0.7. The adjusted weight data is stored in the sampling weight configuration table associated by candidate ID. When storing, verify whether the weight value is within the valid range of 0 to 1, and automatically truncate it to the boundary value when it exceeds.

[0046] When the deviation direction is a negative deviation from the lower threshold limit, the specific process of increasing the sampling weight of the candidate skill test scores lower than the current lower threshold limit is as follows: Obtain the lower limit value of the updated threshold range and the candidate skill test score data, and filter out the candidate records with skill test scores lower than the current lower threshold limit. Perform a gain operation on the sampling weights of this part of the records. The rule of the gain operation is: Set the weight gain factor according to the amplitude lower than the lower limit. The deviation amplitude is divided into three intervals: 0 - 5 points, 5 - 10 points, and above 10 points, corresponding to gain factors of 1.1, 1.3, and 1.6 respectively. For example, if a candidate's skill test score is 58 points and the lower threshold limit is 60 points, with a lower amplitude of 2 points, then the gain factor is 1.1, and the original weight of 1.0 is adjusted to 1.1. The adjusted weight data is stored in the sampling weight configuration table associated by candidate ID. When storing, verify whether the weight value is within the valid range (0 to 2), and automatically truncate it to the boundary value when it exceeds.

[0047] The specific process of generating a dynamic sampling weight adjustment strategy according to the deviation direction type and deviation amplitude, where the sampling weight adjustment strategy includes the piecewise function mapping rules of the weight attenuation factor and the gain factor is as follows: Construct a piecewise mapping relationship table between the deviation amplitude and the weight factor. The piecewise mapping rule is dynamically adjusted according to the skill stability requirements of the target position. For positions with high stability requirements (such as core technology positions), a steeper factor gradient is adopted. For example, when the positive deviation amplitude of the upper threshold limit is 0 - 5 points, the corresponding attenuation factor is 0.8 (original 0.9), and when it is 5 - 10 points, it is 0.6 (original 0.7); for ordinary positions, the default gradient is maintained. The mapping relationship table is stored in the policy library according to the skill label name. The structure of the policy library includes fields such as skill label name, deviation direction, deviation amplitude interval, weight factor, and effective timestamp. When the policy is generated, a version control mechanism is triggered. Each update generates a new version and retains the historical version data, supporting policy rollback and effect comparison analysis.

[0048] S6. Retrain the matching model according to the skill test score sampling weight adjustment strategy and output the updated candidate recommendation list. The specific implementation is as follows: The specific process of extracting the weight decay factor and gain factor mapping table from the sampling weight adjustment strategy is as follows: Read the weight decay factor and gain factor mapping table of the current skill tags of the target position from the policy library generated in step S5. The mapping table is stored classified by skill tag name and deviation direction. When extracting, perform an exact match check on the skill tag name to ensure that only the mapping table entries that are exactly the same as the activated skill tags in the current position profile are extracted. For example, when the activated skill tags of the target position are "Java skill" and "distributed system development", exclude the mapping table data of non-activated tags such as "front-end development". The data structure of the mapping table includes a skill tag name field, a deviation direction field (upper limit positive deviation or lower limit negative deviation), a deviation amplitude interval field (such as 0-5 points, 5-10 points, etc.) and a corresponding weight factor field. The extracted mapping table is sorted by skill tag name and loaded into the training parameter pool in memory. When loading, check the data integrity. If there are missing fields or out-of-bounds values, trigger an exception handling process.

[0049] The specific process of generating a weighted training data set by adjusting the sampling weight distribution of candidate skill test scores based on the weight decay factor and gain factor is as follows: Extract candidate skill test score data from the associated storage database in step S1, and adjust the sampling weight of each candidate according to the sampling weight adjustment strategy in step S5. For candidates with skill test scores higher than the current threshold upper limit, reduce their weight according to the decay factor in the mapping table. For example, if a candidate's "Java skill" score is 95 points (threshold upper limit 90 points) and the decay factor is 0.7, then their weight is adjusted from 1.0 to 0.7; for candidates with scores lower than the current threshold lower limit, increase their weight according to the gain factor. For example, if a candidate's "distributed system development" score is 55 points (threshold lower limit 60 points) and the gain factor is 1.3, then the weight is adjusted from 1.0 to 1.3. The adjusted weight data is combined with the candidate ID and skill tag as a composite primary key and stored in the weighted training data set. The structure of the data set includes candidate ID, skill tag name, skill test score, adjusted weight, original weight and timestamp fields. When storing, use the database transaction lock mechanism to prevent concurrent write conflicts and enable data version control to support historical traceability.

[0050] Retrain the matching model using the weighted training dataset. The specific process of updating the parameters of the matching model by minimizing the prediction error of the weighted samples is as follows: Load the previously trained matching model, and input the weighted training dataset into the model for incremental training. During the training process, the weight of the loss function for each sample is dynamically scaled according to the adjusted weight. The total error of the loss function is calculated as the sum of the weighted errors of all samples. The weighted error is the square difference between the predicted value and the actual value of the sample multiplied by the adjusted weight. When updating the model parameters, the stochastic gradient descent algorithm is used, and the learning rate is dynamically decayed according to the current training epoch. The specific rule is: the initial learning rate is set to 0.01, and the learning rate is multiplied by the decay coefficient 0.9 every 10 training epochs, with the minimum learning rate not less than 0.0001. After training is completed, the updated model parameters are stored in the model version library according to the skill tag name, and the model version number, training timestamp, and the version number of the weighted strategy used are recorded to ensure the traceability of the version association between the model and the weight adjustment strategy.

[0051] Filter the output results of the matching model according to the updated threshold range, and generate an updated candidate recommendation list with the threshold effective timestamp appended. The specific process is as follows: Apply the retrained matching model to the current candidate pool to generate an original recommendation score list. According to the updated threshold range written to the job profile database in step S4, filter the recommendation score list: Only retain the candidates whose skill test scores are within the threshold range. For example, if the updated threshold range for "Java skill" is 85 - 95 points, then filter out the candidates with scores lower than 85 points or higher than 95 points. The filtered recommendation list is sorted by skill tag name. The sorting rule is: Sort in descending order of the recommendation score under the same skill tag, and for those with the same score, sort in ascending order of the candidate ID. The threshold effective timestamp appended to the recommendation list is extracted from the database transaction log in step S4. When extracting, verify the correspondence between the timestamp and the current effective threshold version. If there is a timestamp conflict, trigger an alarm and intervene manually. The finally generated recommendation list is written to the display queue of the recruitment platform. The data structure of the display queue includes fields such as candidate ID, skill tag name, recommendation score, threshold range, and effective timestamp. At the same time, migrate the filtered candidate data to the pending review pool. Each record in the pending review pool is marked with the filtering reason (such as "exceeding the upper limit" or "below the lower limit") and the threshold version number for the recruitment specialist to review.

[0052] In this embodiment, a closed-loop feedback mechanism is constructed to achieve dynamic optimization of the job matching threshold. In traditional recruitment systems, threshold adjustment usually relies on explicit indicators such as the number of resumes and the market supply-demand ratio, lacking quantitative control over the deviation of the candidate ability distribution, which is prone to trigger a vicious cycle of low-quality data contaminating the model. This embodiment introduces the KL divergence value to quantify the deviation degree between the candidate skill test score distribution and the job benchmark curve, and dynamically adjusts the matching threshold correction coefficient through the PID control algorithm. Among them, the proportional coefficient is linked with the correlation coefficient of the skill-performance data, enabling the threshold correction amplitude to be adaptively adjusted according to the data quality feedback, thereby suppressing error accumulation. On this basis, an asymmetric adjustment (downweighting for positive deviation and upweighting for negative deviation) is performed on the sampling weights of high / low-score candidates based on the deviation direction, actively blocking the negative impact of low-quality data on model training. In addition, by strictly binding the threshold effective timestamp and the model version number, the operation timing consistency of threshold update, model training, and result filtering is ensured, avoiding data version conflicts. Compared with the prior art, the systematic integration of distribution deviation detection, dynamic control theory, weight reverse adjustment, and timing synchronization mechanism significantly improves the talent matching accuracy and the long-term stability of the recommendation system while ensuring recruitment efficiency.

[0053] Embodiment 2: Figure 2 The structural schematic diagram of the matching threshold adjustment system for dynamic update of the job portrait according to the present invention is given. The matching threshold adjustment system for dynamic update of the job portrait includes the following modules: Data acquisition module: used to acquire the candidate skill test scores corresponding to the current skill tags of the target job and the performance data after employment; Divergence calculation module: used to calculate the KL divergence value between the distribution curve of the candidate skill test scores within a preset period and the job skill requirement benchmark curve of the target job; Dynamic adjustment module: when the KL divergence value exceeds the KL divergence threshold, used to generate a matching threshold correction coefficient based on the PID control algorithm; Threshold update module: used to update the matching threshold of the current job skill tag according to the matching threshold correction coefficient, and generate the updated threshold range; Strategy generation module: used to determine the sampling weight adjustment strategy of the candidate skill test scores based on the deviation direction between the KL divergence value and the updated threshold range; Model optimization module: used to retrain the matching model according to the skill test score sampling weight adjustment strategy and output the updated candidate recommendation list.

[0054] The calculations involved in the embodiments are all dimensionless numerical calculations, and the preset parameters and threshold selections in the calculations are set by those skilled in the art according to the actual situation.

[0055] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any 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.

[0056] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of the examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and the inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0057] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0058] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division, and there can be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the couplings or direct couplings or communication connections shown or discussed with each other can be through some interfaces, and the indirect couplings or communication connections of the devices or modules can be in electrical, mechanical, or other forms.

[0059] As described above, this is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0060] Finally: The above description is only the preferred embodiments of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A matching threshold adjustment method for dynamic updating of job profiles, characterized in that: The steps include: S1. Obtain the candidate's skill test scores and post-employment performance data corresponding to the current skill tag of the target position; S2. Calculate the KL divergence value between the distribution curve of the candidate's skill test score within a preset period and the benchmark curve of the skill requirements of the target position; S3, when the KL divergence value exceeds the KL divergence threshold, a matching threshold correction coefficient is generated based on the PID control algorithm; S4. Update the matching threshold of the current job skill tag according to the matching threshold correction coefficient to generate an updated threshold range; S5. Determine a sampling weight adjustment strategy for the candidate's skill test score based on the deviation direction of the KL divergence value from the updated threshold range; S6. Retrain the matching model according to the skill test score sampling weight adjustment strategy and output an updated candidate recommendation list.

2. The matching threshold adjustment method for dynamic updating of job profiles according to claim 1 is characterized in that: The proportional coefficient of the PID control algorithm is dynamically adjusted based on the correlation coefficient between the candidate's skill test score and performance data.

3. The matching threshold adjustment method for dynamic updating of job profiles according to claim 1 is characterized in that: Obtain the candidate's skill test scores and post-employment performance data corresponding to the current skill tags of the target position, including: Obtain the recruitment description text of the target position, parse the recruitment description text to generate the current skill label of the target position; Obtain the skill test scores corresponding to the current skill tags of the candidate and the target position from the online assessment system; Obtain the candidate's post-employment performance data in the position corresponding to the current skill tag of the target position from the human resources management system.

4. The matching threshold adjustment method for dynamic updating of job profiles according to claim 1 is characterized in that: Calculate the KL divergence value between the distribution curve of the candidate's skill test score within the preset period and the benchmark curve of the target position's skill requirements, including: The candidate's skill test scores are binned according to the skill label dimension within a preset period to generate a discretized skill test score distribution; Dynamically adjust the bin width based on the distribution of the benchmark curve of the target job skill requirements, so that the distribution of the discretized skill test scores is aligned with the data interval of the benchmark curve; The discretized skill test score distribution is recalculated based on the adjusted bin width, and the kurtosis and skewness parameters of the benchmark curve are extracted; The distribution function of the benchmark curve is modified based on the kurtosis parameter and the skewness parameter to generate a dynamically adapted benchmark curve for target job skill requirements; Calculate the KL divergence value of the discretized skill test score distribution and the dynamically adapted target job skill requirement benchmark curve.

5. The matching threshold adjustment method for dynamic updating of job profiles according to claim 1 is characterized in that: When the KL divergence value exceeds the KL divergence threshold, a matching threshold correction coefficient is generated based on the PID control algorithm, including: Obtain the correlation coefficient between the candidate's skill test score and performance data. The correlation coefficient is calculated using the Pearson product-moment correlation coefficient formula; Dynamically adjust the proportional coefficient of the PID control algorithm based on the degree of deviation between the correlation coefficient and the preset proportional reference value; The integral term is calculated based on the difference between the KL divergence value and the KL divergence threshold, and the differential term is calculated based on the rate of change of the KL divergence value; The proportional coefficient, integral term and differential term are superimposed according to the preset weights to generate the matching threshold correction coefficient.

6. The matching threshold adjustment method for dynamic updating of job profiles according to claim 1 is characterized in that: The matching threshold of the current job skill tag is updated according to the matching threshold correction coefficient to generate an updated threshold range, including: Obtain the original matching threshold of the current job skill tag and the matching threshold correction coefficient generated in step S3; The original matching threshold is linearly adjusted according to the matching threshold correction coefficient to generate a preliminary adjustment threshold; When the initial adjustment threshold exceeds the preset threshold range, it is truncated according to the preset threshold range boundary value to generate a compliance adjustment threshold; Generate an updated threshold range based on the compliance adjustment threshold and the priority weight of the target job skill tag; The updated threshold range is written into the job profile database and the threshold effective timestamp record is triggered.

7. The matching threshold adjustment method for dynamic updating of job profiles according to claim 1 is characterized in that: Based on the deviation direction of the KL divergence value from the updated threshold range, the sampling weight adjustment strategy of the candidate's skill test score is determined, including: Determine the deviation direction of the KL divergence value from the updated threshold range. The deviation direction is divided into positive deviation from the upper threshold limit and negative deviation from the lower threshold limit. When the deviation direction is a positive deviation towards the upper threshold limit, the sampling weight of the candidate skill test scores that are higher than the current upper threshold limit is reduced; When the deviation direction is a negative deviation from the lower threshold, the sampling weight of the skill test scores of candidates below the current lower threshold is increased; Generate a dynamic sampling weight adjustment strategy based on the deviation direction type and deviation amplitude.

8. The matching threshold adjustment method for dynamic updating of job profiles according to claim 7 is characterized in that: The sampling weight adjustment strategy includes a piecewise function mapping rule between weight attenuation factors and gain factors.

9. The matching threshold adjustment method for dynamic updating of job profiles according to claim 1 is characterized in that: Retrain the matching model based on the skill test score sampling weight adjustment strategy and output an updated candidate recommendation list, including: Extract a weight decay factor and gain factor mapping table from the sampling weight adjustment strategy; Adjust the sampling weight distribution of the candidate's skill test scores based on the weight decay factor and the gain factor to generate a weighted training data set; Retrain the matching model using the weighted training data set, and update the matching model parameters by minimizing the prediction error of the weighted samples; The matching model output results are filtered according to the updated threshold range, and an updated candidate recommendation list is generated and the threshold effective timestamp is attached.

10. A matching threshold adjustment system for dynamic updating of job profiles, used to implement the matching threshold adjustment method for dynamic updating of job profiles according to any one of claims 1 to 9, characterized in that: Includes the following modules: Data acquisition module: used to obtain the candidate's skill test scores and post-employment performance data corresponding to the current skill tag of the target position; Divergence calculation module: used to calculate the KL divergence value between the distribution curve of the candidate's skill test score within a preset period and the benchmark curve of the target position's skill requirements; Dynamic adjustment module: when the KL divergence value exceeds the KL divergence threshold, it is used to generate a matching threshold correction coefficient based on the PID control algorithm; Threshold update module: used to update the matching threshold of the current job skill label according to the matching threshold correction coefficient and generate the updated threshold range; Strategy generation module: used to determine the sampling weight adjustment strategy of the candidate's skill test score based on the deviation direction of the KL divergence value and the updated threshold range; Model optimization module: used to retrain the matching model according to the skill test score sampling weight adjustment strategy and output an updated candidate recommendation list.

Images