A professional talent intelligent matching and screening method based on financial and tax business demand

Abstract

The present application relates to the technical field of human resources matching, and particularly relates to a professional talent intelligent matching and screening method based on financial and tax business demand. The method comprises the following steps: constructing and continuously updating a knowledge base based on historical operation data; obtaining project demand information and extracting candidate labels, and determining labels meeting preset standards as screening reference labels; matching and screening candidate object information from the knowledge base based on the screening reference labels; recording screening results and rechecking and confirming the candidate objects, and recording object information that passes the rechecking; after the system runs for a preset time length, calculating a comprehensive retention deviation according to actual execution data of the rechecked and confirmed objects and judging whether the matching strategy is qualified; when it is judged that the matching strategy is not qualified, locating the deviation causes and performing corresponding strategy adjustment operations. Through the technical means of constructing a knowledge base, label screening, retention deviation calculation and cause positioning, the present application improves the accuracy and long-term stability of matching financial and tax talents with business demand.

Description

Technical Field

[0001] This invention relates to the field of human resource matching technology, and in particular to a method for intelligent matching and screening of professional talents based on the needs of financial and tax business. Background Technology

[0002] In the financial and tax services industry, companies typically need to match talent with the corresponding professional skills and industry experience based on client project requirements. Traditional methods of matching financial and tax talent mainly rely on manual resume screening and interview assessments, which suffers from high subjectivity, low efficiency, and inconsistent matching standards. With the increasing diversification and complexity of financial and tax business needs, some companies have begun to introduce automated screening systems based on tag matching, matching candidates by assigning skill tags to their resumes and project requirements. However, existing systems usually only perform a one-time screening at the matching stage, lacking mechanisms for tracking feedback and quantitative evaluation of the matching results. Furthermore, they cannot automatically pinpoint the root causes of matching strategy deficiencies and make targeted corrections based on feedback results. This leads to long-term rigidity in matching strategies, a lack of adaptive optimization, difficulty in ensuring long-term suitability between personnel and positions, and persistently high staff turnover rates.

[0003] Chinese Patent Publication No. CN119130406A discloses an intelligent recruitment system and method based on human-computer interaction and resume mining. The system includes: HR providing job requirements; a robot analyzing the requirements and generating HR requirement tags; the robot searching for resumes online, analyzing them, and extracting corresponding resume tags; filtering resumes based on the HR requirement tags and sorting them; having the robot chat online with job seekers to verify the authenticity of the resume tags and mine temporary tags; assessing the importance of the temporary tags and constructing a job seeker evaluation model based on the HR requirement tags, resume tags, and temporary tags; the robot recommending job seekers who meet the HR's requirements to the HR and generating an evaluation report based on the job seeker evaluation model. This invention improves the efficiency and accuracy of the recruitment process through resume screening, tagging, building a job knowledge graph, and intelligent dialogue verification.

[0004] Therefore, the existing technology has the following problems: The lack of follow-up feedback on the actual performance of recruited personnel after they join the company makes it impossible to automatically locate the root cause of the matching strategy's defects based on long-term adaptability data, and also makes it impossible to achieve adaptive adjustment and closed-loop optimization of the matching strategy. Summary of the Invention

[0005] To address this, the present invention provides a method for intelligent matching and screening of professional talents based on the needs of financial and tax business, in order to overcome the shortcomings of existing technologies, such as the lack of tracking and feedback on the actual performance of recruited personnel after they have joined the company, the inability to automatically locate the root cause of the matching strategy based on long-term adaptability data, and the inability to achieve adaptive adjustment and closed-loop optimization of the matching strategy.

[0006] To achieve the above objectives, this invention provides a method for intelligent matching and screening of professional talent based on financial and tax business needs, comprising: Step S1: Build and continuously update the knowledge base based on historical operational data; Step S2: Obtain the input project requirement information, extract candidate tags from the project requirement information, and determine the candidate tags that meet the preset tag criteria as the screening benchmark tags. Step S3: Based on the filtering benchmark tags, match and filter candidate object information from the knowledge base; Step S4: Record the candidate information obtained after screening, review and confirm the screened candidate, and record the information of the candidates that pass the review. Step S5: After the system has run for a preset period of time, calculate the comprehensive retention deviation based on the actual execution data of the objects that have passed the review, and determine whether the matching strategy is qualified. Step S6: When the matching strategy is determined to be unqualified, locate the cause of the deviation and perform the corresponding strategy adjustment operation according to the cause of the deviation.

[0007] Further, in step S5, the overall retention deviation is calculated based on the actual execution data of the reviewed objects, including: Set a testing cycle and count the actual number of new hires and actual departures of employees who passed the review within the current preset time interval; Get the actual number of new hires and actual number of departures within the previous preset time interval; Based on the actual number of new hires and actual number of departures in the previous preset time interval, and combined with at least one of the following parameters from the previous preset time interval: screening ratio, hiring ratio, and number of received demands, calculate the expected number of departures in the current preset time interval. The overall retention deviation is determined based on the difference between the actual number of departures and the expected number of departures within the current preset time interval.

[0008] Further, in step S5, determining whether the matching strategy is qualified includes: If the overall retention deviation is less than or equal to the preset deviation threshold, the matching strategy is deemed qualified. If the overall retention deviation is greater than the preset deviation threshold, the matching strategy is deemed unqualified.

[0009] Furthermore, in step S6, the causes of positioning deviation include: Obtain the overall retention rate for each of several preset historical time intervals; Calculate the variance of each overall retention rate; When the variance value is less than or equal to the preset variance threshold, the deviation is determined to be the first type of cause, which is that the matching and screening parameters deviate from the standard value. When the variance value is greater than the preset variance threshold, the deviation is determined to be a second type of cause, which is that the mapping relationship between the demand information and the screening benchmark label deviates from the standard value.

[0010] Furthermore, when the cause of the deviation is determined to be the first type of cause, the corresponding strategy adjustment operation is performed according to the cause of the deviation, including: Obtain the ratio of the actual number of employees who passed the review within each historical preset time interval to the industry average number of employees during the same period, and use it as the specific ratio corresponding to each time interval. A specific ratio curve is constructed with the preset time interval as the horizontal axis and the specific ratio value as the vertical axis; Calculate the slope of the specific ratio curve at the current node, and denot it as the specific slope; The screening parameters are adjusted according to the specific slope, and the specific slope is positively correlated with the adjustment range of the screening parameters.

[0011] Furthermore, before adjusting the screening parameters according to the specific slope, the method further includes: Calculate the ratio of the actual number of employees hired in the current preset time interval to the actual number of employees hired in the previous preset time interval, and record it as the employee hiring change ratio. When the change in the number of new hires is greater than one, the adjustment range of the screening parameter is increased according to the change in the number of new hires, and the change in the number of new hires is positively correlated with the adjustment range. When the change in the number of hires is less than or equal to one, the adjustment range of the screening parameter is reduced according to the change in the number of hires, and the change in the number of hires is negatively correlated with the adjustment range.

[0012] Furthermore, after adjusting the screening parameters according to the specific slope, the method further includes: Based on the adjusted screening parameters, the operation of matching candidate information based on the screening benchmark label, calculating the overall retention deviation, and determining whether the matching strategy is qualified is re-executed. If the matching strategy is still unqualified after re-evaluation, the deviation reason will be corrected to the second type of reason, and the strategy adjustment operation corresponding to the second type of reason will be executed.

[0013] Furthermore, when the cause of the deviation is determined to be the second type of cause, the corresponding strategy adjustment operation is executed according to the cause of the deviation, including: Obtain the length of employment for each approved employee from their start date to their departure date; Calculate the ratio of each working duration to the detection cycle, which is used as the individual working percentage for each object; Calculate the average percentage of employed individuals among all those who passed the review, and use this as the overall employed percentage. Based on the mapping relationship between the comprehensive on-the-job ratio adjustment demand information and candidate labels, the mapping relationship is used to control the correlation strength between demand information and each candidate label.

[0014] Furthermore, based on the mapping parameters between the comprehensive on-the-job ratio adjustment demand information and candidate tags, including: The ratio of the number of times each candidate tag appears in a historical preset time interval to the total number of times in the preset time interval is obtained, which is used as the comprehensive frequency of each candidate tag; For candidate tags whose overall frequency is higher than a preset frequency threshold, the adjustment range of their corresponding mapping relationship is reduced. For candidate tags whose overall frequency is lower than or equal to the preset frequency threshold, the adjustment range of their corresponding mapping relationship is maintained.

[0015] Furthermore, after the cause of the deviation is corrected to the second type of cause, the strategy adjustment operation corresponding to the second type of cause is executed, including: Obtain the overall frequency of each candidate tag within a preset historical time interval; The weight correction coefficient for each candidate label is determined based on the comprehensive frequency, and the weight correction coefficient for a candidate label with a higher comprehensive frequency is smaller. The mapping relationship between the demand information and each candidate label is updated based on the weight correction coefficient.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: This embodiment sets a detection cycle to statistically analyze the actual number of new hires and actual number of departures in the current and previous preset time intervals, and calculates the expected number of departures by combining parameters such as the screening ratio, hiring ratio, and demand quantity in the previous preset time interval. Then, the difference between the actual number of departures and the expected number of departures is used as the comprehensive retention deviation, thereby achieving a quantitative evaluation of the retention effect of the matching strategy. At the same time, by comparing the comprehensive retention deviation with a preset deviation threshold determined based on the standard deviation of historical deviation samples, it is determined whether the matching strategy is qualified. If it is unqualified, subsequent deviation positioning and strategy adjustment are triggered. This provides an objective and quantitative triggering mechanism for the closed-loop adaptive optimization of the matching strategy, avoiding the uncertainty caused by subjective judgment, and ensuring that the matching strategy can be identified and the correction process is triggered in a timely manner when the retention rate drops abnormally. This effectively improves the long-term stability and accuracy of matching financial and tax professionals with business needs.

[0017] Furthermore, this embodiment obtains the overall retention rate of several historical preset time intervals and calculates its variance value. The variance value is compared with a preset variance threshold. When the variance value is less than or equal to the preset variance threshold, it is determined to be the first type of reason for the deviation of the matching screening parameters from the standard value. When the variance value is greater than the preset variance threshold, it is determined to be the second type of reason for the deviation of the mapping relationship between the demand information and the screening benchmark label from the standard value. This achieves accurate tracing of the root cause of the matching strategy defect, provides a clear direction for subsequent targeted strategy adjustments, avoids the problem of system instability caused by blindly adjusting all parameters, and effectively improves the efficiency and accuracy of matching strategy optimization.

[0018] Furthermore, this embodiment constructs a unique ratio curve by obtaining the ratio of the actual number of new hires to the industry average number of new hires during each historical preset time interval and calculates the unique slope of the current node. Based on the sign and absolute value of the unique slope, the screening parameters are adjusted in a directional and quantitative manner, achieving dynamic adaptation of the screening criteria to the industry's talent supply and demand situation. At the same time, the adjustment range is corrected a second time by calculating the ratio of changes in new hires, so that the adjustment amount of the screening parameters reflects information from two dimensions: the rate of change of the unique ratio and the trend of changes in the number of new hires. This avoids over-adjustment or under-adjustment caused by abnormal fluctuations in the number of new hires, and improves the accuracy and robustness of the screening parameter adjustment.

[0019] Furthermore, this embodiment calculates the individual employment rate by obtaining the ratio of the employment duration to the detection cycle of each reviewed object, and further calculates the comprehensive employment rate. The comprehensive employment rate is used as a quantitative feedback signal to adjust the mapping relationship between the demand information and the candidate labels, realizing a closed-loop linkage between the label mapping relationship and the actual long-term adaptation effect. At the same time, by obtaining the comprehensive frequency of each candidate label in the historical preset time interval, the adjustment range of the mapping relationship of candidate labels with a comprehensive frequency higher than the preset frequency threshold is reduced, and the adjustment range of the mapping relationship of candidate labels with a comprehensive frequency lower than or equal to the preset frequency threshold is maintained. This realizes differentiated and fine control by adopting a conservative adjustment strategy for high-frequency stable labels to maintain system stability and an active adjustment strategy for low-frequency labels to quickly correct deviations. This ensures the effectiveness of the mapping relationship adjustment while avoiding system oscillation caused by over-adjustment.

[0020] Furthermore, this embodiment corrects the deviation cause to the second type of cause after the first type of cause adjustment is ineffective and performs a global update of the mapping relationship. It obtains the comprehensive frequency of each candidate label and determines the weight correction coefficient based on the comprehensive frequency. The weight correction coefficient corresponding to the candidate label with the higher comprehensive frequency is smaller, and the weight correction coefficient corresponding to the candidate label with the lower comprehensive frequency is larger. Then, based on the weight correction coefficient, the actual on-the-job feedback data of each candidate label is superimposed on its current mapping relationship parameters to generate an updated mapping relationship. This realizes a differentiated mapping relationship update mechanism that adopts conservative updates for high-frequency stable labels to maintain system stability and active updates for low-frequency uncertain labels to make full use of the current feedback data for rapid correction. This fundamentally solves the problem of the mapping relationship between demand information and screening benchmark labels deviating from the standard value, and ensures the accuracy and adaptability of the label extraction process. Attached Figure Description

[0021] Figure 1 This is a flowchart of the intelligent matching and screening method for professional talents based on the needs of financial and tax business in this embodiment. Figure 2 This is a flowchart illustrating the calculation of the comprehensive retention bias in the intelligent matching and screening method for professional talents based on financial and tax business needs in this embodiment. Figure 3 This is a flowchart illustrating the process of determining whether a matching strategy is qualified in the intelligent matching and screening method for professional talents based on financial and tax business needs in this embodiment. Detailed Implementation

[0022] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0023] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0024] Please see Figures 1-3 As shown, Figure 1 This is a flowchart of the intelligent matching and screening method for professional talents based on the needs of financial and tax business in this embodiment. Figure 2 This is a flowchart illustrating the calculation of the comprehensive retention bias in the intelligent matching and screening method for professional talents based on financial and tax business needs in this embodiment. Figure 3 This is a flowchart illustrating the process of determining whether a matching strategy is qualified in the intelligent matching and screening method for professional talents based on financial and tax business needs in this embodiment.

[0025] This embodiment provides a method for intelligent matching and screening of professional talents based on financial and tax business needs, including: Step S1: Build and continuously update the knowledge base based on historical operational data; Step S2: Obtain the input project requirement information, extract candidate tags from the project requirement information, and determine the candidate tags that meet the preset tag criteria as the screening benchmark tags. Step S3: Based on the filtering benchmark tags, match and filter candidate object information from the knowledge base; Step S4: Record the candidate information obtained after screening, review and confirm the screened candidate, and record the information of the candidates that pass the review. Step S5: After the system has run for a preset period of time, calculate the comprehensive retention deviation based on the actual execution data of the objects that have passed the review, and determine whether the matching strategy is qualified. Step S6: When the matching strategy is determined to be unqualified, locate the cause of the deviation and perform the corresponding strategy adjustment operation according to the cause of the deviation.

[0026] In this embodiment, the knowledge base includes the capability feature vectors of candidate objects and their corresponding historical performance data. The system accesses the enterprise's historical financial and tax project database and human resource management system, collecting basic information, project participation records, project completion quality scores, and performance feedback data of candidate objects who have participated in financial and tax business projects such as tax planning, auditing, accounting standards consulting, and financial analysis within a preset historical time interval, thus constructing an initial knowledge base. The capability feature vector of each candidate object includes at least the dimensions of professional skills, industry experience, project complexity handling ability, and historical performance. After each matching task is completed, the knowledge base feeds back the matching result and the candidate object's actual work performance data as incremental samples to iteratively correct the candidate object's capability feature vectors, thereby achieving dynamic evolution and updating of the knowledge base.

[0027] The project requirement information is input by the user through a terminal interface. The system uses a natural language processing engine to perform semantic parsing and structured processing on the project requirement information, extracting an initial set of candidate tags. The system has a preset tag admission standard library, which includes a historical effective usage frequency threshold and tag semantic clarity rules. For each extracted candidate tag, the system queries its occurrence frequency and effective matching rate in historical operation data, and removes candidate tags whose occurrence frequency is lower than a preset frequency threshold or whose effective matching rate is lower than a preset matching rate threshold; candidate tags that simultaneously meet the frequency threshold and matching rate threshold are determined as the screening benchmark tags. The preset frequency threshold is determined based on the distribution characteristics of the occurrence frequency of all candidate tags in historical operation data, specifically by multiplying the median or average of the frequency distribution by a preset coefficient and then rounding down; the preset matching rate threshold is determined based on the matching success rate distribution of each tag in historical operation data, specifically by using the lower quartile of the success rate distribution.

[0028] The system calculates the similarity between the screening benchmark labels and the ability feature vectors of each candidate object in the knowledge base, generating a matching score for each candidate object. Candidate objects are sorted from highest to lowest matching score, and those with matching scores higher than a preset matching threshold are marked as having passed the initial screening. The preset matching threshold is determined based on the distribution of matching scores of candidates in historical matching tasks. Specifically, it can be the value obtained by subtracting a standard deviation from the average matching score of historical candidates, or a preset quantile of the matching score distribution of historical matching tasks.

[0029] The system records the information of candidates who passed the initial screening in step S3 into the screening result database. Each record includes a candidate identifier, a matching score, a screening timestamp, and a corresponding set of screening benchmark tags. The system then enters the review and confirmation stage, where a pre-defined expert evaluation body conducts a comprehensive assessment of the candidates who passed the initial screening. The assessment methods include, but are not limited to, structured interviews, professional ability tests, and verification of past project achievements. Based on the evaluation results, the expert evaluation body makes a review decision of pass or fail for each candidate who passed the initial screening. The system receives and records the information of the candidates who passed the review, including the candidate identifier, the corresponding project requirement information identifier, and the review evaluation data, and enters the information of the candidates who passed the review into the talent recruitment database.

[0030] Specifically, in step S5, the overall retention deviation is calculated based on the actual execution data of the reviewed objects, including: Step S21: Set the detection cycle and count the actual number of new hires and actual departures of the approved candidates within the current preset time interval; Step S22: Obtain the actual number of new hires and actual number of departures within the previous preset time interval; Step S23: Based on the actual number of new hires and actual number of departures in the previous preset time interval, and combined with at least one of the following parameters from the previous preset time interval: screening ratio, hiring ratio, and number of received demands, calculate the expected number of departures in the current preset time interval. Step S24: Determine the overall retention deviation based on the difference between the actual number of employees leaving and the expected number of employees leaving within the current preset time interval.

[0031] In this embodiment, a detection period is set to statistically analyze the actual number of newly hired and departing employees who have passed the review within a current preset time interval. Specifically, the detection period is a pre-set time span used to measure the retention status of employees after they have passed the review and started work. The preset time interval is the basic time unit for the system to execute matching tasks. The system uses the detection period as the window length, and within the current preset time interval, it counts the number of employees who have completed the review and confirmation and have actually started work as the actual number of newly hired employees, and counts the number of employees who have changed from being hired to being departing within the above detection period as the actual number of departing employees. The length of the detection period is determined based on the industry's average employee turnover cycle and the project execution cycle of financial and tax business, and can be set to one year or half a year to ensure that the statistical sample size is sufficient and can reflect the long-term retention trend of the employees.

[0032] The system reads the actual number of new hires and actual number of departures from the historical database for the previous preset time interval adjacent to the current preset time interval. The previous preset time interval has the same duration as the current preset time interval. Based on the actual number of new hires and actual number of departures in the previous preset time interval, and combined with at least one of the following parameters—the screening ratio, the hiring ratio, and the number of received job openings—the system calculates the expected number of departures for the current preset time interval.

[0033] The expected number of departures is a predicted value calculated based on historical data, representing the number of departures that should occur within the current preset time interval. The system obtains the screening ratio and onboarding ratio from the previous preset time interval. The screening ratio is the ratio of the number of candidates who passed the initial screening to the total number of received demands in the previous preset time interval, representing the conversion efficiency from demand to screening. The onboarding ratio is the ratio of the actual number of onboarded employees to the number of candidates who passed the review in the previous preset time interval, representing the conversion efficiency from review to actual onboarding. The system calculates the expected number of departures in the current preset time interval based on the actual number of departures, the actual number of onboarding, and the aforementioned ratio parameters from the previous preset time interval. Specifically, the calculation method is as follows: the ratio of the actual number of departures to the actual number of onboarding in the previous preset time interval is used as the base departure rate. Then, the base departure rate is corrected based on the differences in the screening ratio and onboarding ratio between the current preset time interval and the previous preset time interval. The product of the corrected departure rate and the expected number of employees in the current preset time interval is used as the expected number of departures. When multiple historical preset time intervals are used, the basic turnover rate is the weighted average of the turnover rates of each interval. Each weight is determined according to the time-series distance between each interval and the current time point, and the closer the time-series distance, the greater the weight.

[0034] The overall retention deviation is determined based on the difference between the actual number of departures and the expected number of departures within the current preset time interval. This overall retention deviation characterizes the degree of deviation of the current retention rate from the expected retention rate based on historical data. The system calculates the difference between the actual number of departures and the expected number of departures within the current preset time interval, using this difference as the overall retention deviation. When the difference is positive, it indicates that the actual number of departures is higher than the expected value, and the current retention rate is lower than the retention rate predicted based on historical data; when the difference is negative, it indicates that the actual number of departures is lower than the expected value, and the current retention rate is higher than or equal to the retention rate predicted based on historical data. The system uses the absolute value of the difference to characterize the degree of deviation; a larger absolute value indicates a more severe deviation. The calculation result of the overall retention deviation is used to determine the suitability of the matching strategy in subsequent steps.

[0035] Specifically, in step S5, determining whether the matching strategy is qualified includes: If the overall retention deviation is less than or equal to the preset deviation threshold, the matching strategy is deemed qualified. If the overall retention deviation is greater than the preset deviation threshold, the matching strategy is deemed unqualified.

[0036] In this embodiment, the preset deviation threshold is a pre-set critical value used to measure whether the overall retention deviation is within an acceptable range. The system compares the calculated overall retention deviation with the preset deviation threshold and determines the qualification status of the current matching strategy based on the comparison result. The preset deviation threshold is determined as follows: the system obtains the actual retention rate and expected retention rate corresponding to each interval within several historical preset time intervals, calculates the difference between the actual retention rate and the expected retention rate for each historical interval, constructs a deviation sample set from the differences of each historical interval, calculates the standard deviation of the deviation sample set, and uses a preset multiple of the standard deviation as the preset deviation threshold. The preset multiple is set according to the system's tolerance for matching strategy deviation, specifically a value between one and two standard deviations. When historical data is insufficient, the preset deviation threshold uses an initial setting value, which is determined based on the fluctuation range of the industry's average employee turnover rate.

[0037] When the overall retention deviation is less than or equal to the preset deviation threshold, it indicates that the deviation between the actual number of departures under the current matching strategy and the expected number of departures based on historical data is within the normal fluctuation range, the current retention rate has not decreased significantly, the system determines that the matching strategy is qualified, and maintains the current matching screening parameters and label mapping relationship unchanged. When the overall retention deviation is greater than the preset deviation threshold, it indicates that the actual number of departures under the current matching strategy is significantly higher than the expected value, the current retention rate has decreased abnormally, the system determines that the matching strategy is unqualified, and triggers the deviation cause location and strategy adjustment operation in subsequent step S6. The specific value of the preset deviation threshold can be adaptively adjusted according to the system's running rounds: in the early stage of system operation, due to the small sample size, a larger threshold is used to reduce the false positive rate; as the running rounds increase and the sample size accumulates, the threshold is gradually narrowed to improve sensitivity.

[0038] This embodiment sets a detection cycle to statistically analyze the actual number of new hires and departures in the current and previous preset time intervals. It then calculates the expected number of departures by combining parameters such as the screening ratio, hiring ratio, and demand quantity from the previous preset time interval. The difference between the actual and expected departures is used as the comprehensive retention deviation, thus achieving a quantitative evaluation of the matching strategy's retention effect. Simultaneously, by comparing the comprehensive retention deviation with a preset deviation threshold determined based on the standard deviation of historical deviation samples, it determines whether the matching strategy is qualified. If unqualified, it triggers subsequent deviation location and strategy adjustment. This provides an objective and quantitative triggering mechanism for the closed-loop adaptive optimization of the matching strategy, avoiding the uncertainty caused by subjective judgment. It ensures that the matching strategy can be promptly identified and the correction process triggered when the retention rate abnormally declines, effectively improving the long-term stability and accuracy of matching financial and tax professionals with business needs.

[0039] Specifically, in step S6, the reasons for the positioning deviation include: Obtain the overall retention rate for each of several preset historical time intervals; Calculate the variance of each overall retention rate; When the variance value is less than or equal to the preset variance threshold, the deviation is determined to be the first type of cause, which is that the matching and screening parameters deviate from the standard value. When the variance value is greater than the preset variance threshold, the deviation is determined to be a second type of cause, which is that the mapping relationship between the demand information and the screening benchmark label deviates from the standard value.

[0040] In this embodiment, the comprehensive retention rate of each of several historical preset time intervals is obtained; the variance of each comprehensive retention rate is calculated; when the variance is less than or equal to a preset variance threshold, the deviation is determined to be a first type of cause, which is that the matching screening parameters deviate from the standard value; when the variance is greater than the preset variance threshold, the deviation is determined to be a second type of cause, which is that the mapping relationship between the demand information and the screening benchmark label deviates from the standard value.

[0041] In this embodiment, the overall retention rate is the ratio of the actual number of retained employees to the actual number of new hires among the reviewed candidates within each historical preset time interval. It is used to characterize the long-term fit of the matching and screening results within each historical interval. In response to the determination of an unqualified matching strategy in step S5, the system extracts the overall retention rates of several historical preset time intervals prior to the current preset time interval from the historical database, constructing a time-series sample set of retention rates. The number of historical preset time intervals is determined based on the total system runtime; specifically, six to twelve preset time intervals prior to the current time point can be selected to ensure that the sample size is sufficient to support the statistical significance of variance calculation.

[0042] The preset variance threshold is a critical value used to distinguish whether the fluctuation of the retention rate is within the normal range. It is determined as follows: The system obtains the industry benchmark retention rate fluctuation range, and determines the benchmark variance reference value based on the difference between the upper and lower limits of the industry benchmark retention rate fluctuation range; the system obtains the number of matched tasks and the number of new hires corresponding to each historical preset time interval, determines the confidence weight of each time interval based on the product of the number of matched tasks and the number of new hires, and performs a weighted correction on the benchmark variance reference value based on each confidence weight, using the corrected value as the preset variance threshold. When historical data is insufficient to support variance calculation, the preset variance threshold uses an initial setting value, which is determined based on the experienced fluctuation range of talent turnover in the financial and tax services industry.

[0043] The system calculates the variance of each overall retention rate and compares it with a preset variance threshold. When the variance is less than or equal to the preset variance threshold, it indicates that the overall retention rate is relatively stable over several historical preset time intervals, and the retention rate has not fluctuated significantly. In this case, the reason for the deviation is not the instability of the requirement analysis and tag mapping process, but the matching and screening process itself, i.e., the first type of reason: the matching and screening parameters deviate from the standard value. Specifically, when the retention rate is stable but consistently low, it indicates that the screening benchmark tags themselves can accurately reflect the project requirements, but the screening parameters (such as the matching degree threshold) in the matching and screening process are improperly set. This results in the selected candidates meeting the tag matching degree but lacking actual long-term suitability. In other words, the screening criteria are too lenient or too strict, causing the number of candidates entering the review stage to be inconsistent with the actual job requirements, thereby affecting the long-term retention performance of the final hired personnel.

[0044] When the variance value exceeds a preset variance threshold, it indicates that the overall retention rate fluctuates significantly within several historical preset time intervals, showing an unstable trend. In this case, the deviation is not due to fixed parameter settings in the matching and screening process, but rather to the requirements analysis and tag mapping process—the second type of reason: the mapping relationship between requirements information and screening benchmark tags deviates from the standard value. Specifically, when the retention rate fluctuates significantly, it indicates that the mapping rules for extracting screening benchmark tags from project requirements information within each preset time interval are inconsistent or have deviations. This leads to an unstable correspondence between the screening benchmark tags used in different batches of matching and the actual project requirements. Tags in some batches fail to accurately reflect the core financial and tax business requirements of the project, causing the subsequent tag-based matching, screening, and onboarding process to be based on tags that deviate from actual needs. Ultimately, the long-term suitability of onboarded personnel exhibits random fluctuations. Based on the determination results of the above two types of reasons, the system selects the corresponding strategy adjustment operation to execute, precisely correcting the defects in the matching strategy.

[0045] This embodiment obtains the overall retention rate of several historical preset time intervals and calculates its variance value. The variance value is compared with a preset variance threshold. When the variance value is less than or equal to the preset variance threshold, it is determined to be the first type of reason for the deviation of the matching screening parameters from the standard value. When the variance value is greater than the preset variance threshold, it is determined to be the second type of reason for the deviation of the mapping relationship between the demand information and the screening benchmark label from the standard value. This achieves accurate tracing of the root cause of the matching strategy defect, provides a clear direction for subsequent targeted strategy adjustments, avoids the problem of system instability caused by blindly adjusting all parameters, and effectively improves the efficiency and accuracy of matching strategy optimization.

[0046] Specifically, when the cause of the deviation is determined to be the first type of cause, the corresponding strategy adjustment operation is performed according to the cause of the deviation, including: Obtain the ratio of the actual number of employees who passed the review within each historical preset time interval to the industry average number of employees during the same period, and use it as the specific ratio corresponding to each time interval. A specific ratio curve is constructed with the preset time interval as the horizontal axis and the specific ratio value as the vertical axis; Calculate the slope of the specific ratio curve at the current node, and denot it as the specific slope; The screening parameters are adjusted according to the specific slope, and the specific slope is positively correlated with the adjustment range of the screening parameters.

[0047] In this embodiment, the ratio of the actual number of employees who passed the review within each historical preset time interval to the industry average number of employees during the same period is obtained as the specific ratio corresponding to each time interval; a specific ratio curve is constructed with the preset time interval as the horizontal axis and the specific ratio as the vertical axis; the slope of the specific ratio curve at the current node is calculated and recorded as the specific slope; the screening parameters are adjusted according to the specific slope, and the specific slope is positively correlated with the adjustment range of the screening parameters.

[0048] In this embodiment, the specific ratio is the ratio of the actual number of employees hired by the system within each historical preset time interval to the industry average number of employees hired during the same period. This ratio characterizes the degree of deviation of the system's employee numbers from the overall industry level. When the deviation is determined to be due to a first-category cause, the system retrieves the actual number of employees hired by the system for each of the historical preset time intervals prior to the current time from the historical database. Simultaneously, it retrieves the industry average number of employees hired during the same period within each historical preset time interval from external data sources. The system calculates the ratio of the actual number of employees hired by the system to the industry average number of employees hired during the same period within each historical preset time interval, and uses this ratio as the specific ratio for each time interval. The industry average number of employees hired during the same period is a statistical value representing the average number of employees recruited by companies within the industry during that preset time interval, which can be obtained by accessing an industry data platform or a third-party human resources data service.

[0049] The system plots the specific ratio data points corresponding to each historical preset time interval in a two-dimensional coordinate system, using a preset time interval as the horizontal axis and the specific ratio as the vertical axis. A curve fitting method is used to fit each data point, constructing a continuous specific ratio curve. The curve fitting method can employ polynomial fitting or spline interpolation to reflect the trend of the specific ratio changing over time. The system calculates the derivative of the specific ratio curve at the current node, using the derivative as the slope value at the current node, denoted as the specific slope. The current node is the position on the specific ratio curve corresponding to the current preset time interval. A positive specific slope indicates an upward trend in the specific ratio, meaning the number of new hires in the system is increasing relative to the industry average; a negative specific slope indicates a downward trend in the specific ratio, meaning the number of new hires in the system is decreasing relative to the industry average; the larger the absolute value of the specific slope, the faster the rate of change of the specific ratio.

[0050] The system adjusts the screening parameters in the matching and screening process based on the specific slope. When the specific slope is positive, it indicates that the number of new hires in the system is increasing relative to the industry average. In this case, the screening rigor needs to be increased, i.e., the adjustment range of the screening parameters needs to be increased to make the screening criteria stricter. When the specific slope is negative, it indicates that the number of new hires in the system is decreasing relative to the industry average. In this case, the screening rigor needs to be decreased, i.e., the adjustment range of the screening parameters needs to be decreased to make the screening criteria more lenient. The screening parameter is the matching degree threshold used in step S3 to determine whether a candidate passes the initial screening. The specific slope is positively correlated with the adjustment range of the screening parameter; that is, the larger the absolute value of the specific slope, the larger the adjustment range of the screening parameter. When the specific slope is positive, the screening parameter is adjusted upwards; when the specific slope is negative, the screening parameter is adjusted downwards.

[0051] The specific adjustment method of the screening parameters is as follows: the absolute value of the specific slope is multiplied by a preset adjustment coefficient to obtain the adjustment step size of the screening parameters; when the specific slope is positive, the adjustment step size of the current screening parameters is increased; when the specific slope is negative, the adjustment step size of the current screening parameters is decreased. The preset adjustment coefficient is set according to the system's requirements for the sensitivity of the screening parameter adjustment to balance the system's response speed and stability. The adjustment coefficient can also be adaptively corrected based on historical adjustment effects: if the overall retention deviation decreases after the previous round of adjustment, the adjustment coefficient is maintained or increased; if the overall retention deviation does not decrease or further increases after the previous round of adjustment, the adjustment coefficient is decreased.

[0052] Specifically, before adjusting the screening parameters according to the specific slope, the process also includes: Calculate the ratio of the actual number of employees hired in the current preset time interval to the actual number of employees hired in the previous preset time interval, and record it as the employee hiring change ratio. When the change in the number of new hires is greater than one, the adjustment range of the screening parameter is increased according to the change in the number of new hires, and the change in the number of new hires is positively correlated with the adjustment range. When the change in the number of hires is less than or equal to one, the adjustment range of the screening parameter is reduced according to the change in the number of hires, and the change in the number of hires is negatively correlated with the adjustment range.

[0053] In this embodiment, the ratio of the actual number of employees hired in the current preset time interval to the actual number of employees hired in the previous preset time interval is calculated and denoted as the hiring change ratio. When the hiring change ratio is greater than one, the adjustment range of the screening parameter is increased according to the hiring change ratio, and the hiring change ratio is positively correlated with the adjustment range. When the hiring change ratio is less than or equal to one, the adjustment range of the screening parameter is decreased according to the hiring change ratio, and the hiring change ratio is negatively correlated with the adjustment range.

[0054] In this embodiment, the employee recruitment change ratio is the ratio of the actual number of employees recruited in the current preset time interval to the actual number of employees recruited in the previous preset time interval, used to characterize the temporal trend of employee recruitment. After determining the basic adjustment range of the screening parameters based on the specific slope, the system further calculates the employee recruitment change ratio to correct the adjustment range, so that the adjustment amount adapts to the changing trend of actual employee recruitment. When the ratio of changes in hiring is greater than one, it indicates that the actual number of hires in the current preset time interval has increased compared to the previous preset time interval, and the hiring scale is expanding. In this case, the system increases the adjustment range of the screening parameters based on the ratio of changes in hiring, and the ratio of changes in hiring is positively correlated with the adjustment range, that is, the faster the number of hires increases, the larger the adjustment range of the screening parameters, so as to avoid the risk of quality decline due to the increase in the number of hires when the hiring scale expands. When the ratio of changes in hiring is less than or equal to one, it indicates that the actual number of hires in the current preset time interval has not increased or has even decreased compared to the previous preset time interval, and the hiring scale is contracting or flat. In this case, the system decreases the adjustment range of the screening parameters based on the ratio of changes in hiring, and the ratio of changes in hiring is negatively correlated with the adjustment range, that is, the greater the reduction in the number of hires, the greater the reduction in the adjustment range, so as to reduce the strictness of the screening standards when the hiring scale contracts, ensuring that a sufficient number of candidates pass the screening to meet the job requirements.

[0055] The specific adjustment method is as follows: the system uses the basic adjustment step size determined by the specific slope as a benchmark, and multiplies the basic adjustment step size by the product of the ratio of new hires and the preset correction coefficient to obtain the amplitude correction amount; when the ratio of new hires is greater than one, the amplitude correction amount is added to the basic adjustment step size to obtain the final adjustment step size; when the ratio of new hires is less than or equal to one, the amplitude correction amount is subtracted from the basic adjustment step size to obtain the final adjustment step size.

[0056] This embodiment constructs a unique ratio curve by obtaining the ratio of the actual number of new hires to the industry average number of new hires during each historical preset time interval, and calculates the unique slope of the current node. Based on the sign and absolute value of the unique slope, the screening parameters are adjusted in a directional and quantitative manner, achieving dynamic adaptation of the screening criteria to the industry's talent supply and demand situation. At the same time, the adjustment range is corrected a second time by calculating the ratio of changes in new hires, so that the adjustment amount of the screening parameters reflects information from two dimensions: the rate of change of the unique ratio and the trend of changes in the number of new hires. This avoids over-adjustment or under-adjustment caused by abnormal fluctuations in the number of new hires, and improves the accuracy and robustness of the screening parameter adjustment.

[0057] Specifically, after adjusting the screening parameters according to the specific slope, the process also includes: Based on the adjusted screening parameters, the operation of matching candidate information based on the screening benchmark label, calculating the overall retention deviation, and determining whether the matching strategy is qualified is re-executed. If the matching strategy is still unqualified after re-evaluation, the deviation reason will be corrected to the second type of reason, and the strategy adjustment operation corresponding to the second type of reason will be executed.

[0058] In this embodiment, the operation of matching candidate object information based on the screening benchmark label, calculating the comprehensive retention deviation, and judging whether the matching strategy is qualified is re-executed based on the adjusted screening parameters; if the matching strategy is still unqualified after re-judgment, the deviation reason is corrected to the second type of reason, and the strategy adjustment operation corresponding to the second type of reason is executed.

[0059] In this embodiment, after adjusting the screening parameters, the system uses the adjusted parameters as the new screening criteria and re-executes the operation of matching and screening candidate information from the knowledge base based on the screening benchmark tags. After a preset system running time, the system recalculates the overall retention deviation based on the actual execution data of newly hired and approved candidates. The recalculated overall retention deviation is compared with a preset deviation threshold to determine whether the matching strategy is qualified. This re-execution process is consistent with the operations in steps S3 to S5, that is, re-executes the complete process of matching and screening, review confirmation, onboarding data collection, and retention deviation calculation. If the matching strategy is qualified after re-determination, it indicates that the adjustment of the screening parameters for the first type of reason is effective. The system maintains the adjusted screening parameters, the current deviation processing process ends, and the system waits for the quality assessment in the next preset time interval.

[0060] If the matching strategy still fails after reassessment, it indicates that the deviation is not simply due to a problem with the filtering parameters in the matching process. Specifically, after ruling out the possibility that the filtering parameters deviate from the standard values, the continued failure of the matching strategy suggests that the root cause of the deviation lies in the upstream process, namely, a problem with the mapping rules for extracting and determining the filtering benchmark labels from the project requirement information. In this case, the system corrects the deviation cause to the second type: the mapping relationship between the requirement information and the filtering benchmark labels deviates from the standard values, triggering the strategy adjustment operation corresponding to the second type of cause.

[0061] Specifically, when the cause of the deviation is determined to be the second type of cause, the corresponding strategy adjustment operation is performed according to the cause of the deviation, including: Obtain the length of employment for each approved employee from their start date to their departure date; Calculate the ratio of each working duration to the detection cycle, which is used as the individual working percentage for each object; Calculate the average percentage of employed individuals among all those who passed the review, and use this as the overall employed percentage. Based on the mapping relationship between the comprehensive on-the-job ratio adjustment demand information and candidate labels, the mapping relationship is used to control the correlation strength between demand information and each candidate label.

[0062] Specifically, based on the mapping parameters between the comprehensive on-the-job ratio adjustment demand information and candidate tags, the following parameters are included: The ratio of the number of times each candidate tag appears in a historical preset time interval to the total number of times in the preset time interval is obtained, which is used as the comprehensive frequency of each candidate tag; For candidate tags whose overall frequency is higher than a preset frequency threshold, the adjustment range of their corresponding mapping relationship is reduced. For candidate tags whose overall frequency is lower than or equal to the preset frequency threshold, the adjustment range of their corresponding mapping relationship is maintained.

[0063] In this embodiment, the employment duration of each approved candidate from their onboarding to their departure is obtained; the ratio of each employment duration to the detection period is calculated as the individual employment percentage for each candidate; the average of the individual employment percentages of all approved candidates is calculated as the comprehensive employment percentage; the mapping relationship between demand information and candidate tags is adjusted based on the comprehensive employment percentage, and this mapping relationship is used to control the correlation strength between demand information and each candidate tag. In this embodiment, when the deviation is determined to be a second type of cause, it indicates that the mapping relationship between demand information and screening benchmark tags deviates from the standard value, resulting in an unstable correspondence between the screening benchmark tags used for batch matching and the actual project demand. At this time, the system needs to adjust the mapping relationship, and the adjustment is based on the long-term adaptation level reflected by the actual employment duration of each approved candidate.

[0064] The system retrieves the actual employment duration from the start date to the end date for each approved employee within the current preset time interval from the human resources database. For employees who have not yet left, the employment duration is the length from the start date to the current statistical cutoff date. The system obtains a preset detection period, which is a pre-defined time window length used to measure retention. The system divides the employment duration of each approved employee by the detection period to obtain the individual employment percentage for each employee. The individual employment percentage is used to characterize the actual retention level of each employee within the complete detection period: when the individual employment percentage is greater than or equal to one, it indicates that the employee's employment duration has reached or exceeded the complete detection period, and the retention is good; when the individual employment percentage is less than one, it indicates that the employee left before the end of the detection period, and the retention is insufficient.

[0065] The system calculates the arithmetic mean of the individual employment rates of all approved applicants, and uses this arithmetic mean as the overall employment rate. The overall employment rate is used to characterize the overall retention level of all new hires within the current preset time interval. A higher overall employment rate indicates better overall retention, while a lower overall employment rate indicates worse overall retention.

[0066] The system adjusts the mapping relationship between demand information and candidate tags based on the overall employment ratio. This mapping relationship controls the association strength between each semantic unit in the demand information and each candidate tag, i.e., determining which tags are extracted from the project demand information and what weights are assigned to each tag. The system presets an initial mapping relationship template, which contains the standard correspondence and standard association weights between various semantic units in the demand information and candidate tags. When the overall employment ratio is lower than the preset retention benchmark value, it indicates that the matching degree between the extracted screening benchmark tags under the current mapping relationship and the actual project demand is insufficient, resulting in low long-term suitability of the recruits. The system increases the adjustment range of the mapping relationship, i.e., increases the upward or downward adjustment of the corresponding parameters of the mapping relationship to strengthen the correction of the mapping relationship. When the overall employment ratio is higher than or equal to the preset retention benchmark value, it indicates that the long-term suitability under the current mapping relationship is good. The system decreases the adjustment range of the mapping relationship to maintain or fine-tune the current mapping relationship. The preset retention benchmark value is determined based on the industry average retention rate or the average of the overall retention rate over a historical preset time interval.

[0067] In this embodiment, the ratio of the number of times each candidate tag appears in a historical preset time interval to the total number of preset time intervals is obtained as the comprehensive frequency of each candidate tag. For candidate tags whose comprehensive frequency is higher than a preset frequency threshold, the adjustment range of their corresponding mapping relationship is reduced; for candidate tags whose comprehensive frequency is lower than or equal to the preset frequency threshold, the adjustment range of their corresponding mapping relationship is maintained. After determining the overall adjustment range of the mapping relationship based on the comprehensive on-the-job ratio, the system further obtains the number of times each candidate tag appears in a historical preset time interval, calculates the ratio of the number of times each candidate tag appears to the total number of preset time intervals, and uses the ratio as the comprehensive frequency of each candidate tag. The comprehensive frequency is used to characterize the universality and stability of each candidate tag in historical operation: the higher the comprehensive frequency, the more frequently the tag appears in the demand information in each preset time interval, belonging to the high-frequency general tag in financial and tax business, with clear semantics, stable scope of application, and corresponding mapping relationship has been verified in multiple rounds, and should not be significantly adjusted; the lower the comprehensive frequency, the more frequently the tag belongs to the low-frequency or special tag in specific scenarios, and its mapping relationship has greater uncertainty, requiring greater adjustment space to adapt to the differences in specific project needs.

[0068] The system has a preset frequency threshold, which is determined as follows: The overall frequency distribution of all candidate tags is obtained, the median or average of the overall frequency distribution is calculated, and the median or average is used as the preset frequency threshold. The system compares the overall frequency of each candidate tag with the preset frequency threshold. For candidate tags with an overall frequency higher than the preset frequency threshold, the system reduces the adjustment range of their corresponding mapping relationship; that is, the association weight of high-frequency tags changes less, in order to maintain the verified stable mapping relationship. For candidate tags with an overall frequency lower than or equal to the preset frequency threshold, the system maintains the adjustment range of their corresponding mapping relationship; that is, low-frequency tags are adjusted normally according to the overall adjustment range determined by the overall employment ratio, in order to fully utilize the current retention feedback data to correct their mapping relationship.

[0069] This embodiment calculates the individual employment rate by obtaining the ratio of the employment duration to the detection cycle of each reviewed object, and further calculates the comprehensive employment rate. The comprehensive employment rate is used as a quantitative feedback signal to adjust the mapping relationship between demand information and candidate labels, realizing a closed-loop linkage between the label mapping relationship and the actual long-term adaptation effect. At the same time, by obtaining the comprehensive frequency of each candidate label in the historical preset time interval, the adjustment range of the mapping relationship of candidate labels with a comprehensive frequency higher than a preset frequency threshold is reduced, and the adjustment range of the mapping relationship of candidate labels with a comprehensive frequency lower than or equal to the preset frequency threshold is maintained. This realizes differentiated and fine control by adopting a conservative adjustment strategy for high-frequency stable labels to maintain system stability and an active adjustment strategy for low-frequency labels to quickly correct deviations. While ensuring the effectiveness of the mapping relationship adjustment, it avoids system oscillation caused by over-adjustment.

[0070] Specifically, after the cause of the deviation is corrected to the second type of cause, the strategy adjustment operation corresponding to the second type of cause is executed, including: Obtain the overall frequency of each candidate tag within a preset historical time interval; The weight correction coefficient for each candidate label is determined based on the comprehensive frequency, and the weight correction coefficient for a candidate label with a higher comprehensive frequency is smaller. The mapping relationship between the demand information and each candidate label is updated based on the weight correction coefficient.

[0071] In this embodiment, if the matching strategy still fails after adjusting the screening parameters for the first type of reason and re-executing the matching screening and quality assessment, the system corrects the deviation reason to the second type of reason and executes the corresponding strategy adjustment operation. At this time, since a complete matching screening and assessment cycle has already been completed, the system has accumulated more sufficient actual execution data within the current preset time interval. Therefore, a global update of the mapping relationship between the demand information and each candidate label is needed, rather than simply adjusting the magnitude based on the overall on-the-job percentage. The system obtains the number of times each candidate label appears in the historical preset time interval from the historical database, calculates the ratio of the number of times each candidate label appears to the total number in the preset time interval, and uses this ratio as the overall frequency of each candidate label.

[0072] The system determines the weight correction coefficient for each candidate label based on its overall frequency. This weight correction coefficient characterizes the magnitude of weight adjustment for each candidate label during mapping relationship updates: candidate labels with higher overall frequencies have appeared more frequently in historical operations, indicating stronger mapping stability and correspondingly smaller weight correction coefficients (i.e., smaller update magnitude); conversely, candidate labels with lower overall frequencies have appeared less frequently in historical operations, indicating greater uncertainty in the mapping relationship and correspondingly larger weight correction coefficients (i.e., larger update magnitude). The specific method for determining the weight correction coefficient is as follows: the system uses the reciprocal of the overall frequency as a baseline value, normalizes the baseline values ​​of each candidate label, and ensures that the sum of the weight correction coefficients of all candidate labels equals a preset constant. The normalized value is then used as the weight correction coefficient for each candidate label. This preset constant is set according to the system's requirements for the global update intensity of the mapping relationship.

[0073] The system updates the mapping relationship between demand information and each candidate label based on the weight correction coefficient. Specifically, the system obtains the actual employment feedback data corresponding to each candidate label within the current preset time interval, multiplies the actual employment feedback data of each candidate label by the weight correction coefficient, and uses this product as the mapping relationship update amount for each candidate label. This update amount is then added to the mapping relationship parameters corresponding to each candidate label to generate the updated mapping relationship. For candidate labels with high overall frequency, due to the small weight correction coefficient, the update amount of their mapping relationship parameters is small, maintaining a stable mapping based on historical verification. For candidate labels with low overall frequency, due to the large weight correction coefficient, the update amount of their mapping relationship parameters is large, allowing for rapid correction of unstable mapping relationships by fully utilizing current feedback data.

[0074] This embodiment corrects the deviation cause to a second cause after the first type of cause adjustment fails and performs a global update of the mapping relationship. It obtains the comprehensive frequency of each candidate label and determines the weight correction coefficient based on the comprehensive frequency. The weight correction coefficient corresponding to the candidate label with the higher comprehensive frequency is smaller, and the weight correction coefficient corresponding to the candidate label with the lower comprehensive frequency is larger. Then, based on the weight correction coefficient, the actual on-the-job feedback data of each candidate label is superimposed on its current mapping relationship parameters to generate an updated mapping relationship. This realizes a differentiated mapping relationship update mechanism that adopts conservative updates for high-frequency stable labels to maintain system stability and active updates for low-frequency uncertain labels to make full use of the current feedback data for rapid correction. It fundamentally solves the problem of the mapping relationship between demand information and screening benchmark labels deviating from the standard value, and ensures the accuracy and adaptability of the label extraction process.

[0075] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for intelligent matching and screening of professional talents based on the needs of financial and tax business, characterized in that, include: Step S1: Build and continuously update the knowledge base based on historical operational data; Step S2: Obtain the input project requirement information, extract candidate tags from the project requirement information, and determine the candidate tags that meet the preset tag criteria as the screening benchmark tags. Step S3: Based on the filtering benchmark tags, match and filter candidate object information from the knowledge base; Step S4: Record the candidate information obtained after screening, review and confirm the screened candidate, and record the information of the candidates that pass the review. Step S5: After the system has run for a preset period of time, calculate the comprehensive retention deviation based on the actual execution data of the objects that have passed the review, and determine whether the matching strategy is qualified. Step S6: When the matching strategy is determined to be unqualified, locate the cause of the deviation and perform the corresponding strategy adjustment operation according to the cause of the deviation.

2. The intelligent matching and screening method for professional talents based on financial and tax business needs as described in claim 1, characterized in that, In step S5, the overall retention deviation is calculated based on the actual execution data of the reviewed objects, including: Set a testing cycle and count the actual number of new hires and actual departures of employees who passed the review within the current preset time interval; Get the actual number of new hires and actual number of departures within the previous preset time interval; Based on the actual number of new hires and actual number of departures in the previous preset time interval, and combined with at least one of the following parameters from the previous preset time interval: screening ratio, hiring ratio, and number of received demands, calculate the expected number of departures in the current preset time interval. The overall retention deviation is determined based on the difference between the actual number of departures and the expected number of departures within the current preset time interval.

3. The intelligent matching and screening method for professional talents based on financial and tax business needs as described in claim 2, characterized in that, In step S5, determining whether the matching strategy is qualified includes: If the overall retention deviation is less than or equal to the preset deviation threshold, the matching strategy is deemed qualified. If the overall retention deviation is greater than the preset deviation threshold, the matching strategy is deemed unqualified.

4. The intelligent matching and screening method for professional talents based on financial and tax business needs as described in claim 3, characterized in that, In step S6, the reasons for the positioning deviation include: Obtain the overall retention rate for each of several preset historical time intervals; Calculate the variance of each overall retention rate; When the variance value is less than or equal to the preset variance threshold, the deviation is determined to be the first type of cause, which is that the matching and screening parameters deviate from the standard value. When the variance value is greater than the preset variance threshold, the deviation is determined to be a second type of cause, which is that the mapping relationship between the demand information and the screening benchmark label deviates from the standard value.

5. The intelligent matching and screening method for professional talents based on financial and tax business needs as described in claim 4, characterized in that, When the cause of the deviation is determined to be the first type of cause, the corresponding strategy adjustment operation is executed according to the cause of the deviation, including: Obtain the ratio of the actual number of employees who passed the review within each historical preset time interval to the industry average number of employees during the same period, and use it as the specific ratio corresponding to each time interval. A specific ratio curve is constructed with the preset time interval as the horizontal axis and the specific ratio value as the vertical axis; Calculate the slope of the specific ratio curve at the current node, and denot it as the specific slope; The screening parameters are adjusted according to the specific slope, and the specific slope is positively correlated with the adjustment range of the screening parameters.

6. The intelligent matching and screening method for professional talents based on financial and tax business needs as described in claim 5, characterized in that, Before adjusting the screening parameters according to the specific slope, the process also includes: Calculate the ratio of the actual number of employees hired in the current preset time interval to the actual number of employees hired in the previous preset time interval, and record it as the employee hiring change ratio. When the change in the number of new hires is greater than one, the adjustment range of the screening parameter is increased according to the change in the number of new hires, and the change in the number of new hires is positively correlated with the adjustment range. When the change in the number of new hires is less than or equal to one, the adjustment range of the screening parameter is reduced according to the change in the number of new hires, and the change in the number of new hires is negatively correlated with the adjustment range.

7. The intelligent matching and screening method for professional talents based on financial and tax business needs as described in claim 6, characterized in that, After adjusting the screening parameters according to the specific slope, the process also includes: Based on the adjusted screening parameters, the operation of matching candidate information based on the screening benchmark label, calculating the overall retention deviation, and determining whether the matching strategy is qualified is re-executed. If the matching strategy is still unqualified after re-evaluation, the reason for the deviation will be corrected to the second type of reason, and the strategy adjustment operation corresponding to the second type of reason will be executed.

8. The intelligent matching and screening method for professional talents based on financial and tax business needs as described in claim 4, characterized in that, When the cause of the deviation is determined to be the second type of cause, the corresponding strategy adjustment operation is executed according to the cause of the deviation, including: Obtain the length of employment for each approved employee from their start date to their departure date; Calculate the ratio of each working duration to the detection cycle, which is used as the individual working percentage for each object; Calculate the average percentage of employed individuals among all those who passed the review, and use this as the overall employed percentage. Based on the mapping relationship between the comprehensive on-the-job ratio adjustment demand information and candidate labels, the mapping relationship is used to control the correlation strength between demand information and each candidate label.

9. The intelligent matching and screening method for professional talents based on financial and tax business needs as described in claim 8, characterized in that, The mapping parameters between the comprehensive on-the-job ratio adjustment demand information and candidate tags include: The ratio of the number of times each candidate tag appears in a historical preset time interval to the total number of times in the preset time interval is obtained, which is used as the comprehensive frequency of each candidate tag; For candidate tags whose overall frequency is higher than a preset frequency threshold, the adjustment range of their corresponding mapping relationship is reduced. For candidate tags whose overall frequency is lower than or equal to the preset frequency threshold, the adjustment range of their corresponding mapping relationship is maintained.

10. The intelligent matching and screening method for professional talents based on financial and tax business needs as described in claim 7, characterized in that, Once the cause of the deviation is corrected to the second type of cause, the strategy adjustment operation corresponding to the second type of cause is executed, including: Obtain the overall frequency of each candidate tag within a preset historical time interval; The weight correction coefficient for each candidate label is determined based on the comprehensive frequency, and the weight correction coefficient for a candidate label with a higher comprehensive frequency is smaller. The mapping relationship between the demand information and each candidate label is updated based on the weight correction coefficient.

Images