Artificial intelligence-based talent recruitment promotion effect evaluation system

Abstract

The invention discloses a talent recruitment promotion effect evaluation system based on artificial intelligence. The system comprises a promotion effect evaluation data acquisition module, a data optimization module, a recruitment promotion effect stage evaluation module, a stage evaluation dynamic weight calculation module and a recruitment promotion effect comprehensive evaluation module. The invention relates to the technical field of information data processing, in particular to a talent recruitment promotion effect evaluation system based on artificial intelligence, which innovatively proposes a mode of combining instant evaluation and long-term prediction evaluation to shorten the recruitment promotion effect evaluation period. A multi-dimensional quantitative index, a dynamic target value mechanism and a cross-stage feature fusion mechanism are innovatively proposed, and the accuracy of evaluation results of all recruitment promotion stages is effectively improved; the method for dynamically adjusting the weight based on the clustering analysis and optimization algorithm is innovatively provided, the adaptability of the evaluation process is improved, the limitation of static weight setting is eliminated, and the accuracy of the comprehensive evaluation result of the recruitment promotion effect is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of information data processing, and particularly to a talent recruitment promotion effect evaluation system based on artificial intelligence. Background Art

[0002] The talent recruitment promotion effect evaluation system based on artificial intelligence collects and analyzes multi-source data generated during the large-scale recruitment promotion process, combines artificial intelligence technology, quantitatively evaluates the effects achieved by talent recruitment promotion, and further provides data-driven decision-making basis for the adjustment and optimization of enterprise recruitment strategies to maximize the talent recruitment promotion effect; however, there are technical problems in the traditional talent recruitment promotion effect evaluation system that it is difficult to present the evaluation results of promotion effects in real time, resulting in a fault phenomenon in the promotion effect evaluation cycle, thereby making it impossible to effectively comprehensively evaluate the entire process of talent recruitment promotion effects; there are technical problems in the traditional talent recruitment promotion effect evaluation system that the evaluation results in the immediate evaluation stage of promotion effects cannot achieve quantitative and accurate evaluation and the accuracy of the evaluation results in the long-term prediction evaluation stage of promotion effects is insufficient, resulting in inaccurate evaluation results in each recruitment promotion stage; the traditional talent recruitment promotion effect evaluation system has the technical problem of using static weight allocation, ignoring the dynamic changes of multi-dimensional factors in the recruitment promotion scenario, resulting in the weight setting deviating from the actual situation, thus making the evaluation results of talent recruitment promotion effects inaccurate. Summary of the Invention

[0003] In view of the above situation, to overcome the defects of the prior art, the present invention provides a talent recruitment promotion effect evaluation system based on artificial intelligence. Aiming at the technical problem that it is difficult to present the evaluation results of the promotion effect in a timely manner in the traditional talent recruitment promotion effect evaluation system, which leads to a fault in the promotion effect evaluation cycle and further affects the comprehensive evaluation effect of the entire recruitment process. This solution innovatively proposes to finely divide the recruitment promotion effect evaluation stage into six stages, and adopt a combination of immediate evaluation and long-term prediction evaluation, effectively eliminating the problem of the evaluation cycle fault existing in the traditional method, ensuring the real-time and continuity of the evaluation results at each stage of the recruitment activity, significantly shortening the recruitment promotion effect evaluation cycle, reducing the promotion effect evaluation from monthly to daily, comprehensively covering the promotion effect evaluation of the entire talent recruitment process, and significantly improving the real-time and accuracy of the recruitment promotion effect evaluation; aiming at the technical problems that the evaluation results in the immediate evaluation stage of the promotion effect in the traditional talent recruitment promotion effect evaluation system cannot be quantitatively and accurately evaluated and the accuracy of the evaluation results in the long-term prediction evaluation stage of the promotion effect is insufficient, resulting in inaccurate evaluation results at each recruitment promotion stage. This solution innovatively proposes that in the immediate evaluation stage of the promotion effect, through multi-dimensional quantitative indicators and dynamic target value mechanisms, it can reflect the actual recruitment situation in real time, provide a more realistic reference standard for the evaluation results in the immediate evaluation stage, and enhance the accuracy and practicability of the quantification of the evaluation results; in the long-term prediction evaluation stage of the promotion effect, through cross-stage feature fusion mechanisms, establish the internal dependence relationship between the employment and probationary period stages, comprehensively integrate long-term data features, and significantly improve the prediction accuracy; it can effectively improve the accuracy of the evaluation results at each recruitment promotion stage; aiming at the technical problem that the traditional talent recruitment promotion effect evaluation system uses static weight allocation, ignoring the dynamic changes of multi-dimensional factors in the recruitment promotion scenario, resulting in the weight setting deviating from the actual situation, and thus making the evaluation results of the talent recruitment promotion effect inaccurate. This solution innovatively proposes a method for dynamically adjusting weights based on clustering analysis and particle swarm optimization algorithm, flexibly adjusting the weight combination according to the actual situation of the recruitment promotion scenario, improving the adaptability and accuracy of the evaluation process, eliminating the limitations of static weight setting, and considering the dynamic changes of multi-dimensional factors, thus significantly improving the accuracy and operability of the recruitment promotion effect evaluation results.

[0004] The technical solution adopted by the present invention is as follows: The talent recruitment promotion effect evaluation system based on artificial intelligence provided by the present invention includes a promotion effect evaluation data acquisition module, a data optimization module, a recruitment promotion effect stage evaluation module, a stage evaluation dynamic weight calculation module, and a recruitment promotion effect comprehensive evaluation module;

[0005] The promotion effect evaluation data acquisition module specifically collects the original data for evaluating the recruitment promotion effect through the recruitment channel platform and the company's personnel system;

[0006] The data optimization module is used to optimize the original data for evaluating the recruitment promotion effect. First, through multi-dimensional evaluation data cleaning, invalid data is eliminated. Then, data standardization processing and data logical verification are carried out to ensure data logical consistency and business rationality. Finally, feature extraction for evaluating the promotion effect is implemented to obtain optimized data for evaluating the recruitment promotion effect;

[0007] The recruitment promotion effect stage evaluation module is specifically divided into six stages according to the natural nodes of the recruitment process, and these six stages are further combined into an immediate evaluation stage and a long-term prediction evaluation stage. The evaluation results of the talent recruitment recommendation effect in each stage of the immediate evaluation stage and the long-term prediction evaluation stage are calculated respectively, and the evaluation results of each stage are combined to obtain a set of recruitment promotion effect stage evaluation results;

[0008] The stage evaluation dynamic weight calculation module is specifically to obtain real-time talent recruitment promotion scenario labels through the K-means clustering algorithm, and find the recruitment promotion weight combination corresponding to this label in the historical recruitment promotion scenario data as the initial weight combination set. The particle swarm optimization algorithm is used to optimize on the basis of the initial weight combination set to obtain the optimal stage evaluation dynamic weight combination;

[0009] The recruitment promotion effect comprehensive evaluation module is specifically to combine the optimal stage evaluation dynamic weight combination with the set of recruitment promotion effect stage evaluation results through the weighted average method to obtain the comprehensive evaluation result.

[0010] Furthermore, the promotion effect evaluation data acquisition module is specifically to collect the original data for evaluating the recruitment promotion effect through the recruitment channel platform and the company's personnel system; the original data for evaluating the recruitment promotion effect includes historical promotion effect evaluation data, historical recruitment promotion scenario data, real-time recruitment promotion scenario data, and real-time promotion effect evaluation data; both the historical recruitment promotion effect evaluation data and the real-time recruitment promotion effect evaluation data include job exposure data, job click data, resume submission data, interview invitation data, employment situation data, probationary situation data, recruitment activity information data, and job information data; the historical recruitment promotion effect evaluation data also includes the evaluation scores for the talent recruitment and employment stage, the evaluation scores for the talent recruitment and probationary stage, and the historical comprehensive evaluation value of the talent recruitment promotion effect; both the historical recruitment promotion scenario data and the real-time recruitment promotion scenario data include recruitment activity information data and job information data; the historical recruitment promotion scenario data also includes historical talent recruitment promotion scenario labels and recruitment promotion weight combinations.

[0011] Further, the data optimization module is used to optimize the original data for evaluating the recruitment promotion effect, including multi-dimensional evaluation data cleaning, data standardization processing, data logical verification, and extraction of evaluation features for the promotion effect, to obtain optimized data for evaluating the recruitment promotion effect; specifically, it includes the following steps:

[0012] Multi-dimensional evaluation data cleaning is used to identify and process missing values, outliers, and duplicate records in the original data.

[0013] Data standardization processing is used to unify the data formats from different sources and structures. Specifically, it includes numerical data processing, categorical data processing, standardization of time fields, standardization of status fields, and structuring of text fields.

[0014] Data logical verification is used to ensure the logical consistency and business rationality between different data fields. Specifically, it uses preset business logic rules to check relationship constraints and eliminates illogical data.

[0015] Extraction of evaluation features for the promotion effect is used to screen out the core evaluation features related to the recruitment promotion effect from the data. Specifically, through statistical analysis methods, statistical feature selection is performed on the data to obtain the key evaluation features reflecting the recruitment effect.

[0016] Further, the phased evaluation module for the recruitment promotion effect is used to quantitatively evaluate the implementation effect of the recruitment promotion activity in different stages, including phased division of recruitment, immediate evaluation stage of the promotion effect, long-term prediction evaluation stage of the promotion effect, and output of the phased evaluation results of the recruitment promotion effect; specifically, it includes the following steps:

[0017] Phased division of recruitment is used to break down the entire recruitment process into quantifiable recruitment stages. Specifically, the recruitment stages are divided and defined according to the natural nodes of the recruitment process. The recruitment stages include the job exposure stage, job click stage, job application stage, job interview stage, employment stage, and probationary period completion stage; further, the job exposure stage, job click stage, job application stage, and job interview stage are divided into the immediate evaluation stage, and the employment stage and probationary period completion stage are divided into the long-term prediction evaluation stage.

[0018] The immediate evaluation stage of the promotion effect is used to quantitatively analyze the promotion effect of the recruitment promotion activity in the short term. Specifically, by calculating the ratio of the actual performance in each stage to the preset fixed benchmark value and combining the key influencing dimensions of each stage, the promotion effect stage score of each stage in the immediate evaluation stage is obtained; it includes the following steps:

[0019] Calculate the score for the job exposure stage to evaluate the effectiveness of the job exposure stage. Calculate the ratio of the actual exposure times to the fixed benchmark exposure times, and combine the job promotion display position and display period to obtain the score for the job exposure stage ;

[0020] Calculate the score for the job click stage to evaluate the effectiveness of the job click stage. Calculate the ratio of the actual click times to the fixed benchmark click times, and combine the dwell time at the time of click to calculate the promotion score for this stage to obtain the score for the job click stage ;

[0021] Calculate the score for the job application stage to evaluate the effectiveness of the job application stage. Calculate the ratio of the actual resume application volume to the fixed benchmark application volume, and combine the channel quality and candidate matching degree to calculate the promotion score for this stage to obtain the score for the job application stage ;

[0022] Calculate the score for the job interview stage to evaluate the effectiveness of the job interview stage. Calculate the ratio of the actual number of interview invitations to the fixed benchmark number of interviews, and combine the timeliness of the interview arrangement and the interview participation rate to calculate the promotion score for this stage to obtain the score for the job interview stage ;

[0023] Calculate the dynamic target value for the recruitment stage, which is used to adjust the target value of the talent recruitment promotion effect in real time. Specifically, build a dynamic target value calculation model through a GRU neural network model, use the historical promotion dynamic target value calculation data as training data to train the dynamic target value calculation model, and use the real-time promotion dynamic target value calculation data as input data to pass it to the trained dynamic target value calculation model to obtain the dynamic target values for each stage of the recruitment activity ; The formula used is as follows:

[0024] ;

[0025] ;

[0026] In the formula, represents the hidden state at the current moment, represents the GRU model running function, represents the input data at the current moment, represents the hidden state at the previous moment, represents the predicted value of the dynamic target value for the i-th stage, represents the weight matrix of the output layer, represents the bias term of the output layer, represents the Sigmoid activation function;

[0027] Calculate the evaluation results of the immediate evaluation stage. Specifically, combine the dynamic target value with the actual score of each stage in the immediate evaluation stage to obtain the evaluation results of the recruitment promotion effect of each stage in the immediate evaluation stage. The formula used is as follows:

[0028] ;

[0029] ;

[0030] In the formula, represents the evaluation result of the recruitment promotion effect in the i-th stage, represents the weight coefficient for balancing the historical benchmark and dynamic prediction, represents the actual score in the i-th stage, represents the historical highest score in the i-th stage, represents the fixed benchmark value in the i-th stage;

[0031] The long-term prediction evaluation stage for promotion effect is used to predict the long-term talent recruitment promotion effect. Specifically, by constructing a long-term prediction evaluation model, predict the long-term promotion effect of talent recruitment, which specifically includes the following steps:

[0032] Construct a long-term prediction evaluation model, which specifically includes the following steps:

[0033] Obtain the evaluation features of the promotion effect in the employment stage to capture the time-series dependence of the dynamic changes in the employment stage. Specifically, process the employment situation data through a long short-term memory network to obtain the time-series features of the employment stage ;

[0034] Obtain the evaluation features of the promotion effect in the regularization stage to capture the context dependence of the data in the regularization stage. Specifically, process the regularization situation data through a bidirectional long short-term memory network to learn the forward and backward dependencies simultaneously and output features containing bidirectional context information ;

[0035] Cross-stage feature fusion is used to establish the cross-stage dependence relationship between the employment stage and the regularization stage. Specifically, decompose the evaluation features of the promotion effect in the employment stage into employment shared features and employment private features, and at the same time decompose the evaluation features of the promotion effect in the regularization stage into regularization shared features and regularization private features. Dynamically evaluate the importance of the shared features and private features through the attention mechanism, calculate the weights of the shared features, and weight the sum of the shared features and the sum of the private features according to the weights to generate comprehensive features. The formula used is as follows:

[0036] ;

[0037] ;

[0038] In the formula, Indicates the shared features for employment Indicates the private features for employment Indicates the shared features for regularization Indicates the private features for regularization Indicates the weights of shared features Indicates the comprehensive features Indicates the attention weight matrix;

[0039] Obtain the evaluation result of the promotion effect in the employment stage, which is used to evaluate the promotion effect in the employment stage. Specifically, by calculating the weights of employment features and performing non-linear processing, the evaluation result of the recruitment promotion effect in the employment stage is obtained; the formula used is as follows:

[0040] ;

[0041] ;

[0042] In the formula, Indicates the weights of employment features Indicates the attention weight matrix in the employment stage Indicates the evaluation result of the recruitment promotion effect in the employment stage Indicates the evaluation weight matrix in the employment stage Indicates the bias term parameter in the employment stage Indicates the element-wise product;

[0043] Obtain the evaluation result of the promotion effect in the regularization stage, which is used to evaluate the promotion effect in the regularization stage. Specifically, by performing non-linear processing on the comprehensive features The evaluation result of the recruitment promotion effect in the regularization stage is obtained ;

[0044] Real-time prediction of the long-term promotion evaluation of talent recruitment. Specifically, the employment situation data, regularization situation data, recruitment activity information data, and job information data in the historical recruitment promotion effect evaluation data are used for model training, and the employment situation data, regularization situation data, recruitment activity information data, and job information data in the real-time recruitment promotion effect evaluation data are used as the input data of the trained long-term prediction evaluation model to obtain the long-term promotion evaluation prediction result of talent recruitment; the long-term promotion evaluation prediction result of talent recruitment includes the evaluation result of the promotion effect in the employment stage and the evaluation result of the promotion effect in the regularization stage;

[0045] Obtain the phased evaluation result set of the recruitment promotion effect. Specifically, the evaluation results of the promotion effect in each recruitment stage are combined to obtain the phased evaluation result set of the recruitment promotion effect.

[0046] Furthermore, the stage evaluation dynamic weight calculation module specifically includes grouping of talent recruitment and promotion scenarios, obtaining an initial weight combination, and obtaining an optimal weight combination to obtain an optimal stage evaluation dynamic weight combination, including the following steps:

[0047] Grouping of talent recruitment and promotion scenarios, specifically using the K-means clustering algorithm to perform clustering analysis on historical recruitment and promotion scenario data and real-time recruitment and promotion scenario data. After clustering, the real-time recruitment and promotion scenario data is assigned to the clustering clusters matching the historical recruitment and promotion scenario data to obtain real-time talent recruitment and promotion scenario labels;

[0048] Obtaining an initial weight combination, specifically retrieving all historical recruitment and promotion scenario data with the same label in the historical recruitment and promotion scenario data according to the real-time talent recruitment and promotion scenario label. Each piece of historical recruitment and promotion scenario data corresponds to a set of recruitment and promotion weight combinations. From all historical recruitment and promotion scenario data with the same label, by selecting the top 20 historical records closest to the recruitment and promotion start time of the real-time recruitment and promotion scenario data, the recruitment and promotion weight combinations of these 20 pieces of historical recruitment and promotion scenario data are summarized to generate an initial weight combination set;

[0049] Obtaining an optimal weight combination, specifically using the particle swarm optimization algorithm to obtain an optimal weight combination; including the following steps:

[0050] Particle initialization, taking each weight combination in the initial weight combination set as a particle in the particle swarm algorithm. Each particle represents a candidate weight combination, and the dimension of the particle corresponds to the weight value of each stage of recruitment and promotion;

[0051] Definition of fitness function, specifically calculating the fitness value of the particle through the fitness function; the formula used is as follows:

[0052] ;

[0053] In the formula, represents the fitness function, represents the i-th particle individual, represents the number of samples, represents the comprehensive evaluation value of the talent recruitment and promotion effect obtained through the weight combination represented by the current particle, represents the actual comprehensive evaluation value of the talent recruitment and promotion effect;

[0054] Particle update, specifically including velocity update, position update, and recording the optimal position of the particle;

[0055] Output the optimal weight combination. When the fitness value of a particle individual is higher than the fitness threshold and the maximum number of iterations is reached, terminate the search and obtain the global optimal position of the particle individual, where the global optimal position of the particle individual is the optimal stage evaluation dynamic weight combination.

[0056] Further, the comprehensive evaluation module for the recruitment promotion effect specifically combines the optimal stage evaluation dynamic weight combination with the recruitment promotion effect stage evaluation result set through a weighted average method to obtain the comprehensive evaluation result of the recruitment promotion effect. Based on the comprehensive evaluation result of the recruitment promotion effect, it helps enterprises optimize the recruitment promotion strategy and achieve a more accurate and efficient talent recruitment process.

[0057] The beneficial effects achieved by the present invention using the above solution are as follows:

[0058] (1) Aiming at the technical problem in the traditional talent recruitment promotion effect evaluation system that it is difficult to present the evaluation result of the promotion effect in a timely manner, resulting in a fault in the promotion effect evaluation cycle, which in turn affects the comprehensive evaluation effect of the entire recruitment process. This solution innovatively proposes to finely divide the recruitment promotion effect evaluation stage into six stages and adopt a combination of immediate evaluation and long-term prediction evaluation, effectively eliminating the problem of the evaluation cycle fault in the traditional method, ensuring the real-time and continuous evaluation results of each stage of the recruitment activity, significantly shortening the recruitment promotion effect evaluation cycle, reducing the promotion effect evaluation from monthly to daily, comprehensively covering the promotion effect evaluation of the entire talent recruitment process, and significantly improving the real-time and accuracy of the recruitment promotion effect evaluation.

[0059] (2) Aiming at the technical problems in the traditional talent recruitment promotion effect evaluation system that the evaluation result in the immediate evaluation stage of the promotion effect cannot be quantitatively and accurately evaluated and the accuracy of the evaluation result in the long-term prediction evaluation stage of the promotion effect is insufficient, resulting in inaccurate evaluation results in each recruitment promotion stage. This solution innovatively proposes that in the immediate evaluation stage of the promotion effect, through multi-dimensional quantitative indicators and a dynamic target value mechanism, it can reflect the actual recruitment situation in real time, provide a more realistic reference standard for the evaluation result in the immediate evaluation stage, and enhance the accuracy and practicality of the quantitative evaluation result; in the long-term prediction evaluation stage of the promotion effect, through a cross-stage feature fusion mechanism, an internal dependence relationship between the employment and regularization stages is established, comprehensively integrating long-term data features, significantly improving the prediction accuracy; it can effectively improve the accuracy of the evaluation results in each recruitment promotion stage.

[0060] (3) Aiming at the technical problem that the traditional talent recruitment promotion effect evaluation system uses static weight allocation, ignores the dynamic changes of multi-dimensional factors in the recruitment promotion scenario, resulting in the deviation of weight setting from the actual situation, and thus making the evaluation results of talent recruitment promotion effect inaccurate, this solution innovatively proposes a method for dynamically adjusting weights based on clustering analysis and particle swarm optimization algorithm. According to the actual situation of the recruitment promotion scenario, it flexibly adjusts the weight combination, improves the adaptability and accuracy of the evaluation process, eliminates the limitations of static weight setting, and takes into account the dynamic changes of multi-dimensional factors, thus significantly improving the accuracy and operability of the evaluation results of recruitment promotion effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a schematic diagram of the modules of the talent recruitment promotion effect evaluation system based on artificial intelligence provided by the present invention;

[0062] Figure 2 It is a schematic diagram of the process of the stage evaluation module for recruitment promotion effect;

[0063] Figure 3 It is a schematic diagram of the process of the stage evaluation dynamic weight calculation module;

[0064] Figure 4 It is a schematic diagram of the process of the immediate evaluation stage of promotion effect in the stage evaluation module for recruitment promotion effect;

[0065] Figure 5 It is a schematic diagram of the process of constructing a long-term prediction evaluation model in the long-term prediction evaluation stage of promotion effect;

[0066] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0067] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the 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 of the embodiments; based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention.

[0068] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. indicating the orientation or position relationship are based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the system or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation to the present invention.

[0069] Example 1. Refer to Figure 1 The talent recruitment promotion effect evaluation system based on artificial intelligence provided by the present invention includes a promotion effect evaluation data acquisition module, a data optimization module, a recruitment promotion effect stage evaluation module, a stage evaluation dynamic weight calculation module, and a recruitment promotion effect comprehensive evaluation module;

[0070] The promotion effect evaluation data acquisition module is specifically configured to collect the original data of the recruitment promotion effect evaluation through the recruitment channel platform and the company's personnel system, and send the data to the data optimization module;

[0071] The data optimization module receives the data sent by the promotion effect evaluation data acquisition module, eliminates invalid data through multi-dimensional evaluation data cleaning, then performs data standardization processing and data logical verification to ensure data logical consistency and business rationality, and finally implements promotion effect evaluation feature extraction to obtain the optimized data of the recruitment promotion effect evaluation, and sends the data to the recruitment promotion effect stage evaluation module and the stage evaluation dynamic weight calculation module;

[0072] The recruitment promotion effect stage evaluation module receives the data sent by the data optimization module. Specifically, it is divided into six stages according to the natural nodes of the recruitment process, and these six stages are further combined into an immediate evaluation stage and a long-term prediction evaluation stage. The evaluation results of the talent recruitment recommendation effect in each stage of the immediate evaluation stage and the long-term prediction evaluation stage are calculated respectively, and the evaluation results of each stage are combined to obtain a recruitment promotion effect stage evaluation result set, and the data is sent to the recruitment promotion effect comprehensive evaluation module;

[0073] The stage evaluation dynamic weight calculation module receives the data sent by the data optimization module. Through the K-means clustering algorithm, it obtains the real-time talent recruitment promotion scenario label, and searches for the recruitment promotion weight combination corresponding to this label in the historical recruitment promotion scenario data as the initial weight combination set. The particle swarm optimization algorithm is used to optimize on the basis of the initial weight combination set to obtain the optimal stage evaluation dynamic weight combination, and the data is sent to the recruitment promotion effect comprehensive evaluation module;

[0074] The recruitment promotion effect comprehensive evaluation module receives the data sent by the recruitment promotion effect stage evaluation module and the stage evaluation dynamic weight calculation module, and combines the optimal stage evaluation dynamic weight combination with the recruitment promotion effect stage evaluation result set through the weighted average method to obtain the comprehensive evaluation result.

[0075] Example 2. Refer to Figure 1, this embodiment is based on the above embodiment. The promotion effect evaluation data acquisition module is specifically configured to collect the original data for evaluating the recruitment promotion effect through the recruitment channel platform and the company's personnel system. The original data for evaluating the recruitment promotion effect includes historical promotion effect evaluation data, historical recruitment promotion scenario data, real-time recruitment promotion scenario data, and real-time promotion effect evaluation data. The historical recruitment promotion effect evaluation data and the real-time recruitment promotion effect evaluation data both include job exposure data, job click data, resume submission data, interview invitation data, employment situation data, probationary conversion situation data, recruitment activity information data, and job information data. The historical recruitment promotion effect evaluation data further includes the evaluation score for the talent recruitment and employment stage, the evaluation score for the talent recruitment probationary conversion stage, and the comprehensive evaluation value of the historical talent recruitment promotion effect. The historical recruitment promotion scenario data and the real-time recruitment promotion scenario data both include recruitment activity information data and job information data. The historical recruitment promotion scenario data further includes historical talent recruitment promotion scenario tags and recruitment promotion weight combinations. The recruitment promotion weight combination includes the weight values for each stage of job exposure, job click, job submission, job interview, employment, and probationary conversion. The job exposure data includes the number of exposures, display positions, and display time periods of the recruitment promotion channels. The job click data includes the number of clicks and click dwell times of the recruitment promotion channels. The resume submission data includes the resume submission volume, submission time, job source channels, and job seeker information for each job. The job seeker information includes education level, age, experience, and salary requirements. The interview invitation data includes the number of interview invitations, interview arrangement times, whether to participate in the interview, and actual participation status for each job. The employment situation data includes the employment results, employment times, employed positions, and job source channels of the interviewees. The probationary conversion situation data includes the probationary conversion results, probationary conversion times, probationary conversion positions, and performance evaluations during the probationary period of the employed personnel. The recruitment activity information data includes the start time, end time, promotion channels, promotion budget expenditures, and promotion forms of the recruitment promotion. The job information data includes the recruitment job name, job category, work location, salary range, and number of recruits.

[0076] Embodiment 3, refer to Figure 1 , this embodiment is based on the above embodiment. The data optimization module is used to optimize the original data for evaluating the recruitment promotion effect, including multi-dimensional evaluation data cleaning, data standardization processing, data logical verification, and extraction of promotion effect evaluation features, to obtain optimized data for evaluating the recruitment promotion effect. Specifically, it includes the following steps:

[0077] Multi-dimensional evaluation data cleaning is used to identify and process missing values, outliers, and duplicate records in the original data. Specifically, missing values are filled with the average value, outliers are identified and removed through statistical distribution methods, and duplicate records are de-duplicated through the data unique identifier field.

[0078] Data standardization processing is used to unify the data formats from different sources and structures, specifically including numerical data processing, categorical data processing, time field standardization, status field standardization, and text field structuring;

[0079] The numerical data processing specifically adopts the min-max normalization method for numerical features to map the data to a unified interval;

[0080] The categorical data processing specifically implements one-hot encoding for categorical data to convert it into a boolean vector form;

[0081] The time field standardization is used to ensure the comparability of the time dimension. Specifically, it adopts the unified time format conversion technology to uniformly convert the original time field into the standard timestamp format;

[0082] The status field standardization adopts the discrete status unified mapping technology to convert the unstructured status field into boolean data;

[0083] The text field structuring adopts natural language processing technology to clean, segment, and extract keywords from the unstructured text field to achieve structured expression;

[0084] Data logical verification is used to ensure the logical consistency and business rationality between different data fields. Specifically, it uses the preset business logic rules to check the relationship constraints and eliminates the illogical data; the business logic rules include that the click count shall not be greater than the exposure count, the delivery time shall be later than the click time, the interview invitation shall be based on valid delivery records, the employed personnel shall be included in the interviewed personnel, the regularization time shall be later than the employment time, and the start and end times of the recruitment activity shall include the timestamps of all relevant behavior data;

[0085] Recruitment promotion effect evaluation feature extraction is used to screen the core evaluation features related to the recruitment promotion effect from the data. Specifically, through statistical analysis methods, statistical feature selection is performed on the data to obtain the key evaluation features reflecting the recruitment effect.

[0086] Example 4, refer to Figure 1 、 Figure 2 and Figure 5 Based on the above example, the recruitment promotion effect stage evaluation module is used to conduct phased quantitative evaluation on the implementation effects of the recruitment promotion activities in different stages, including recruitment stage division, promotion effect immediate evaluation stage, promotion effect long-term prediction evaluation stage, and recruitment promotion effect stage evaluation result output; specifically includes the following steps:

[0087] The phased division of recruitment is used to break down the entire recruitment process into quantifiable recruitment stages. Specifically, the recruitment stages are divided and defined according to the natural nodes of the recruitment process. The recruitment stages include the job exposure stage, the job click stage, the job application stage, the job interview stage, the employment stage, and the probationary period completion stage. Further, the job exposure stage, the job click stage, the job application stage, and the job interview stage are classified as the immediate evaluation stage, and the employment stage and the probationary period completion stage are classified as the long-term prediction evaluation stage.

[0088] The immediate evaluation stage of promotion effect is used to quantitatively analyze the promotion effect of recruitment promotion activities in the short term. Specifically, by calculating the ratio of the actual performance in each stage to the preset fixed benchmark value and combining the key influencing dimensions of each stage, the promotion effect stage score of each stage in the immediate evaluation stage is obtained.

[0089] The long-term prediction evaluation stage of promotion effect is used to predict the long-term promotion effect of talent recruitment. Specifically, by constructing a long-term prediction evaluation model, the long-term promotion effect of talent recruitment is predicted. The specific steps are as follows:

[0090] Construct a long-term prediction evaluation model, which specifically includes the following steps:

[0091] Obtain the promotion effect evaluation features in the employment stage to capture the time series dependence relationship of the dynamic changes in the employment stage. Specifically, by processing the employment situation data through a long short-term memory network, the time series features in the employment stage are obtained. ;

[0092] Obtain the promotion effect evaluation features in the probationary period completion stage to capture the context dependence of the data in the probationary period completion stage. Specifically, by processing the probationary period completion situation data through a bidirectional long short-term memory network, the dependence relationships in both the forward and reverse directions are learned, and features containing bidirectional context information are output. ;

[0093] Cross-stage feature fusion is used to establish the cross-stage dependence relationship between the employment stage and the probationary period completion stage. Specifically, the promotion effect evaluation features in the employment stage are decomposed into employment shared features and employment private features, and at the same time, the promotion effect evaluation features in the probationary period completion stage are decomposed into probationary period completion shared features and probationary period completion private features. Through the attention mechanism, the importance of the shared features and the private features is dynamically evaluated, the weights of the shared features are calculated, and the sum of the shared features and the sum of the private features are weighted according to the weights to generate comprehensive features. The formula used is as follows:

[0094] ;

[0095] ;

[0096] ;

[0097] ;

[0098] In the formula, represents the employment sharing feature, represents the employment private feature, represents the regularization sharing feature, represents the regularization private feature, represents the weight of the sharing feature, represents the comprehensive feature, represents the attention weight matrix;

[0099] Obtain the promotion effect evaluation result in the employment stage, which is used to evaluate the promotion effect in the employment stage. Specifically, by calculating the employment feature weight and performing non-linear processing, the recruitment promotion effect evaluation result in the employment stage is obtained. The formula used is as follows:

[0100] ;

[0101] ;

[0102] In the formula, represents the employment feature weight, represents the attention weight matrix in the employment stage, represents the recruitment promotion effect evaluation result in the employment stage, represents the evaluation weight matrix in the employment stage, represents the bias term parameter in the employment stage, represents the element-wise product;

[0103] Obtain the promotion effect evaluation result in the regularization stage, which is used to evaluate the promotion effect in the regularization stage. Specifically, by performing non-linear processing on the comprehensive feature the recruitment promotion effect evaluation result in the regularization stage is obtained. The formula used is as follows:

[0104] ;

[0105] In the formula, represents the recruitment promotion effect evaluation result in the regularization stage, represents the evaluation weight matrix in the employment stage, represents the bias term parameter in the employment stage;

[0106] Real-time prediction of the long-term promotion evaluation of talent recruitment. Specifically, the data on employment status, regularization status, recruitment activity information, and position information in the historical recruitment promotion effect evaluation data are used for model training. The data on employment status, regularization status, recruitment activity information, and position information in the real-time recruitment promotion effect evaluation data are used as the input data of the trained long-term prediction evaluation model to obtain the long-term promotion evaluation prediction result of talent recruitment. The long-term promotion evaluation prediction result of talent recruitment includes the promotion effect evaluation result in the employment stage and the promotion effect evaluation result in the regularization stage.

[0107] Obtain the phased evaluation result set of the recruitment promotion effect. Specifically, the promotion effect evaluation results of each recruitment stage are combined to obtain the phased evaluation result set of the recruitment promotion effect. ; The formula used is as follows:

[0108] ;

[0109] In the formula, represents the phased evaluation result set of the recruitment promotion effect, represents the evaluation result of the recruitment promotion effect in the position exposure stage, represents the evaluation result of the recruitment promotion effect in the position click stage, represents the evaluation result of the recruitment promotion effect in the position application stage, represents the evaluation result of the recruitment promotion effect in the position interview stage, represents the evaluation result of the recruitment promotion effect in the employment stage, represents the evaluation result of the recruitment promotion effect in the regularization stage.

[0110] By performing the above operations, aiming at the technical problem in the traditional talent recruitment promotion effect evaluation system that it is difficult to present the promotion effect evaluation result in a timely manner, resulting in a break in the promotion effect evaluation cycle, and further affecting the comprehensive evaluation effect of the entire recruitment process. This solution innovatively proposes to finely divide the recruitment promotion effect evaluation stage into six stages, and adopts a combination of immediate evaluation and long-term prediction evaluation, effectively eliminating the problem of the evaluation cycle break in the traditional method, ensuring the real-time and continuous evaluation results of each stage of the recruitment activity, significantly shortening the recruitment promotion effect evaluation cycle, compressing the promotion effect evaluation from monthly to daily, comprehensively covering the promotion effect evaluation of the entire talent recruitment process, and significantly improving the real-time and accuracy of the recruitment promotion effect evaluation.

[0111] Example 5, refer to Figure 1 、 Figure 2 and Figure 4 Based on the above example, in the immediate evaluation stage of the promotion effect, the specific steps are as follows:

[0112] Calculate the score for the job exposure stage to evaluate the effectiveness of the job exposure stage. Calculate the ratio of the actual number of exposures to the fixed benchmark number of exposures, and combine the job promotion display position and display period. The formula used is as follows:

[0113] ;

[0114] In the formula, represents the score for the job exposure stage, represents the actual number of exposures, represents the fixed benchmark number of exposures, represents the weight of the job promotion display position, represents the weight of the job promotion display period;

[0115] Calculate the score for the job click stage to evaluate the effectiveness of the job click stage. Calculate the ratio of the actual number of clicks to the fixed benchmark number of clicks, and combine the dwell time at the time of clicking to calculate the promotion score for this stage. The formula used is as follows:

[0116] ;

[0117] In the formula, represents the score for the job click stage, represents the actual number of clicks, represents the fixed benchmark number of clicks, represents the weight of the dwell time when clicking on the job;

[0118] Calculate the score for the job application stage to evaluate the effectiveness of the job application stage. Calculate the ratio of the actual number of resume submissions to the fixed benchmark number of submissions, and combine the channel quality and candidate matching degree to calculate the promotion score for this stage. The formula used is as follows:

[0119] ;

[0120] In the formula, represents the score for the job application stage, represents the actual number of resume submissions, represents the fixed benchmark number of resume submissions, represents the weight of the job promotion channel, represents the weight of the candidate matching degree for job promotion;

[0121] Calculate the score for the job interview stage to evaluate the effectiveness of the job interview stage. Calculate the ratio of the actual number of interview invitations to the fixed benchmark number of interviews, and combine the timeliness of the interview arrangement and the interview participation rate to calculate the promotion score for this stage. The formula used is as follows:

[0122] ;

[0123] In the formula, represents the score in the job interview stage, represents the actual number of interviewees, represents the fixed benchmark number of interviewees, represents the timeliness weight of the job interview arrangement, represents the weight of the interviewees' arrival;

[0124] Calculation of the dynamic target value in the recruitment stage, which is used to adjust the target value of the talent recruitment promotion effect in real time. Specifically, a dynamic target value calculation model is constructed through a GRU neural network model. The historical promotion dynamic target value calculation data is used as training data to train the dynamic target value calculation model. The real-time promotion dynamic target value calculation data is used as input data and passed to the trained dynamic target value calculation model to obtain the dynamic target values of each stage of the recruitment activity. ;

[0125] The historical promotion dynamic target value calculation data includes the job exposure data, job click data, resume submission data, interview invitation data, recruitment activity information data, and job information data in the historical promotion effect evaluation data; the real-time promotion dynamic target value calculation data includes the job exposure data, job click data, resume submission data, interview invitation data, recruitment activity information data, and job information data in the real-time promotion effect evaluation data; the formula used is as follows:

[0126] ;

[0127] ;

[0128] In the formula, represents the hidden state at the current moment, represents the GRU model running function, represents the input data at the current moment, represents the hidden state at the previous moment, represents the predicted value of the dynamic target value in the i-th stage, represents the weight matrix of the output layer, represents the bias term of the output layer, represents the Sigmoid activation function;

[0129] Calculate the evaluation results in the immediate evaluation stage. Specifically, combine the dynamic target value with the actual score of each stage in the immediate evaluation stage to obtain the recruitment promotion effect evaluation results of each stage in the immediate evaluation stage; the formula used is as follows:

[0130] ;

[0131] ;

[0132] ;

[0133] wherein, represents the evaluation result of the recruitment promotion effect in the i-th stage, represents the weight coefficient for balancing the historical benchmark and dynamic prediction, represents the actual score in the i-th stage, represents the historical highest score in the i-th stage, represents the fixed benchmark value in the i-th stage.

[0134] By performing the above operations, aiming at the technical problems in the traditional talent recruitment promotion effect evaluation system that the evaluation results in the immediate evaluation stage of the promotion effect cannot achieve quantitative and accurate evaluation, and the accuracy of the evaluation results in the long-term prediction evaluation stage of the promotion effect is insufficient, resulting in inaccurate evaluation results in each recruitment promotion stage, this solution innovatively proposes that in the immediate evaluation stage of the promotion effect, through multi-dimensional quantitative indicators and dynamic target value mechanisms, it can reflect the actual recruitment situation in real time, provide a more realistic reference standard for the evaluation results in the immediate evaluation stage, and enhance the accuracy and practicality of the quantitative evaluation results; in the long-term prediction evaluation stage of the promotion effect, through the cross-stage feature fusion mechanism, establish the internal dependence relationship between the employment and regularization stages, comprehensively integrate the long-term data features, and significantly improve the prediction accuracy; it can effectively improve the accuracy of the evaluation results in each recruitment promotion stage.

[0135] Example Six, refer to Figure 1 and Figure 3 Based on the above embodiment, the specific steps of the stage evaluation dynamic weight calculation module for obtaining the optimal stage evaluation dynamic weight combination by grouping the talent recruitment promotion scenarios, obtaining the initial weight combination, and obtaining the optimal weight combination are as follows:

[0136] Grouping the talent recruitment promotion scenarios, specifically using the K-means clustering algorithm to perform clustering analysis on the historical recruitment promotion scenario data and real-time recruitment promotion scenario data. After clustering, the real-time recruitment promotion scenario data is assigned to the clustering cluster matching the historical recruitment promotion scenario data to obtain the real-time talent recruitment promotion scenario label;

[0137] Obtaining the initial weight combination, specifically according to the real-time talent recruitment promotion scenario label, retrieving all historical recruitment promotion scenario data with the same label in the historical recruitment promotion scenario data. Each piece of historical recruitment promotion scenario data corresponds to a set of recruitment promotion weight combinations. From all historical recruitment promotion scenario data with the same label, by selecting the first 20 historical records with the closest recruitment promotion start time to the real-time recruitment promotion scenario data, summarize the recruitment promotion weight combinations of these 20 pieces of historical recruitment promotion scenario data to generate the initial weight combination set;

[0138] Obtain the optimal weight combination, specifically, use the particle swarm optimization algorithm to obtain the optimal weight combination; the steps are as follows:

[0139] Particle initialization: Take each weight combination in the initial weight combination set as a particle in the particle swarm algorithm. Each particle represents a candidate weight combination, and the dimension of the particle corresponds to the weight value at each stage of recruitment promotion;

[0140] Definition of fitness function: Specifically, calculate the fitness value of the particle through the fitness function; the formula used is as follows:

[0141] ;

[0142] In the formula, represents the fitness function, represents the i-th particle individual, represents the number of samples, represents the comprehensive evaluation value of the talent recruitment promotion effect obtained through the weight combination represented by the current particle, represents the actual comprehensive evaluation value of the talent recruitment promotion effect;

[0143] Particle update: Specifically, it includes velocity update, position update, and recording the optimal position of the particle;

[0144] The velocity update is specifically to update the velocity of the particle according to the current velocity, individual optimal position, and global optimal position of the particle, and push the particle to move in the direction of the optimal solution;

[0145] The position update is specifically to adjust the position of the particle according to the updated velocity, and the new particle position represents a new weight combination;

[0146] The recording of the optimal position of the particle is specifically to record the historical optimal position of each particle and the global optimal position of the group;

[0147] Output the optimal weight combination. When the fitness value of the particle individual is higher than the fitness threshold and the maximum number of iterations is reached, terminate the search and obtain the global optimal position of the particle individual. The global optimal position of the particle individual is the optimal stage evaluation dynamic weight combination.

[0148] By performing the above operations, for the technical problem that the traditional talent recruitment promotion effect evaluation system uses static weight allocation, ignores the dynamic changes of multi-dimensional factors in the recruitment promotion scenario, resulting in the deviation of weight settings from the actual situation, and thus making the evaluation results of talent recruitment promotion inaccurate, this solution innovatively proposes a method for dynamically adjusting weights based on cluster analysis and particle swarm optimization algorithm. According to the actual situation of the recruitment promotion scenario, the weight combination is flexibly adjusted, improving the adaptability and accuracy of the evaluation process, eliminating the limitations of static weight settings, and considering the dynamic changes of multi-dimensional factors, thereby significantly improving the accuracy and operability of the evaluation results of recruitment promotion effects.

[0149] Example Seven. Refer to Figure 1 , this example is based on the above example. The comprehensive evaluation module for recruitment promotion effect specifically combines the optimal stage evaluation dynamic weight combination with the recruitment promotion effect stage evaluation result set through the weighted average method to obtain the comprehensive evaluation result of recruitment promotion effect. According to the comprehensive evaluation result of recruitment promotion effect, it helps enterprises optimize recruitment promotion strategies and achieve a more accurate and efficient recruitment process. The formula used is as follows:

[0150] ;

[0151] ;

[0152] In the formula, represents the optimal stage evaluation dynamic weight combination, represents the recruitment promotion weight in the job exposure stage, represents the recruitment promotion weight in the job click stage, represents the recruitment promotion weight in the job application stage, represents the recruitment promotion weight in the job interview stage, represents the recruitment promotion weight in the employment stage, represents the recruitment promotion weight in the regularization stage, represents the comprehensive evaluation result of recruitment promotion effect, represents the recruitment promotion weight in the jth recruitment stage, represents the evaluation result of the jth recruitment stage effect.

[0153] It should be noted that, in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0154] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention.

[0155] The present invention and its embodiments have been described above. Such description is not restrictive. What is shown in the drawings is only one of the embodiments of the present invention, and the actual structure is not limited thereto. In summary, if those of ordinary skill in the art are inspired by it and, without departing from the gist of the present invention, design similar structural modes and embodiments to this technical solution without creative efforts, they shall fall within the protection scope of the present invention.

Claims

1. An artificial intelligence-based talent recruitment promotion effect evaluation system, characterized in that: It includes a promotion effect evaluation data acquisition module, a data optimization module, a recruitment promotion effect stage evaluation module, a stage evaluation dynamic weight calculation module, and a recruitment promotion effect comprehensive evaluation module; The promotion effect evaluation data acquisition module is specifically for collecting the original data of the recruitment promotion effect evaluation; The data optimization module is used to optimize the original data of the recruitment promotion effect evaluation, specifically including multi-dimensional evaluation data cleaning, data standardization processing, data logical verification, and promotion effect evaluation feature extraction, to obtain the optimized data of the recruitment promotion effect evaluation; The recruitment promotion effect stage evaluation module is specifically divided into six stages according to the natural nodes of the recruitment process, and further combines these six stages into an immediate evaluation stage and a long-term prediction evaluation stage, respectively calculates the evaluation results of the talent recruitment recommendation effect in each stage of the immediate evaluation stage and the long-term prediction evaluation stage, and combines the evaluation results of each stage to obtain the recruitment promotion effect stage evaluation result set; including recruitment stage division, promotion effect immediate evaluation stage, promotion effect long-term prediction evaluation stage, and recruitment promotion effect stage evaluation result output; The stage evaluation dynamic weight calculation module is specifically to obtain the real-time talent recruitment promotion scenario label through the clustering algorithm, and find the recruitment promotion weight combination corresponding to this label in the historical recruitment promotion scenario data as the initial weight combination set, and use the particle swarm optimization algorithm to optimize on the basis of the initial weight combination set to obtain the optimal stage evaluation dynamic weight combination; The recruitment promotion effect comprehensive evaluation module is specifically to combine the optimal stage evaluation dynamic weight combination with the recruitment promotion effect stage evaluation result set through the weighted average method to obtain the comprehensive evaluation result.

2. The talent recruitment promotion effect evaluation system based on artificial intelligence according to claim 1, wherein: The recruitment promotion effect stage evaluation module is used to quantitatively evaluate the implementation effect of the recruitment promotion activity in different stages in stages; specifically includes the following steps: Recruitment stage division is used to disassemble the entire recruitment process into quantifiable recruitment stages, specifically to divide and define the recruitment stages according to the natural nodes of the recruitment process. The recruitment stages include the job exposure stage, the job click stage, the job application stage, the job interview stage, the employment stage, and the probationary conversion stage; further divide the job exposure stage, the job click stage, the job application stage, and the job interview stage into the immediate evaluation stage, and divide the employment stage and the probationary conversion stage into the long-term prediction evaluation stage; The promotion effect immediate evaluation stage is used to quantitatively analyze the promotion effect of the recruitment promotion activity in the short term. Specifically, by calculating the ratio of the actual performance in each stage to the preset fixed benchmark value, and combining the key influence dimensions of each stage, the promotion effect stage score of each stage in the immediate evaluation stage is obtained; The promotion effect long-term prediction evaluation stage is used to predict the long-term talent recruitment promotion effect. Specifically, by constructing a long-term prediction evaluation model, the long-term promotion effect of talent recruitment is predicted; Obtain the recruitment promotion effect stage evaluation result set, specifically by combining the promotion effect evaluation results of each recruitment stage to obtain the recruitment promotion effect stage evaluation result set.

3. The talent recruitment promotion effect evaluation system based on artificial intelligence according to claim 1, characterized in that: In the immediate evaluation stage of the promotion effect, the following specific steps are included: Calculate the score for the job exposure stage to evaluate the effectiveness of the job exposure stage. Calculate the ratio of the actual number of exposures to the fixed benchmark number of exposures, and combine the job promotion display position and display time period to obtain the score for the job exposure stage ; Calculate the score for the job click stage to evaluate the effectiveness of the job click stage. Calculate the ratio of the actual number of clicks to the fixed benchmark number of clicks, and calculate the promotion score for this stage in combination with the dwell time at the time of clicking to obtain the score for the job click stage ; Calculate the score for the job application stage to evaluate the effectiveness of this stage. Calculate the ratio of the actual number of resume submissions to the fixed benchmark number of submissions, and combine the channel quality and candidate matching degree to calculate the promotion score for this stage, thus obtaining the score for the job application stage ; Calculate the score for the job interview stage to evaluate the effectiveness of this stage. Calculate the ratio of the actual number of interview invitations to the fixed benchmark number of interviews, and combine the timeliness of the interview arrangement and the interview participation rate to calculate the promotion score for this stage, thus obtaining the score for the job interview stage ; Calculation of dynamic target values in the recruitment stage, which is used to adjust the target values of talent recruitment promotion in real time. Specifically, a dynamic target value calculation model is constructed through a GRU neural network model. The historical promotion dynamic target value calculation data is used as training data to train the dynamic target value calculation model. The real-time promotion dynamic target value calculation data is used as input data and passed to the trained dynamic target value calculation model to obtain the dynamic target values of each stage of the recruitment activity ; The formula used is as follows: ； ； In the formula, represents the hidden state at the current moment, represents the GRU model running function, represents the input data at the current moment, represents the hidden state at the previous moment, represents the predicted value of the dynamic target value at the i-th stage, represents the weight matrix of the output layer, represents the bias term of the output layer, represents the Sigmoid activation function; Calculate the evaluation results in the immediate evaluation stage. Specifically, combine the dynamic target value with the actual scores of each stage in the immediate evaluation stage to obtain the evaluation results of the recruitment promotion effect for each stage in the immediate evaluation stage. The formula used is as follows: ； ； Wherein, represents the evaluation result of the recruitment promotion effect in the i-th stage, represents the weight coefficient for balancing the historical benchmark and the dynamic prediction, represents the actual score in the i-th stage, represents the historical highest score in the i-th stage, represents the fixed benchmark value in the i-th stage.

4. The talent recruitment promotion effect evaluation system based on artificial intelligence according to claim 1, characterized in that: In the long-term prediction and evaluation stage of the promotion effect, the following specific steps are included: Construct a long-term prediction and evaluation model, which specifically includes the following steps: Obtain the promotion effect evaluation features in the recruitment stage to capture the time series dependencies of the dynamic changes in the recruitment stage. Specifically, process the recruitment data through a long short-term memory network to obtain the time series features in the recruitment stage ; Obtain the promotion effect evaluation features in the regularization stage, which are used to capture the context dependencies of the data in the regularization stage. Specifically, process the regularization situation data through a bidirectional long short-term memory network, learn the forward and backward dependencies simultaneously, and output features containing bidirectional context information ; Cross-stage feature fusion is used to establish the cross-stage dependence relationship between the employment stage and the regularization stage. Specifically, decompose the promotion effect evaluation features in the employment stage into employment shared features and employment private features, and at the same time decompose the promotion effect evaluation features in the regularization stage into regularization shared features and regularization private features. Dynamically evaluate the importance of shared features and private features through the attention mechanism, calculate the weights of the shared features, and weight the sum of the shared features and the sum of the private features according to the weights to generate comprehensive features. The formula used is as follows: ； ； Wherein, represents the employment sharing feature, represents the employment private feature, represents the regularization sharing feature, represents the regularization private feature, represents the weight of the sharing feature, represents the comprehensive feature, represents the attention weight matrix; Obtain the evaluation results of the promotion effect in the employment stage to evaluate the promotion effect in the employment stage. Specifically, calculate the employment feature weights and perform non-linear processing to obtain the evaluation results of the recruitment promotion effect in the employment stage. The formula used is as follows: ； ； In the formula, represents the employment feature weight, represents the attention weight matrix in the employment stage, represents the evaluation result of the recruitment promotion effect in the employment stage, represents the evaluation weight matrix in the employment stage, represents the bias term parameter in the employment stage, represents the element-wise product; Obtain the evaluation results of the promotion effect in the regularization stage, which are used to evaluate the promotion effect in the regularization stage. Specifically, by performing non-linear processing on the comprehensive features to obtain the evaluation results of the recruitment promotion effect in the regularization stage ; Real-time prediction of the long-term promotion evaluation of talent recruitment is specifically to use the employment situation data, regularization situation data, recruitment activity information data, and job information data in the historical recruitment promotion effect evaluation data for model training, and use the employment situation data, regularization situation data, recruitment activity information data, and job information data in the real-time recruitment promotion effect evaluation data as the input data of the trained long-term prediction and evaluation model to obtain the long-term promotion evaluation prediction results of talent recruitment. The long-term promotion evaluation prediction results of talent recruitment include the evaluation results of the promotion effect in the employment stage and the evaluation results of the promotion effect in the regularization stage.

5. The talent recruitment promotion effect evaluation system based on artificial intelligence according to claim 1, characterized in that: The specific steps of the stage evaluation dynamic weight calculation module specifically include grouping the talent recruitment promotion scenarios, obtaining the initial weight combination, and obtaining the optimal weight combination to obtain the optimal stage evaluation dynamic weight combination, including the following steps: Group the talent recruitment promotion scenarios. Specifically, use the K-means clustering algorithm to perform clustering analysis on the historical recruitment promotion scenario data and the real-time recruitment promotion scenario data. After clustering, the real-time recruitment promotion scenario data is assigned to the clustering cluster that matches the historical recruitment promotion scenario data to obtain the real-time talent recruitment promotion scenario label. Obtain the initial weight combination. Specifically, according to the real-time talent recruitment promotion scenario label, retrieve all the historical recruitment promotion scenario data with the same label in the historical recruitment promotion scenario data. Each piece of historical recruitment promotion scenario data corresponds to a set of recruitment promotion weight combinations. From all the historical recruitment promotion scenario data with the same label, select the first 20 historical records whose recruitment promotion start time is closest to the real-time recruitment promotion scenario data, and summarize the recruitment promotion weight combinations of these 20 pieces of historical recruitment promotion scenario data to generate an initial weight combination set. Obtain the optimal weight combination. Specifically, use the particle swarm optimization algorithm to obtain the optimal weight combination, including the following steps: Particle initialization: Each weight combination in the initial weight combination set is used as a particle in the particle swarm optimization algorithm. Each particle represents a candidate weight combination, and the dimension of the particle corresponds to the weight value at each stage of recruitment promotion. Fitness function definition: Specifically, calculate the fitness value of the particle through the fitness function. The formula used is as follows: ； In the formula, represents the fitness function, represents the i-th particle individual, represents the number of samples, represents the comprehensive evaluation value of the talent recruitment promotion effect obtained through the weight combination represented by the current particle, represents the comprehensive evaluation value of the actual talent recruitment promotion effect; Particle update: Specifically, it includes velocity update, position update, and recording the optimal position of the particle. Output the optimal weight combination: When the individual fitness value of the particle is higher than the fitness threshold and the maximum number of iterations is reached, terminate the search and obtain the global optimal position of the particle individual. The global optimal position of the particle individual is the optimal stage evaluation dynamic weight combination.

6. The talent recruitment promotion effect evaluation system based on artificial intelligence according to claim 1, characterized in that: The comprehensive evaluation module for recruitment promotion effect specifically combines the optimal stage evaluation dynamic weight combination with the recruitment promotion effect stage evaluation result set through the weighted average method to obtain the comprehensive evaluation result of the recruitment promotion effect. According to the comprehensive evaluation result of the recruitment promotion effect, it helps the enterprise optimize the recruitment promotion strategy and achieve a more accurate and efficient talent recruitment process.

7. The talent recruitment promotion effect evaluation system based on artificial intelligence according to claim 1, characterized in that: The promotion effect evaluation data acquisition module specifically collects the original data for evaluating the recruitment promotion effect through the recruitment channel platform and the company's personnel system. The original data for evaluating the recruitment promotion effect includes historical promotion effect evaluation data, historical recruitment promotion scenario data, real-time recruitment promotion scenario data, and real-time promotion effect evaluation data. The historical recruitment promotion effect evaluation data and the real-time recruitment promotion effect evaluation data both include job exposure data, job click data, resume submission data, interview invitation data, employment situation data, probationary status data, recruitment activity information data, and job information data. The historical recruitment promotion effect evaluation data also includes the evaluation score for the talent recruitment and employment stage, the evaluation score for the talent recruitment and probationary stage, and the historical comprehensive evaluation value of the talent recruitment promotion effect. The historical recruitment promotion scenario data and the real-time recruitment promotion scenario data both include recruitment activity information data and job information data. The historical recruitment promotion scenario data also includes the historical talent recruitment promotion scenario label and the recruitment promotion weight combination.

8. The talent recruitment promotion effect evaluation system based on artificial intelligence according to claim 1, characterized in that: The data optimization module is used to optimize the original data for evaluating the recruitment promotion effect, and specifically includes the following steps: Multi-dimensional evaluation data cleaning: Used to identify and process missing values, outliers, and duplicate records in the original data. Data standardization processing: Used to unify the data formats from different sources and structures. Specifically, it includes numerical data processing, categorical data processing, time field standardization, status field standardization, and text field structuring. Data logical verification is used to ensure the logical consistency and business rationality between different data fields. Specifically, preset business logic rules are used for relationship constraint checks to eliminate illogical data. Promotion effect evaluation feature extraction: Used to screen the core evaluation features related to the recruitment promotion effect from the data.

Images