Operating personnel post selection method
Through automated processing and clustering model optimization, the problems of low efficiency and low accuracy of existing job selection are solved, and efficient and accurate job selection is achieved to adapt to changes in market demand.
Patent Information
- Application Number
- CN202510284874.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-11
- Publication Date
- 2025-07-29
AI Technical Summary
The existing job selection methods are inefficient and have low accuracy, and mainly rely on manual screening, which has problems of high time consumption and subjective bias.
By extracting relevant texts from the worker from the database, using natural language processing technology to automatically extract key information, and perform numerical processing and evaluation based on quantitative strategies and clustering models, combined with the Q-Learning algorithm to optimize weights, calculate the matching degree between candidates and positions, and realize automated selection.
It improves the efficiency and accuracy of job selection, reduces the time and subjective bias of manual screening, can more accurately reflect the fit between candidates and positions, and adapts to market changes through continuous learning and optimization of models.
Smart Images
Figure CN120387714A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data processing, and in particular, to a method for selecting job personnel positions. Background Art
[0002] In the workplace, position selection is a common method used by enterprises to select talents. Job personnel with stronger capabilities can enter higher job ranks through position selection, thereby participating in higher-level affairs. The existing position selection generally adopts the method of manual screening, that is, the selectors determine the candidates for promotion based on their impressions of the candidates' work performance. However, this method is inefficient and has a low accuracy rate. Summary of the Invention
[0003] In view of this, the purpose of the embodiments of the present application is to provide a method for selecting job personnel positions, which can improve the problems of low efficiency and low accuracy rate in the existing technology.
[0004] To achieve the above technical purpose, the technical solution adopted by the present application is as follows:
[0005] The embodiments of the present application provide a method for selecting job personnel positions, and the method includes:
[0006] Extract relevant texts of the job personnel from a database, where the relevant texts include resumes, professional skill assessment results, and past work performance records;
[0007] Extract first key information from the relevant texts, where the first key information includes first work experience, first educational background, and first professional skills;
[0008] Based on a preset quantization strategy, numerically process the first key information;
[0009] Generate a candidate ability profile of the job personnel according to the numerically processed first key information;
[0010] Calculate the matching degree between the ability profile and the clustering center obtained according to a preset clustering model;
[0011] Based on the matching degree, determine the position selection result of the job personnel.
[0012] Further, after determining the position selection result of the job personnel according to the clustering result, the method further includes:
[0013] Use the candidate ability profile corresponding to the position selection result as a training set to input into the clustering model for correcting the clustering center.
[0014] Further, generating the candidate ability profile of the operator based on the numerically processed first key information includes:
[0015] Normalize all types of the first key information to obtain a work experience index value, an educational background index value, and a professional skill index value;
[0016] Based on the historical relevant text, determine the first weight values of the work experience index, the educational background index, and the professional skill index;
[0017] According to the work experience index value, the educational background index value, the professional skill index value, and the corresponding first weight values, obtain the candidate ability profile.
[0018] Further, the determining the first weight values of the work experience index, the educational background index, and the professional skill index based on the historical relevant text includes:
[0019] Obtain the numerically processed second key information of the operator corresponding to the historical job selection result from the historical relevant text, where the second key information includes second work experience, second educational background, and second professional skill;
[0020] Input the second key information into the Q-Learning algorithm model to obtain the first weight values. The Q-Learning algorithm model is configured to, after receiving the second key information, input the second key information into the reward function of the Q-Learning algorithm model, and update the Q-table of the Q-Learning algorithm model based on the calculation result of the reward function. According to the updated Q-table, obtain the first weight values.
[0021] Further, the reward function is:
[0022]
[0023] A t is the ability selected at time step t;
[0024] S t is the state at time step t, which can represent the operator's current working environment, task type, or any factor affecting performance;
[0025] R(S t ,A t ) is the reward obtained by performing action A t in state S t ;
[0026] P i is the i-th performance indicator in the second key information;
[0027] W i is the weight corresponding to the i-th performance indicator. These weights may be set based on expert experience or business requirements initially and will be adjusted during the Q-Learning process;
[0028] N is the total number of performance indicators;
[0029] f i (P i (A t , S t )) is the function value of the i-th performance indicator P i in the state S t when performing the action A t , and this function maps the performance indicator to the reward space.
[0030] Furthermore, extracting the first key information from the relevant text includes:
[0031] Performing word segmentation on the relevant text to obtain a number of words;
[0032] Performing part-of-speech tagging on the words to obtain the part-of-speech of each word;
[0033] Based on a domain dictionary or knowledge base, identifying the named entities of the words after part-of-speech tagging;
[0034] Based on a natural language processing model, extracting the first key information according to the named entities.
[0035] Furthermore, the extracting the first key information based on the natural language processing model according to the named entities includes:
[0036] Based on the natural language processing model, extracting the named entities used to represent the year and job title to form the first work experience; extracting the named entities used to represent the school name and educational level to form the first educational background; and extracting the named entities used to represent the performance score to form the first professional skill.
[0037] Furthermore, the preset quantization strategy includes a first strategy, a second strategy, and a third strategy;
[0038] The numerical processing of the first key information based on the preset quantization strategy includes:
[0039] Assigning weights to the first key information so that the first work experience, the first educational background, and the first professional skill are assigned corresponding second weights;
[0040] According to the first strategy, assign a first value to the first work experience, where the first strategy is to determine the first value according to the number of years of work corresponding to the first work experience;
[0041] According to the second strategy, assign a second value to the first educational background, where the second strategy is to determine the second value according to the educational attainment corresponding to the first educational background;
[0042] According to the third strategy, assign a third value to the first professional skill, where the third strategy is to determine the third value according to all performance scores corresponding to the first professional skill;
[0043] Based on the first value, the second value, and the third value, numerically process the first key information.
[0044] Further, after taking the candidate ability portrait corresponding to the job selection result as a node and inputting it into the clustering model, the method further includes:
[0045] S1: Based on all nodes of the clustering model, obtain an initial clustering center, where the first work experience corresponding to the initial clustering center is the median value of the first work experience of all nodes; the first educational background corresponding to the initial clustering center is the median value of the first educational background of all nodes; the first professional skill corresponding to the initial clustering center is the median value of the first professional skill of all nodes;
[0046] S2: Calculate the first Euclidean distance between all nodes and the initial clustering center, and screen the nodes with the first Euclidean distance less than or equal to the first distance threshold to obtain screened nodes;
[0047] S3: Calculate the weighted average of the first work experience, the weighted average of the first educational background, and the weighted average of the first professional skill of all screened nodes. Based on the weighted average of the first work experience, the weighted average of the first educational background, and the weighted average of the first professional skill, form a weighted node, and calculate the second Euclidean distance between the weighted node and the initial clustering center;
[0048] S4: If the second Euclidean distance is less than or equal to the second distance threshold, then set the weighted node as the corrected clustering center; if the second Euclidean distance is greater than the second distance threshold, then set the weighted node as the initial clustering center, and repeat S2 and S3 until a weighted node with a second Euclidean distance less than or equal to the second distance threshold from the initial clustering center is obtained.
[0049] The invention adopting the above technical solution has the following advantages:
[0050] In the technical solution provided by this application, by extracting relevant texts from the database and automatically extracting key information using technologies such as natural language processing, the automation of the selection process is achieved, reducing the time and effort required for manual screening and improving the selection efficiency. Based on the preset quantization strategy and clustering model, this solution can quickly perform numerical processing and evaluation on the capabilities of candidates, avoiding the delays and uncertainties that may occur in the manual screening process, making the selection process faster and more efficient. Through the preset quantization strategy, the key information of candidates (such as work experience, educational background, professional skills, etc.) can be numerically processed, making the evaluation results more objective and accurate, and avoiding the subjective biases and misjudgments that may occur in the manual screening process. This technical solution uses the clustering model to calculate the matching degree of the candidate's ability portrait, thereby determining the selection result, which can more accurately reflect the fit between the candidate and the job requirements and improve the accuracy of the selection. This technical solution also includes a correction mechanism for the clustering center, that is, as new selection data is added, the model can continuously learn and optimize, enabling the preset clustering model to continuously improve the accuracy and efficiency of the selection, so as to better adapt to the changing market and talent needs. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] This application can be further illustrated by the non-limiting embodiments given in the drawings. It should be understood that the following drawings only show some embodiments of this application, so they should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other relevant drawings can also be obtained based on these drawings.
[0052] Figure 1 It is a flowchart provided for an embodiment of this application.
[0053] Figure 2 It is a flowchart of steps 121 - 124 provided for an embodiment of this application.
[0054] Figure 3 It is a flowchart of steps 131 - 135 provided for an embodiment of this application.
[0055] Figure 4 It is a flowchart of steps 141 - 143 provided for an embodiment of this application.
[0056] Figure 5 It is a flowchart of steps 1421 - 1422 provided for an embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0057] The present application will be described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that in the description of the drawings or the specification, similar or identical parts are all denoted by the same reference numerals, and the implementation manners not shown or described in the drawings are in the forms known to those of ordinary skill in the art. In the description of the present application, the terms "first", "second", etc. are only used for distinguishing descriptions and cannot be construed as indicating or implying relative importance.
[0058] Please refer to Figure 1 , the present application provides a method for selecting job positions for workers. Among them, the method for selecting job positions for workers may include the following steps:
[0059] Step 110: Extract the relevant texts of the worker from the database, where the relevant texts include resumes, professional skill assessment results, and past work performance records;
[0060] Step 120: Extract the first key information from the relevant texts, where the first key information includes first work experience, first educational background, and first professional skills;
[0061] Step 130: Numerically process the first key information based on a preset quantization strategy;
[0062] Step 140: Generate a candidate ability profile of the worker according to the numerically processed first key information;
[0063] Step 150: Calculate the matching degree between the ability profile and the clustering center obtained according to the preset clustering model;
[0064] Step 160: Determine the job position selection result of the worker based on the matching degree.
[0065] The following will elaborate on each step of the method for selecting job positions for workers in detail, as follows:
[0066] In step 110, the relevant texts are actively transmitted to the database by relevant personnel. The database in this embodiment is used to store the relevant texts of the workers.
[0067] When extracting the relevant texts, corresponding SQL query statements can be written for extracting the relevant texts.
[0068] As Figure 2 shown, step 120 may specifically include the following steps:
[0069] Step 121: Perform word segmentation on the relevant texts to obtain a number of words;
[0070] Step 122: Perform part-of-speech tagging on the words to obtain the part-of-speech of each word;
[0071] Step 123: Based on the domain dictionary or knowledge base, identify the named entities of the words after part-of-speech tagging;
[0072] Step 124: Based on the natural language processing model, extract the first key information according to the named entities.
[0073] In step 121, it can be implemented in the following ways:
[0074] First, clean the relevant text to remove useless characters, punctuation marks, etc.
[0075] Second, select an appropriate word segmentation method according to the specific application scenario and requirements. Specifically, the word segmentation methods include the dictionary matching method and the statistics-based word segmentation method.
[0076] The dictionary matching method includes forward maximum matching (FMM), reverse maximum matching (RMM), bidirectional maximum matching, and word-by-word traversal method.
[0077] Forward maximum matching (FMM) means scanning the text from left to right and matching the longest vocabulary as much as possible each time.
[0078] Reverse maximum matching (RMM) scans the text from right to left and also matches the longest vocabulary as much as possible each time.
[0079] Bidirectional maximum matching: Combine the results of FMM and RMM and select the optimal word segmentation method.
[0080] Word-by-word traversal method: Traverse all possible vocabulary combinations in a certain order (such as from front to back or from back to front) to find the optimal word segmentation result.
[0081] The statistics-based word segmentation methods include:
[0082] Hidden Markov model (HMM): Use HMM to model the text and determine the optimal word segmentation method by calculating probabilities.
[0083] Conditional random field (CRF): CRF is a discriminative model that can output the optimal label sequence (i.e., the word segmentation result) under the condition of a given input sequence.
[0084] Neural network method: Common neural network models include convolutional neural network (CNN), recurrent neural network (RNN) and its variants (such as LSTM, GRU), etc. These models can automatically learn the features in the text and output accurate word segmentation results.
[0085] In step 122, an existing NLP library can be used to implement part-of-speech tagging. The types of tags include work experience, school names and major names in educational background, etc.
[0086] Currently, many NLP libraries (such as NLTK, SpaCy, jieba, etc.) provide the function of part-of-speech tagging. These libraries usually contain pre-trained models and dictionaries, which can be directly used for part-of-speech tagging.
[0087] In step 123, named entity recognition tools (such as the named entity recognition modules of SpaCy and NLTK) can be used to perform named entity recognition on the text.
[0088] In step 124, a pre-trained NLP model can be selected, or a custom model can be constructed according to specific requirements. These models can be based on machine learning or deep learning algorithms and have been trained to understand the semantics and structure in the text and extract the first key information. The specific logic is as follows:
[0089] Based on the natural language processing model, extract the named entities used to represent the year and job title to form the first work experience; extract the named entities used to represent the school name and educational level to form the first educational background; extract the named entities used to represent the performance score to form the first professional skill.
[0090] Among them, identify the entity pairs containing the year and job title. Determine the correspondence between the year and the job according to the context. Form structured data, such as {year: '2020', job title: 'Project Manager'}, and aggregate to form the first work experience.
[0091] Educational background extraction: Identify the entities of the school name and educational level. Determine the correspondence between the school and the educational level according to the context. Form structured data, such as {school: 'Peking University', educational level:'master'}, and aggregate to form the first educational background.
[0092] Professional skill extraction: Identify the performance score entity. If the performance score is directly related to a specific skill or project, extract the associated information. Form structured data, such as {skill / project: 'Project Management', performance score: '90 points'} (if applicable), or only extract the performance score for subsequent evaluation.
[0093] In step 130, the preset quantization strategies include the first strategy, the second strategy, and the third strategy. The first strategy is to determine the first value according to the working years corresponding to the first work experience; the second strategy is to determine the second value according to the educational level corresponding to the first educational background; the third strategy is to determine the third value according to all the performance scores corresponding to the first professional skill.
[0094] As Figure 3 shown, step 130 mainly includes the following steps:
[0095] Step 131: Assign weights to the first key information so that the first work experience, the first educational background, and the first professional skills are assigned corresponding second weights;
[0096] Step 132: According to the first strategy, assign a first numerical value to the first work experience,
[0097] Step 133: According to the second strategy, assign a second numerical value to the first educational background,
[0098] Step 134: According to the third strategy, assign a third numerical value to the first professional skills,
[0099] Step 135: Based on the first numerical value, the second numerical value, and the third numerical value, perform numerical processing on the first key information.
[0100] In step 131, the subjective weighting method can be used, that is, through multiple rounds of anonymous questionnaires, the opinions of experts are made to tend to be consistent, so as to determine the weight values of the first work experience, the first educational background, and the first professional skills.
[0101] In step 135, the purpose of numerical processing of the first key information is achieved by multiplying the first numerical value, the second numerical value, and the third numerical value by the corresponding second weights.
[0102] In step 140, as Figure 4 shown, it includes the following steps:
[0103] Step 141: Normalize all types of the first key information to obtain a work experience index value, an educational background index value, and a professional skills index value;
[0104] Step 142: Based on the historical relevant text, determine the first weight values of the work experience index, the educational background index, and the professional skills index;
[0105] Step 143: According to the work experience index value, the educational background index value, the professional skills index value, and the corresponding first weight values, obtain the candidate ability portrait.
[0106] As Figure 5 shown, step 142 includes the following steps:
[0107] Step 1421: In the historical relevant text, screen the numerically processed second key information of the operators corresponding to the historical job selection results, and the second key information includes the second work experience, the second educational background, and the second professional skills;
[0108] Step 1422: Input the second key information into the Q-Learning algorithm model to obtain the first weight value. The Q-Learning algorithm model is configured to, after receiving the second key information, input the second key information into the reward function of the Q-Learning algorithm model, update the Q-table of the Q-Learning algorithm model based on the calculation result of the reward function, and obtain the first weight value according to the updated Q-table.
[0109] In this embodiment, through Step 1422, the first weight value is continuously updated based on the historical relevant texts of the candidates who have been promoted. The main reason is that: the candidates who have been promoted are more suitable candidates for the corresponding positions, indicating that their second work experience, second educational background, and second professional skills are more in line with the requirements of the corresponding positions. The weights calculated based on the second work experience, second educational background, and second professional skills of multiple candidates who have been promoted are more in line with the requirements of position selection.
[0110] The Q-Learning algorithm model in Step 1422 can be:
[0111] Using the Q-learning algorithm to adjust and optimize the weights of the ability profile based on the historical work data of the operators, the idea of reinforcement learning can be borrowed. The various abilities of the operators can be regarded as different "actions", and the performance in the historical work data is used as a "reward" to guide the learning process. The following is a possible method framework:
[0112] I. Initialize the ability profile and the Q-table
[0113] Define the ability profile:
[0114] According to the business requirements, determine the key abilities that need to be evaluated for the operators, such as communication ability, technical ability, teamwork ability, etc.
[0115] Assign an initial weight to each ability, and these weights can be set based on expert experience or business requirements.
[0116] Initialize the Q-table:
[0117] Create a Q-table, where the rows represent different states (which can be regarded as different work scenarios or tasks of the operators here), and the columns represent different abilities (actions).
[0118] Initialize all the values in the Q-table to 0 or a small value.
[0119] II. Collect historical work data
[0120] Collect the historical work data of the operators, including their performance in different tasks, the number of tasks completed, quality, efficiency, etc.
[0121] These data will be used as a "reward" signal to update the Q-table.
[0122] III. Design the reward function
[0123] Based on the historical work data, design a reward function that can reflect the performance of the operators in each task based on different capabilities.
[0124] The reward function can be discrete (such as a fixed reward based on task completion) or continuous (such as a weighted reward based on task completion quality).
[0125] IV. Application of the Q-learning algorithm
[0126] State selection:
[0127] At each time step, select the state according to the current work task or scenario.
[0128] Action selection:
[0129] Select an action (i.e., a capability) according to the ε-greedy policy. The value of ε determines the balance between exploration (randomly selecting a capability) and exploitation (selecting the current optimal capability).
[0130] Execute the action and observe the result:
[0131] Execute the selected action (capability) and observe the actual performance of the operator in this task. [[ID=3�]]
[0132] Calculate the reward:
[0133] Use the reward function to calculate the reward obtained by the operator after executing this action.
[0134] Update the Q-table:
[0135] According to the reward function of Q-learning (such as Q(s,a)←Q(s,a)+α[r+γmax_a'Q(s',a')-Q(s,a)]), use the obtained reward and the next state to update the value of the corresponding state-action pair in the Q-table.
[0136] V. Adjust and optimize the weights of the capability profile
[0137] After multiple iterations, the Q-table will contain the expected return of each state-action pair.
[0138] The weights of the ability profile can be adjusted according to the values in the Q-table. Specifically, for each ability (action), calculate its average Q-value across all states, and then use these average Q-values as the new weights.
[0139] VI. Continuous Iteration and Optimization
[0140] As new historical work data is collected, the Q-table can be continuously updated and the weights of the ability profile adjusted.
[0141] The effectiveness of the ability profile can be periodically re-evaluated and necessary adjustments made according to business requirements.
[0142] The reward function in step 1422 can be:
[0143]
[0144] A t is the ability selected at time step t;
[0145] S t is the state at time step t, which can represent the operator's current working environment, task type, or any factor affecting performance;
[0146] R(S t ,A t ) is the reward obtained for performing action A t in state S t ;
[0147] P i is the i-th performance indicator in the second key information (such as task completion rate, error rate, efficiency score, etc.);
[0148] W i is the weight corresponding to the i-th performance indicator. These weights may be set based on expert experience or business requirements initially and will be adjusted during the Q-Learning process;
[0149] N is the total number of performance indicators;
[0150] f i (P i (A t ,S t )) is the function value of the i-th performance indicator P i when performing action A t in state S t . This function maps the performance indicator to the reward space. For example, if P i is the task completion rate, then f i can be a linear function that converts the completion rate to a reward value (such as multiplying the completion rate by a constant).
[0151] In step 150, the specific process of training the preset clustering model may include: obtaining the ability portraits of a large number of promoted workers, dividing these ability portraits into a training set and a test set, inputting them into the initialized clustering model. After training, the clustering model will cluster to obtain several clusters. Each cluster includes a clustering center, and each cluster represents a type of job rank, such as management type, skill type, administrative type, etc. And each clustering center can represent the optimal ability portrait required for the corresponding job rank.
[0152] Therefore, in step 160, calculate the matching degree between the candidate's ability portrait and the clustering center. If a certain candidate has the highest matching degree with the clustering center representing the management type, then this candidate is suitable for promotion to a management type position; if a certain candidate has the highest matching degree with the clustering center representing the skill type, then this candidate is suitable for promotion to a skill type position.
[0153] After step 160, step 170 is also included.
[0154] Step 170 is: taking the candidate ability portrait corresponding to the job selection result as the training set and inputting it into the clustering model to correct the clustering center.
[0155] In step 170, after determining the selection result, input the ability portraits of the selected candidates again
[0156] Step 170 is: taking the candidate ability portrait corresponding to the job selection result as the training set and inputting it into the clustering model to achieve the purpose of correcting the clustering center.
[0157] In this embodiment, after taking the candidate ability portrait corresponding to the job selection result as a node and inputting it into the clustering model, the method further includes:
[0158] S1: Based on all the nodes of the clustering model, obtain the initial clustering center. Among them, the first work experience corresponding to the initial clustering center is the median value of the first work experience of all nodes; the first educational background corresponding to the initial clustering center is the median value of the first educational background of all nodes; the first professional skill corresponding to the initial clustering center is the median value of the first professional skills of all nodes;
[0159] S2: Calculate the first Euclidean distance between all nodes and the initial clustering center, and screen the nodes whose first Euclidean distance is less than or equal to the first distance threshold to obtain the screened nodes;
[0160] S3: Calculate the weighted average of the first work experience, the weighted average of the first educational background, and the weighted average of the first professional skills of all screened nodes. Based on the weighted average of the first work experience, the weighted average of the first educational background, and the weighted average of the first professional skills, form a weighted node, and calculate the second Euclidean distance between the weighted node and the initial cluster center;
[0161] S4: If the second Euclidean distance is less than or equal to the second distance threshold, then set the weighted node as the corrected cluster center; if the second Euclidean distance is greater than the second distance threshold, then set the weighted node as the initial cluster center, and repeat S2 and S3 until a weighted node with a second Euclidean distance less than or equal to the second distance threshold from the initial cluster center is obtained.
[0162] In S1, the median values of all the first work experience, the first educational background, and the first professional skills are used to obtain the initial cluster center. It is not sensitive to outliers, can reflect the mainstream data distribution, and represents the ability level of typical candidates, making the initial cluster center more stable and accelerating the convergence of the algorithm (the convergence condition in this embodiment is that the second Euclidean distance from the initial cluster center is less than or equal to the second distance threshold).
[0163] According to S1 - S4, based on the first key ability parameters of the selected successful candidates, the dynamic correction of the cluster center is realized, making the cluster center more accurate, thereby improving the accuracy of candidate selection.
[0164] The above are only the embodiments of the present application and are not used to limit the protection scope of the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims
1. A method for selecting job positions for personnel, characterized in that, The method includes: extracting relevant texts of the operator from a database, where the relevant texts include a resume, professional skill assessment results, and past work performance records; extracting first key information from the relevant texts, where the first key information includes first work experience, first educational background, and first professional skills; numerically processing the first key information based on a preset quantization strategy; generating a candidate ability profile of the operator according to the numerically processed first key information; calculating the matching degree between the ability profile and the clustering center obtained according to a preset clustering model; determining the job selection result of the operator based on the matching degree.
2. The method according to claim 1, wherein: After determining the job selection result of the operator according to the clustering result, the method further includes: using the candidate ability profile corresponding to the job selection result as a node to input into the clustering model for correcting the clustering center.
3. The method according to claim 1, characterized in that: The generating a candidate ability profile of the operator according to the numerically processed first key information includes: performing normalization processing on all types of the first key information to obtain a work experience index value, an educational background index value, and a professional skill index value; determining first weight values of the work experience index, the educational background index, and the professional skill index based on historical relevant texts; obtaining the candidate ability profile according to the work experience index value, the educational background index value, the professional skill index value, and the corresponding first weight values.
4. The method according to claim 3, characterized in that: The determining first weight values of the work experience index, the educational background index, and the professional skill index based on historical relevant texts includes: obtaining second key information after numerical processing of the operator corresponding to the historical job selection result from historical relevant texts, where the second key information includes second work experience, second educational background, and second professional skills; inputting the second key information into a Q-Learning algorithm model to obtain the first weight values. The Q-Learning algorithm model is configured to, after receiving the second key information, input the second key information into a reward function of the Q-Learning algorithm model, update a Q table of the Q-Learning algorithm model based on a calculation result of the reward function, and obtain the first weight values according to the updated Q table.
5. The method according to claim 4, wherein: The reward function is: A t The ability selected at time step t; S t is the state at time step t, which can represent the current working environment of the operator, the task type, or any factor affecting performance; R(S t ,A t ) is the reward obtained when performing action A t in state S t ; P i is the i-th performance indicator in the second key information; W i is the weight corresponding to the i-th performance indicator. These weights may be set based on expert experience or business requirements initially and will be adjusted during the Q-Learning process; N is the total number of performance indicators; f i (P i (A t ,S t )) is the function value when the i-th performance indicator P i performs the action A t in the state S t . This function maps the performance indicator to the reward space.
6. The method according to claim 1, characterized in that: The extracting first key information from the relevant texts includes: performing word segmentation on the relevant texts to obtain a number of words; performing part-of-speech tagging on the words to obtain the part of speech of each word; identifying named entities of the words after part-of-speech tagging based on a domain dictionary or knowledge base; extracting the first key information based on a natural language processing model according to the named entities.
7. The method according to claim 6, characterized in that: The extracting the first key information based on a natural language processing model according to the named entities includes: Based on the natural language processing model, extract the named entities used to represent the year and job title to form the first work experience; extract the named entities used to represent the school name and education level to form the first education background; extract the named entities used to represent the performance score to form the first professional skill.
8. The method according to claim 1, characterized in that: The preset quantization strategy includes a first strategy, a second strategy, and a third strategy; Based on the preset quantization strategy, numerically process the first key information, including: Assign weights to the first key information so that the first work experience, the first education background, and the first professional skill are assigned corresponding second weights; According to the first strategy, assign a first numerical value to the first work experience, and the first strategy is to determine the first numerical value according to the working years corresponding to the first work experience; According to the second strategy, assign a second numerical value to the first education background, and the second strategy is to determine the second numerical value according to the education level corresponding to the first education background; According to the third strategy, assign a third numerical value to the first professional skill, and the third strategy is to determine the third numerical value according to all the performance scores corresponding to the first professional skill; Based on the first numerical value, the second numerical value, and the third numerical value, numerically process the first key information.
9. The method according to claim 2, characterized in that: After using the candidate ability portrait corresponding to the job selection result as a node and inputting it into the clustering model, the method further includes: S1: Based on all the nodes of the clustering model, obtain an initial clustering center. Among them, the first work experience corresponding to the initial clustering center is the median value of the first work experience of all the nodes; the first education background corresponding to the initial clustering center is the median value of the first education background of all the nodes; the first professional skill corresponding to the initial clustering center is the median value of the first professional skill of all the nodes; S2: Calculate the first Euclidean distance between all the nodes and the initial clustering center, and filter out the nodes whose first Euclidean distance is less than or equal to the first distance threshold to obtain the filtered nodes; S3: Calculate the weighted average of the first work experience, the weighted average of the first education background, and the weighted average of the first professional skill of all the filtered nodes. Based on the weighted average of the first work experience, the weighted average of the first education background, and the weighted average of the first professional skill, form a weighted node, and calculate the second Euclidean distance between the weighted node and the initial clustering center; S4: If the second Euclidean distance is less than or equal to the second distance threshold, then set the weighted node as the corrected clustering center; if the second Euclidean distance is greater than the second distance threshold, then set the weighted node as the initial clustering center, and repeat S2 and S3 until a weighted node with a second Euclidean distance less than or equal to the second distance threshold from the initial clustering center is obtained.