Talent data and post matching method based on weight adjustment and related equipment

Abstract

The invention discloses a talent data and post matching method based on weight adjustment and related equipment, which can effectively improve the matching accuracy, help recruiters to quickly position adaptive candidates, reduce the manual screening cost and improve the recruitment efficiency, thereby meeting the requirements of quick recruitment of enterprises and efficient job hunting of talents. The method comprises the following steps: capturing multi-source data by utilizing a web crawler, and carrying out natural language processing on the multi-source data to obtain structured data; post time sequence characteristics of the structured post data in the structured data are extracted; when it is determined that the post time sequence features meet post requirements, performing multi-dimensional analysis on the post time sequence features to correct deviations and fuse weights, and then generating target post time sequence features; extracting talent features of the structured talent data in the structured data; and converting the target post time sequence features and the talent features into target vectors, and inputting the target vectors into a deep learning model to obtain a prediction matching degree of talent data and posts.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a talent data and job matching method based on weight adjustment and related equipment. Background Art

[0002] In the current job market, both companies and job seekers are in urgent need of improving the efficiency and accuracy of recruitment. Companies want to accurately recruit suitable talents, reduce costs and improve efficiency, and job seekers also hope to be accurately matched to suitable positions to increase their chances of successful job search.

[0003] Therefore, the current recruitment process uses talent data to match job information, which can effectively save time in talent data screening. However, most current matching algorithms use fixed weights, making it difficult to flexibly adjust to changing circumstances. For example, traditional methods often use static weights for factors like job skill requirements and work experience, failing to adapt to changing job requirements brought about by industry development and failing to fully account for the dynamic improvement of talent's own capabilities.

[0004] That is, the existing matching technology is insufficient in its ability to process real-time updated talent data and job information, resulting in delayed and inaccurate matching results, which cannot meet the needs of companies for rapid recruitment and talents for efficient job hunting. Summary of the Invention

[0005] In order to solve the above technical problems, the present application provides a talent data and job matching method based on weight adjustment and related equipment.

[0006] The technical solution provided in this application is described below: In a first aspect, the present application provides a method for matching talent data with positions based on weight adjustment, the method comprising: Using a web crawler to crawl multi-source data and performing natural language processing on the multi-source data to obtain structured data, wherein the structured data includes structured job data and structured talent data; Extracting job temporal features of structured job data from the structured data; When it is determined that the position time series characteristics meet the position requirements, the position time series characteristics are subjected to multi-dimensional analysis to correct deviations and integrate weights to generate target position time series characteristics; extracting talent features of structured talent data from the structured data; After converting the target position time series characteristics and talent characteristics into target vectors, the vectors are input into a deep learning model to obtain the predicted matching degree between talent data and positions; Obtaining recruitment feedback data, and obtaining actual matching degree between talent data and positions based on the recruitment feedback data; Comparing the predicted matching degree with the actual matching degree to obtain a comparison result; Adjusting the feature weight of the target job time series feature according to the comparison result to obtain an adjusted job feature; A visual matching report is generated based on the adjusted job characteristics and talent characteristics.

[0007] Optionally, extracting the position time series features of the structured position data in the structured data includes: Determining time-related feature dimensions based on the structured job data; Extracting original time series information containing time tags from the structured job data based on the time-related feature dimension; The original time series information is subjected to standardization processing such as supplementing missing values, removing duplicate records, and unifying the format to obtain processed information; The processed information is converted into the structured job time series features.

[0008] Optionally, when it is determined that the position time series characteristics meet the position requirements, the position time series characteristics are subjected to multi-dimensional analysis to correct deviations and integrate weights to generate target position time series characteristics, which is characterized by including: Determine whether the job timing characteristics meet the preset job requirements; If so, analyze the time series characteristics of the position from multiple dimensions, including time granularity, data distribution, and feature correlation, to identify and correct deviations caused by outliers, missing values, and differences in statistical caliber; Based on job attributes and historical matching data, initial weights are assigned to the revised job time series features of each dimension and then integrated; The weighted job time series features are integrated to generate the target job time series features.

[0009] Optionally, extracting talent features of structured talent data from the structured data includes: Determining characteristic dimensions based on the structured talent data, the characteristic dimensions including basic attributes, skills and qualifications, resume and experience, and performance; extracting original talent information from the structured talent data according to the feature dimensions; After verifying the authenticity of the original talent information and standardizing it to a unified format, the target talent information is obtained; The target talent information is converted into structured talent characteristics.

[0010] Optionally, obtaining recruitment feedback data and obtaining the actual matching degree between talent data and positions based on the recruitment feedback data includes: Determine the scope of recruitment feedback data, which includes interview evaluations, probationary period assessment results, and job suitability scores; Obtaining recruitment feedback data according to the feedback data range; Converting the recruitment feedback data into quantitative scores to obtain quantitative feedback data; The actual matching degree between the talent data and the position is calculated based on the quantitative feedback data.

[0011] Optionally, comparing the predicted matching degree with the actual matching degree to obtain a comparison result includes: Determine the deviation value indicator of the difference between the predicted match and the actual match; Substituting the predicted matching degree and the actual matching degree into the deviation value index for numerical calculation to obtain an index value; A comparison result is obtained according to the indicator value, wherein the comparison result includes difference data between the predicted matching degree and the actual matching degree.

[0012] Optionally, adjusting the feature weight of the target job time series feature according to the comparison result to obtain the adjusted job feature includes: Determine the bias-weight mapping relationship; Determining a feature weight adjustment rule based on the degree of difference in the comparison result according to the deviation-weight mapping relationship; Associating the difference data in the comparison result with each sub-feature of the target position time series feature, and locating the feature item corresponding to the deviation; The weights of the located feature items are increased or decreased according to the feature weight adjustment rules; The weights of each sub-feature after adjustment are integrated to form the adjusted job feature.

[0013] A second aspect of the present application provides a talent data and job matching system based on weight adjustment, the system comprising: a first acquisition unit, configured to crawl multi-source data using a web crawler and perform natural language processing on the multi-source data to obtain structured data, wherein the structured data includes structured job data and structured talent data; A first extraction unit is used to extract the position time series features of the structured position data in the structured data; A first generating unit is configured to, when it is determined that the position time series feature meets the position requirements, perform multi-dimensional analysis on the position time series feature, correct deviations and fuse weights, and generate a target position time series feature; a second extraction unit, configured to extract talent features of structured talent data from the structured data; A second acquisition unit converts the target position time series characteristics and talent characteristics into a target vector, and inputs the target vector into a deep learning model to obtain a predicted matching degree between the talent data and the position; A third acquisition unit is used to acquire recruitment feedback data and obtain the actual matching degree between talent data and positions based on the recruitment feedback data; a comparing unit, configured to compare the predicted matching degree with the actual matching degree to obtain a comparison result; an adjusting unit, configured to adjust the feature weight of the target job time series feature according to the comparison result to obtain an adjusted job feature; The second generating unit generates a visual matching report based on the adjusted job characteristics and talent characteristics.

[0014] A third aspect of the present application provides a talent data and job matching system based on weight adjustment, the system comprising: processor, memory, input and output units, and buses; The processor is connected to the memory, the input and output unit, and the bus; The memory stores a program, and the processor calls the program to execute the method as described in the first aspect and any one of the first aspects.

[0015] In a fourth aspect, the present application provides a computer-readable storage medium, on which a program is stored. When the program is executed on a computer, the method described in the first aspect and any one of the first aspects is executed.

[0016] It can be seen from the above technical solutions that this application has the following beneficial effects: 1. This application uses web crawlers to capture multi-source recruitment data, and combines natural language processing technology to convert unstructured text into structured job data and structured talent data, solving the problems of fragmented and chaotic format of traditional data acquisition, and providing a comprehensive and standardized data foundation for accurate matching.

[0017] 2. When extracting job temporal characteristics, this application corrects deviations and integrates weights through multi-dimensional analysis to generate target job characteristics that are more in line with actual needs, avoiding the limitations of traditional methods that ignore dynamic changes in jobs.

[0018] 3. This application converts target job characteristics and talent characteristics into vectors, inputs them into a deep learning model to predict matching, compares the predicted and actual matching degrees with recruitment feedback data, and dynamically adjusts the weights of job characteristics to solve the problems of static matching and deviation accumulation in traditional models, significantly improving prediction accuracy.

[0019] 4. This application generates a visual matching report based on the adjusted features, which intuitively displays the matching between talents and positions in multiple dimensions, helping recruiters to quickly locate suitable candidates, reduce manual screening costs, and improve recruitment efficiency, thereby meeting the needs of enterprises for rapid recruitment and talents for efficient job hunting. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in this application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 This is a schematic diagram of an embodiment of the talent data and job matching method based on weight adjustment of this application; Figure 2 This is a schematic diagram of another embodiment of the talent data and job matching method based on weight adjustment of this application; Figure 3 This is a schematic diagram of another embodiment of the talent data and job matching method based on weight adjustment of this application; Figure 4 This is a schematic diagram of another embodiment of the talent data and job matching method based on weight adjustment of this application; Figure 5 This is a schematic diagram of another embodiment of the talent data and job matching method based on weight adjustment of this application; Figure 6 This is a schematic diagram of another embodiment of the talent data and job matching method based on weight adjustment of this application; Figure 7 This is a schematic diagram of another embodiment of the talent data and job matching method based on weight adjustment of this application; Figure 8 This is a schematic diagram of an embodiment of a talent data and job matching system based on weight adjustment in this application; Figure 9 This is a schematic diagram of another embodiment of the talent data and job matching system based on weight adjustment of this application. DETAILED DESCRIPTION

[0022] Currently available matching technologies are insufficient in their ability to process real-time updated talent data and job information, resulting in delayed and inaccurate matching results, which cannot meet the needs of companies for rapid recruitment and talents for efficient job hunting.

[0023] Based on this, this application provides a talent data and job matching method based on weight adjustment and related equipment, which can effectively improve matching accuracy, help recruiters quickly locate suitable candidates, reduce manual screening costs, and improve recruitment efficiency, thereby meeting the needs of enterprises for rapid recruitment and talents for efficient job hunting.

[0024] See also Figure 1 In a first aspect, the present application provides a method for matching talent data with positions based on weight adjustment, the method comprising: 101. Using a web crawler to crawl multi-source data, and performing natural language processing on the multi-source data to obtain structured data, wherein the structured data includes structured job data and structured talent data; 102. Extracting job time series features of structured job data from the structured data; 103. When it is determined that the position time series characteristics meet the position requirements, the position time series characteristics are subjected to multi-dimensional analysis to correct deviations and integrate weights to generate target position time series characteristics; 104. Extracting talent features of structured talent data from the structured data; 105. After converting the target position time series characteristics and talent characteristics into a target vector, input the vector into a deep learning model to obtain the predicted matching degree between the talent data and the position; 106. Obtain recruitment feedback data, and obtain the actual matching degree between talent data and positions based on the recruitment feedback data; 107. Compare the predicted matching degree with the actual matching degree to obtain a comparison result; 108. Adjust the feature weight of the target position time series feature according to the comparison result to obtain an adjusted position feature; 109. Generate a visual matching report based on the adjusted job characteristics and talent characteristics.

[0025] In an embodiment of the present application, a web crawler is first used to crawl multi-source data, and the multi-source data is subjected to natural language processing to obtain structured data. The structured data includes structured job data and structured talent data. Then, the job time series features of the structured job data in the structured data are extracted. When it is determined that the job time series features meet the job requirements, the job time series features are subjected to multi-dimensional analysis to correct deviations and integrate weights to generate target job time series features. Talent features of the structured talent data in the structured data are further extracted. The target job time series features and talent features are then converted into target vectors, which are then input into a deep learning model to obtain a predicted matching degree between the talent data and the job. After obtaining recruitment feedback data, the team then uses this data to determine the actual match between the talent data and the position. The team then compares the predicted match with the actual match to obtain a comparison result. Based on the comparison result, the feature weights of the target position's time series features are adjusted to obtain the adjusted position features. Finally, a visual match report is generated based on the adjusted position features and talent features.

[0026] In step 101, a web crawler is first used to capture data from multiple sources. These sources include job postings from major recruitment websites, resumes from professional talent databases, official recruitment requirements from companies, and talent market reports from relevant industries, among other sources. The web crawler automatically accesses these data sources, collecting large amounts of unstructured information, such as job descriptions and work experience descriptions from resumes. Once the data is captured, natural language processing is performed on this multi-source data to convert the disorganized text into structured data.

[0027] Among them, the structured job data will clearly include fixed fields such as job title, industry, work location, educational requirements, professional skills requirements, and years of work experience. The structured talent data will cover specific information such as name, graduation school, major, years of work experience, skills mastered, past project experience, and expected position. After the structured data is processed, step 102 is executed.

[0028] In step 102, after obtaining the structured data, the job time series features can be extracted from the structured job data. Specifically, since job requirements are not fixed and will change over time, for example, the same job may have increased or decreased skill requirements or fluctuating salary ranges in different quarters due to business adjustments.

[0029] Therefore, it is necessary to sort out structured job data in chronological order and analyze the changes in job information at different time nodes, so as to extract characteristics that change over time, such as the update frequency of job skill requirements, the changing trend of salary levels, the increase and decrease in the number of recruits, etc. These characteristics can well reflect the dynamic changes in job requirements.

[0030] In step 103, after extracting the job temporal features, it is necessary to first determine whether these features meet the job requirements. The job requirements include both the company's core demand standards for the job and the general standards for similar jobs in the industry. If it is determined that the requirements are met, it is necessary to conduct a multi-dimensional analysis of these job temporal features, checking whether there are deviations from multiple angles such as the rationality of skill matching and the adaptability of experience requirements. For example, some temporal features may be abnormal due to accidental errors during data capture and need to be corrected. At the same time, it is also necessary to perform fusion weighting processing based on the importance of different features in the job requirements. For example, for technical positions, the temporal features of skill updates are more important than the salary fluctuation features, so they will be given a higher weight. Therefore, after the above processing, the target job temporal features that can reflect the core requirements of the position can be generated.

[0031] In step 104, while processing job time series characteristics, talent characteristics are also extracted from structured talent data. Talent characteristics are the attributes of a talent, including professional proficiency, work experience, project experience, the authority of certifications, and personal professionalism. By mining and integrating structured talent data, scattered information is refined into talent characteristics that reflect the talent's comprehensive qualities and capabilities.

[0032] In step 105 , after processing in steps 103 and 104 , the target position temporal features and talent features are obtained, and the target position temporal features and talent features need to be further converted into a target vector.

[0033] Because deep learning models cannot directly process text or discrete feature information, they need to map the target job's temporal features and talent characteristics into a high-dimensional vector space. This allows each feature to be represented as a vector, and the distance or similarity between vectors can reflect the degree of correlation between the features. These target vectors are then fed into a pre-trained deep learning model. Through calculations using a multi-layer neural network, the deep learning model learns the relationship between job and talent characteristics, and outputs a predicted match between the talent data and the job.

[0034] In step 106, after the predicted match is obtained, it is combined with recruitment feedback data. This feedback data comes from the actual recruitment process, such as the company's evaluation of the talent after the interview, the talent's response to accepting or rejecting the job, and the talent's performance evaluation during the probationary period. Based on this recruitment feedback data, the talent's actual work performance meets the job requirements, thereby determining the actual match between the talent data and the job, providing a basis for verifying the accuracy of the predicted match.

[0035] In step 107, after obtaining the actual matching degree and the predicted matching degree, the predicted matching degree and the actual matching degree are compared, and a specific comparison result is obtained by calculating a difference index between the two.

[0036] In step 108, the feature weights of the target position's temporal features are adjusted based on the comparison results. For example, if a position's temporal features were given a high weight during prediction but had little impact on the actual matching results, the weight of that feature should be reduced. Conversely, if a feature played a key role in the actual matching but had a low weight, its weight should be increased. After these adjustments, adjusted position features that better reflect the actual recruitment situation can be obtained.

[0037] In step 109, a visual matching report is generated based on the adjusted job characteristics and talent characteristics. The report uses a variety of chart forms such as bar charts, line charts, heat maps, etc., combined with text descriptions, to intuitively display the matching status of talent characteristics and adjusted job characteristics, such as which characteristics have a high matching degree, which ones have gaps, and the specific size of the gaps. The visual report can enable corporate recruiters and job seekers to clearly understand the matching results, providing a reference for recruitment decisions and job search choices.

[0038] Please refer to Figure 2 According to some embodiments of the present invention, the extraction of the position temporal features of the structured position data in the structured data in step 102 may specifically include, but is not limited to, the following: 201. Determine a time-related feature dimension based on the structured job data; 202. Extracting original time series information containing time stamps from the structured job data based on the time-related feature dimension; 203. Performing standardization processing on the original time series information by supplementing missing values, removing duplicate records, and unifying the format to obtain processed information; 204. Convert the processed information into the structured job time series features.

[0039] In the embodiment of the present application, time-related feature dimensions are determined based on structured job data. Specifically, structured job data contains a lot of time-related information, such as the job posting time, the expected recruitment deadline, the time range corresponding to the required years of work experience for the job, and historical records of recruitment cycles for similar jobs in the past. By sorting and analyzing this information, it is clear which dimensions are closely related to time, such as the update time dimension of job information, the time node dimension of each stage of the recruitment process, and the time span requirement dimension for talent work experience. These determined time-related feature dimensions will provide a clear direction for the subsequent extraction of original time series information containing time tags.

[0040] Based on the time-related feature dimension, raw time series information with time stamps is extracted from structured job data. Specifically, since the time-related feature dimension has been clarified, information with specific time stamps can be targeted and extracted from structured job data. For example, within the job information update time dimension, the specific time of each job update and the content changes before and after the update are extracted; within the recruitment deadline dimension, the deadlines for different recruitment batches are extracted; within the work experience requirement dimension, the time range involved in the job requirement of "3-5 years of work experience" and the corresponding skill improvement time nodes are extracted. All of this extracted information has clear time stamps and constitutes the raw time series information, which is the basis for subsequent standardization processing.

[0041] The original time series information is standardized by supplementing missing values, removing duplicate records, and unifying the format to obtain processed information. Specifically, the original time series information extracted from structured job data may have some problems, such as missing time tags, such as a job update record that does not indicate the specific update time, or duplicate records, where the same job information at the same time point is extracted multiple times; and inconsistent time formats, such as "year / month / day" and "month / day / year".

[0042] Therefore, standardization processing is required. For missing time tags, reasonable supplements should be made based on the logic of the previous and subsequent times. For example, the missing update time in the middle can be calculated based on the time of two adjacent updates. For duplicate records, they can be removed by comparing the time tags and information content. For time with inconsistent formats, they should be uniformly converted into the standard format of "year-month-day". The processed information obtained after these processes can ensure the accuracy and consistency of the time series information, and prepare for the conversion into structured job time series features.

[0043] The processed information is converted into structured job time series features. Specifically, this processed information is further integrated and refined by combining previously identified time-related feature dimensions. For example, updated information for the same job at different times is arranged chronologically to form a time series of job updates. Deadlines for different recruitment batches are organized to form a time series distribution of the recruitment cycle. The time range of work experience requirements is combined with the corresponding skill requirements to form a time series feature of skill growth over time.

[0044] Through such transformation, we can finally obtain structured job temporal characteristics, which can clearly reflect the changing patterns of jobs in the time dimension and provide a reliable basis for subsequent analysis and processing related to job temporal characteristics.

[0045] Please refer to Figure 3 According to some embodiments of the present invention, in step 103, when it is determined that the job time series characteristics meet the job requirements, the job time series characteristics are subjected to multi-dimensional analysis to correct deviations and fuse weights, and then target job time series characteristics are generated, which may specifically include, but is not limited to, the following: 301. Determine whether the position time sequence characteristics meet the preset position requirements; 302. If yes, analyze the time series characteristics of the position from multiple dimensions, including time granularity, data distribution, and feature correlation, to identify and correct deviations caused by outliers, missing values, and statistical caliber differences; 303. Based on job attributes and historical matching data, assign initial weights to the revised job time series features of each dimension and perform fusion; 304. Integrate the weighted job time series features to generate the target job time series features.

[0046] In practical applications, the first step is to determine whether the job's temporal characteristics meet the preset job requirements. These requirements encompass both the core job requirements established by the company based on its business development and the prevailing industry standards for similar positions, such as the average years of experience required for the position and required professional qualifications. By comparing the extracted job temporal characteristics against these preset requirements one by one, for example, checking whether the skill requirements reflected in the job temporal characteristics cover the preset core skills and whether salary trends fluctuate within the industry standard salary range, the team determines whether these characteristics meet the basic conditions for further processing.

[0047] If the job time series characteristics meet the preset requirements, the job time series characteristics will be analyzed in multiple dimensions from the dimensions of time granularity, data distribution and feature correlation, and then the deviations caused by outliers, missing values ​​and differences in statistical caliber will be identified and corrected.

[0048] From the perspective of time granularity, it is necessary to check the consistency of features under different time units, such as whether the trends of monthly data and quarterly data are smoothly connected, and whether there are gaps or abnormal fluctuations caused by different data collection frequencies.

[0049] In terms of data distribution, by drawing distribution graphs and other methods, we can observe whether the distribution of characteristic values ​​conforms to normal rules. For example, whether the salary time series characteristics of a certain position have extreme values ​​that are far higher or far lower than the reasonable range, these may be outliers.

[0050] Feature correlation analysis explores the intrinsic connections between the temporal characteristics of different positions, such as whether there is a reasonable correlation between the update of job skill requirements and salary adjustments. If there is a contradictory correlation, it may be caused by data anomalies.

[0051] For identified outliers, interpolation or deletion methods are used for correction. For missing values, the mean filling method is used to fill them according to their position and characteristic attributes in the time series. For deviations caused by differences in statistical caliber of different data sources, such as different ways of calculating work experience on different platforms, unified statistical standards are needed to ensure the consistency and accuracy of the data.

[0052] After completing the deviation correction, initial weights are assigned to the corrected temporal characteristics of each dimension of the position based on position attributes and historical matching data, and then they are integrated. Among them, the position attributes determine the inherent importance of different characteristics. For example, for management positions, the weight of temporal characteristics reflecting team coordination ability and strategic planning experience should be higher than the temporal characteristics of basic execution skills; for technical positions, the updating temporal characteristics of core technical capabilities may be more critical.

[0053] At the same time, we need to further combine historical matching data and analyze past cases of successful matching between this or similar positions and talents to observe which time series characteristics play a decisive role in the matching process. For example, in historical matching for a certain position, the skill update frequency is highly correlated with the success of the match. In this case, its proportion will be appropriately increased when allocating initial weights.

[0054] After assigning initial weights, a weighted summation fusion method is used to integrate the weighted position time series features to generate the target position time series features. During this integration process, it is important to ensure that the weighted fusion features complement and synergize with each other to accurately capture the dynamic core requirements of the position.

[0055] For example, after weighting, the core skill update and salary adjustment timing characteristics of a particular position may hold higher weights. When integrating these characteristics, it's important to emphasize the impact of their synergistic changes on job demand, while also considering the auxiliary information reflected by other, lower-weighted characteristics. Through this integration, the resulting target position timing characteristics can more accurately and comprehensively represent the actual demand for the position at different time points, paving the way for subsequent matching analysis with talent characteristics.

[0056] Please refer to Figure 4 According to some embodiments of the present invention, the talent features of the structured talent data extracted from the structured data in step 104 may specifically include, but are not limited to, the following: 401. Determine characteristic dimensions based on the structured talent data, where the characteristic dimensions include basic attributes, skills and qualifications, resume and experience, and performance; 402. Extracting original talent information from the structured talent data based on the feature dimensions; 403. After verifying the authenticity of the original talent information and standardizing the format to obtain target talent information; 404. Convert the target talent information into structured talent characteristics.

[0057] Based on the above plan, I will explain the steps in order from 401 to 404, combined with the processing logic of structured talent data, step by step to ensure that each step is naturally connected.

[0058] In the embodiment of the present application, characteristic dimensions are determined based on structured talent data, and these characteristic dimensions specifically include basic attributes, skills and qualifications, resumes and experience, and performance. Among them, basic attributes involve the basic identity and background information of talents, which are the initial positioning of talents. Skills and qualifications reflect the professional capabilities and recognitions possessed by talents, and are a key reference for job matching. Resume and experience reflect the growth trajectory and practical experience of talents, which can reflect their potential to adapt to the position. Performance directly demonstrates the achievements of talents in past work and is a strong proof of their ability. By determining these four characteristic dimensions, the key information in structured talent data can be comprehensively and systematically covered.

[0059] After determining the characteristic dimensions, we can extract raw talent information from structured talent data based on these characteristic dimensions. Specifically, from the basic attribute dimension, we will extract information such as name, age, gender, education, major, and place of origin; from the skill qualification dimension, we will extract the name and proficiency of professional skills mastered, professional qualification certificates obtained, professional and technical titles, training courses attended, etc.; from the resume experience dimension, we will extract information such as the name of the unit where the employee worked, the position held, the start and end time of work, the main job responsibilities, the names and roles of projects participated in, internship experience, etc.; from the performance dimension, we will extract the performance appraisal results of past work, the rewards and punishments received, and the project results data. This extracted raw talent information is the specific filling of the four characteristic dimensions.

[0060] After extracting the raw talent information, it needs to be verified for authenticity and standardized to a uniform format to obtain the target talent information. Authenticity verification is performed to ensure the reliability of the information. For example, for academic qualifications, this can be verified through official channels such as the China Higher Education Student Information and Career Center (CHSICC). Work experience can be verified by contacting the original employer or verifying resignation certificates. For skills and qualifications, the certificate number, issuing agency, and validity period are verified.

[0061] Regarding format standardization, adjustments are made to address potential format differences in structured talent data from different sources. For example, academic qualifications are standardized to "bachelor's," "master's," and "doctoral," eliminating variations such as "undergraduate" and "graduated from a bachelor's degree." Skill proficiency levels are standardized into "entry," "proficient," and "masterful." Dates are standardized to "year-month-day." This standardization effectively eliminates redundancy and inconsistencies in the information, transforming raw talent information into accurate and standardized target talent information, paving the way for subsequent conversion into structured talent profiles.

[0062] After acquiring the target talent information, it can be converted into structured talent characteristics. This conversion process involves structuring and integrating the standardized target talent information according to four characteristic dimensions to form characteristic data with clear logic and a fixed format. For example, the target talent information in the basic attribute dimension will be converted into a structured data group containing various basic information; the skill qualification dimension will form a structured feature indexed by the skill name, including proficiency, certificate status, etc.; the resume experience dimension will be organized into a structured sequence arranged in chronological order, containing work details for each stage; and the performance dimension will be converted into a structured record containing specific performance indicators and results data. Through such a conversion, the final structured talent characteristics obtained not only retain the comprehensiveness and accuracy of the talent information, but also have a clear structure and logic, and can be effectively matched and analyzed with the previously generated target position time series characteristics.

[0063] Please refer to Figure 5 According to some embodiments of the present invention, obtaining recruitment feedback data in step 106 and obtaining the actual matching degree between talent data and positions based on the recruitment feedback data may specifically include, but are not limited to, the following: 501. Determine the scope of recruitment feedback data, which includes data types such as interview evaluations, probationary period assessment results, and job suitability scores; 502. Obtain recruitment feedback data according to the feedback data range; 503. Convert the recruitment feedback data into quantitative scores to obtain quantitative feedback data; 504. Calculate the actual matching degree between the talent data and the position based on the quantitative feedback data.

[0064] Following the above plan, I will start by determining the scope of recruitment feedback data and gradually explain the process of obtaining data, converting data, and calculating the actual match degree, maintaining the coherence and logic of the content.

[0065] In the embodiment of the present application, the scope of recruitment feedback data includes data types such as interview evaluation, probationary period assessment results and job adaptation score. Interview evaluation usually comes from the interviewer's record after communicating with the candidate, covering the candidate's professional skills performance, communication skills, adaptability, professional quality and other aspects of the evaluation content; the probationary period assessment result is a comprehensive assessment record of the company's work task completion, teamwork performance, and familiarity with the job within a period of time after the candidate joins the company; the job adaptation score is a score data given from the perspective of the candidate's work style, values ​​and the fit with the company and the position in combination with the specific requirements of the position. Clarifying these data types can ensure that the feedback data obtained subsequently is comprehensive and targeted.

[0066] After determining the scope of recruitment feedback data, recruitment feedback data can be obtained based on this scope. Specifically, for interview evaluation, it is necessary to retrieve the various evaluation records entered by the interviewer from the company's recruitment management system, including text descriptions and preliminary scores; the probationary period assessment results must be extracted from the employee assessment files of the human resources department. These files will record in detail the specific work completed by the candidate during the probation period, the results achieved, and the shortcomings.

[0067] Job fit scores are derived from evaluations by direct supervisors, feedback from colleagues, and comprehensive assessments by the HR department. Therefore, it's necessary to aggregate and collect these scattered scoring data from various channels. During this acquisition process, we ensure the integrity and authenticity of the data to avoid missing key information. We also verify the data against corresponding talent and job information to ensure that each piece of feedback is accurately linked to a specific talent and position.

[0068] After obtaining recruitment feedback data, since it may be in a variety of formats, ranging from textual descriptions to ambiguous ratings, it is necessary to convert it into quantitative scores to obtain quantifiable feedback data. Specifically, for textual descriptions in interview evaluations, such as "the candidate has solid professional skills and good communication skills," "solid professional skills" can be assigned a score of 8-10, and "good communication skills" can be assigned a score of 7-9 based on the preset scoring criteria. The specific score is then determined based on the level of detail in the description. For probationary period assessment results, the completion of work tasks can be converted into corresponding scores based on indicators such as the quantity, quality, and efficiency of completed tasks, using the scoring system. For example, exceeding task completion can be assigned a score of 9-10, while nearly completing tasks can be assigned a score of 6-8. If the job fit rating is originally graded, such as "Excellent," "Good," and "Fair," these scores can be assigned 9-10, 7-8, and 4-6, respectively. This conversion unifies the originally diverse feedback data into a calculable quantitative score.

[0069] Based on the quantitative feedback data obtained, the actual matching degree between talent data and positions can be calculated. First, the weight of each quantitative feedback data in the actual matching degree calculation needs to be determined. For example, the probationary period assessment results directly reflect the actual work performance of the talent in the position, and the weight may be higher than the interview evaluation. The position adaptation score reflects the compatibility of long-term development and will also occupy a certain weight.

[0070] Then, the quantitative feedback data is weighted according to the weight of each data point to obtain a comprehensive score, which represents the actual match between the talent data and the position. For example, if the interview evaluation quantitative score is 80 points (weighted 30%), the probationary period assessment quantitative score is 85 points (weighted 50%), and the position fit score quantitative score is 78 points (weighted 20%), then the actual match is 80 × 30% + 85 × 50% + 78 × 20% = 82.1 points. This actual match can directly reflect the degree of fit between the talent and the position in actual work.

[0071] Please refer to Figure 6 According to some embodiments of the present invention, the comparison of the predicted matching degree and the actual matching degree in step 107 to obtain a comparison result may specifically include, but is not limited to, the following: 601. Determine a deviation value indicator of the difference between the predicted matching degree and the actual matching degree; 602. Substitute the predicted matching degree and the actual matching degree into the deviation value index to perform numerical calculation to obtain an index value; 603. Obtain a comparison result based on the indicator value, where the comparison result includes difference data between the predicted matching degree and the actual matching degree.

[0072] In the embodiment of the present application, a deviation value indicator of the difference between the predicted matching degree and the actual matching degree is determined. The setting of this indicator should be combined with the characteristics of the predicted matching degree and the actual matching degree, while referring to the experience of past comparative analysis and the actual business scenarios of job and talent matching.

[0073] Among them, since the predicted matching degree and the actual matching degree are usually presented in the form of percentages or scores, the deviation value indicator needs to be able to accurately quantify the degree of difference between the two. Common deviation value indicators can include absolute error, that is, the absolute value of the difference between the predicted matching degree and the actual matching degree, which can intuitively reflect the size of the gap between the two.

[0074] After determining the deviation index, the previously obtained predicted and actual match values ​​are substituted into these indexes for numerical calculations to obtain specific index values. For example, if the predicted match between a talent and a position is 80% and the actual match is 70%, substituting this into the absolute error index calculation yields a value of 10%. Substituting this into the relative error index calculation yields the ratio of 10% to 70%, which is approximately 14.29%. This calculation converts the abstract difference in match values ​​into concrete values, quantifying the gap between prediction and actual match, and providing the data foundation for subsequent comparative results.

[0075] Comparison results are generated based on the calculated index values. These results include the difference between the predicted and actual match degrees. For example, when the absolute error index value is 10% and the relative error is 14.29%, the comparison results clearly show that the predicted match between this group of talents and positions is 10% higher than the actual match, with a relative deviation of 14.29%. Furthermore, for multiple position-talent matching cases, the deviation index values ​​of all cases are summarized to analyze the overall difference trends. For example, if the absolute error of most cases is within 5%, it indicates that the predicted and actual match degrees are generally close. However, if the absolute error of individual cases exceeds 20%, it indicates that there are significant differences in these cases.

[0076] These comparison results, which include specific difference data and overall trends, can clearly reflect the degree of consistency between the predicted matching degree and the actual matching degree, and provide a reference basis for the subsequent adjustment of the feature weights of the target position's time series characteristics.

[0077] Please refer to Figure 7 According to some embodiments of the present invention, in step 108, adjusting the feature weight of the target job temporal feature according to the comparison result to obtain the adjusted job feature may specifically include, but is not limited to, the following: 701. Determine a bias-weight mapping relationship; 702. Determine, according to the deviation-weight mapping relationship, a feature weight adjustment rule based on the degree of difference in the comparison result; 703. Associate the difference data in the comparison result with each sub-feature of the target position time series feature, and locate the feature item corresponding to the deviation; 704. Increase or decrease the weight of the located feature item according to the feature weight adjustment rule; 705. Integrate the weights of each sub-feature after adjustment to form the adjusted position feature.

[0078] In the embodiments of this application, by analyzing the differences between past predicted and actual match degrees, and the magnitude of the adjustments to the weights of the time series features of each position each time a difference occurs, a corresponding pattern between the degree of deviation and the weight adjustment is found. For example, when a certain feature causes a large difference between the predicted and actual match degrees, the magnitude of the weight adjustment is often also large. This pattern is fixed in the form of a mapping table or functional relationship, forming a deviation-weight mapping relationship.

[0079] Once the deviation-weight mapping relationship is determined, this relationship can be used to develop feature weight adjustment rules based on the degree of difference in the comparison results. For example, if the mapping relationship shows that when the difference is between 10% and 20%, the corresponding feature weight needs to be reduced by 5%, then this can be used as an adjustment rule. When the difference exceeds 30%, the weight is reduced by 15%, and so on. These rules will clarify the specific weight adjustment direction and magnitude corresponding to different degrees of difference, ensuring a regular basis for subsequent feature weight adjustments.

[0080] Next, we need to correlate the difference data in the comparison results with the sub-features of the target position's time series characteristics to identify the feature items corresponding to the deviation. The difference data in the comparison results reflects the gap between the predicted and actual match levels, which is often caused by improper weighting of certain sub-features.

[0081] By analyzing the causes of the discrepancies in the data, we can trace them back to specific sub-features in the time series characteristics of the target position. For example, we found that the predicted match for a certain position was much higher than the actual match. Further analysis revealed that the weight of the sub-feature "project experience requirements" for that position was too high, causing this feature to be overemphasized in the prediction, while its importance in actual recruitment was not that high. Therefore, we identified this sub-feature as the feature item corresponding to the deviation.

[0082] After identifying the feature items corresponding to the deviation, their weights are adjusted according to the established feature weight adjustment rules. For example, if the sub-feature "Project Experience Requirements" is identified as the cause of the deviation, and the comparison results show a 25% discrepancy, then the weight of this sub-feature will be reduced by 10%, based on the adjustment rule that requires a 10% reduction in weight for a 20%-30% discrepancy. This adjustment brings the feature weights more in line with actual hiring practices and reduces matching bias caused by inappropriate weighting.

[0083] After adjusting the weights of each sub-feature, integrate all adjusted sub-feature weights to form the adjusted job profile. This integration process ensures that the new weights of each sub-feature can synergistically reflect the actual needs of the position. For example, after a position is adjusted, the weight of the "Core Skill Requirements" sub-feature is increased, while the weight of the "Project Experience Requirements" sub-feature is decreased. During the integration, the impact of these changes in weights on the overall needs of the position should be highlighted, while also taking into account the weights of other sub-features. The resulting adjusted job profile can more accurately match talent characteristics, improving matching accuracy.

[0084] See also Figure 8 The second aspect of this application provides a talent data and job matching system based on weight adjustment, including: The first acquisition unit 801 is configured to crawl multi-source data using a web crawler and perform natural language processing on the multi-source data to obtain structured data, wherein the structured data includes structured job data and structured talent data; A first extraction unit 802 is configured to extract job temporal features of structured job data from the structured data; The first generating unit 803 is configured to, when it is determined that the job time series feature meets the job requirements, perform multi-dimensional analysis on the job time series feature, correct deviations and fuse weights, and generate a target job time series feature; A second extraction unit 804 is configured to extract talent features of structured talent data from the structured data; The second acquisition unit 805 converts the target position time series characteristics and talent characteristics into a target vector, and inputs the target vector into a deep learning model to obtain a predicted matching degree between the talent data and the position; The third acquisition unit 806 is used to acquire recruitment feedback data and obtain the actual matching degree between talent data and positions based on the recruitment feedback data; A comparison unit 807 is configured to compare the predicted matching degree with the actual matching degree to obtain a comparison result; An adjusting unit 808 is configured to adjust the feature weight of the target job time series feature according to the comparison result to obtain an adjusted job feature; The second generating unit 809 generates a visual matching report based on the adjusted job characteristics and talent characteristics.

[0085] See also Figure 9 , this application also provides a talent data and job matching system based on weight adjustment, including: Processor 901, memory 902, input / output unit 903, bus 904; The processor 901 is connected to the memory 902 , the input and output unit 903 , and the bus 904 ; The storage 902 stores a program, and the processor 901 calls the program to execute any of the above methods.

[0086] The present application also relates to a computer-readable storage medium, on which a program is stored. When the program is run on a computer, the computer is caused to execute any of the above methods.

[0087] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0088] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0089] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0090] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0091] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, read-only memory), random access memory (RAM, random access memory), disk or optical disk, and other media that can store program code.

Claims

1. A talent data and job matching method based on weight adjustment, characterized in that: The method comprises: Using a web crawler to crawl multi-source data and performing natural language processing on the multi-source data to obtain structured data, wherein the structured data includes structured job data and structured talent data; Extracting job temporal features of structured job data from the structured data; When it is determined that the position time series characteristics meet the position requirements, the position time series characteristics are subjected to multi-dimensional analysis to correct deviations and integrate weights to generate target position time series characteristics; extracting talent features of structured talent data from the structured data; After converting the target position time series characteristics and talent characteristics into target vectors, the vectors are input into a deep learning model to obtain the predicted matching degree between talent data and positions; Obtaining recruitment feedback data, and obtaining actual matching degree between talent data and positions based on the recruitment feedback data; Comparing the predicted matching degree with the actual matching degree to obtain a comparison result; Adjusting the feature weight of the target job time series feature according to the comparison result to obtain an adjusted job feature; A visual matching report is generated based on the adjusted job characteristics and talent characteristics.

2. The talent data and job matching method based on weight adjustment according to claim 1 is characterized in that: Extracting the job time series features of the structured job data in the structured data includes: Determining time-related feature dimensions based on the structured job data; Extracting original time series information containing time tags from the structured job data based on the time-related feature dimension; The original time series information is subjected to standardization processing such as supplementing missing values, removing duplicate records, and unifying the format to obtain processed information; The processed information is converted into job time series features.

3. The talent data and job matching method based on weight adjustment according to claim 1 is characterized in that: When it is determined that the position time series characteristics meet the position requirements, the position time series characteristics are subjected to multi-dimensional analysis to correct deviations and integrate weights to generate target position time series characteristics, including: Determine whether the job timing characteristics meet the preset job requirements; If so, analyze the time series characteristics of the position from multiple dimensions, including time granularity, data distribution, and feature correlation, to identify and correct deviations caused by outliers, missing values, and differences in statistical caliber; Based on job attributes and historical matching data, initial weights are assigned to the revised job time series features of each dimension and then integrated; The weighted job time series features are integrated to generate the target job time series features.

4. The talent data and job matching method based on weight adjustment according to claim 1 is characterized in that: Extracting talent features of structured talent data from the structured data includes: Determining characteristic dimensions based on the structured talent data, the characteristic dimensions including basic attributes, skills and qualifications, resume and experience, and performance; extracting original talent information from the structured talent data according to the feature dimensions; After verifying the authenticity of the original talent information and standardizing it to a unified format, the target talent information is obtained; The target talent information is converted into structured talent characteristics.

5. The talent data and job matching method based on weight adjustment according to claim 1 is characterized in that: Obtaining recruitment feedback data, and obtaining the actual matching degree between talent data and positions based on the recruitment feedback data, including: Determine the scope of recruitment feedback data, which includes interview evaluations, probationary period assessment results, and job suitability scores; Obtaining recruitment feedback data according to the feedback data range; Converting the recruitment feedback data into quantitative scores to obtain quantitative feedback data; The actual matching degree between the talent data and the position is calculated based on the quantitative feedback data.

6. The talent data and job matching method based on weight adjustment according to claim 1 is characterized in that: Comparing the predicted matching degree with the actual matching degree to obtain a comparison result includes: Determine the deviation value indicator of the difference between the predicted match and the actual match; Substituting the predicted matching degree and the actual matching degree into the deviation value index for numerical calculation to obtain an index value; A comparison result is obtained according to the indicator value, wherein the comparison result includes difference data between the predicted matching degree and the actual matching degree.

7. The talent data and job matching method based on weight adjustment according to claim 1 is characterized in that: Adjusting the feature weight of the target position time series feature according to the comparison result to obtain the adjusted position feature includes: Determine the bias-weight mapping relationship; Determining a feature weight adjustment rule based on the degree of difference in the comparison result according to the deviation-weight mapping relationship; Associating the difference data in the comparison result with each sub-feature of the target position time series feature, and locating the feature item corresponding to the deviation; The weights of the located feature items are increased or decreased according to the feature weight adjustment rules; The weights of each sub-feature after adjustment are integrated to form the adjusted job feature.

8. A talent data and job matching system based on weight adjustment, characterized in that: The system comprises: a first acquisition unit, configured to crawl multi-source data using a web crawler and perform natural language processing on the multi-source data to obtain structured data, wherein the structured data includes structured job data and structured talent data; A first extraction unit is used to extract the position time series features of the structured position data in the structured data; A first generating unit is configured to, when it is determined that the position time series feature meets the position requirements, perform multi-dimensional analysis on the position time series feature, correct deviations and fuse weights, and generate a target position time series feature; a second extraction unit, configured to extract talent features of structured talent data from the structured data; A second acquisition unit converts the target position time series characteristics and talent characteristics into a target vector, and inputs the target vector into a deep learning model to obtain a predicted matching degree between the talent data and the position; A third acquisition unit is used to acquire recruitment feedback data and obtain the actual matching degree between talent data and positions based on the recruitment feedback data; a comparing unit, configured to compare the predicted matching degree with the actual matching degree to obtain a comparison result; an adjusting unit, configured to adjust the feature weight of the target job time series feature according to the comparison result to obtain an adjusted job feature; The second generating unit generates a visual matching report based on the adjusted job characteristics and talent characteristics.

9. A talent data and job matching system based on weight adjustment, characterized in that: The system comprises: processor, memory, input and output units, and buses; The processor is connected to the memory, the input and output unit, and the bus; The memory stores a program, and the processor calls the program to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program, and when the program is executed on a computer, the method according to any one of claims 1 to 7 is executed.

Images