A person-job matching system based on intelligent data analysis

Abstract

The present invention discloses a person-job matching system based on intelligent data analysis, comprising a matching degree training set acquisition module, used for acquiring a job matching degree training set D, wherein the matching degree training set acquisition module is built with a screening submodule, and the screening submodule is used for acquiring a second historical entrant from a first historical entrant; a self-supervised learning module, used for inputting the job matching degree training set into a pre-training model, and training to obtain a person-job matching model; a recommendation module, used for inputting a job-seeking feature vector of a current job seeker into the person-job matching model, and outputting a high-matching job; the present invention screens out a second historical entrant from the first historical entrant, and uses the features of the second historical entrant with a high matching degree that have been tested in practice as the training set of the person-job matching model, so that the model can learn excellent job-seeking features in the high-quality training set, thereby providing a higher quality person-job matching degree for recruiters or job seekers.

Description

Technical Field

[0001] The present application relates to the field of person-job matching, and specifically to a person-job matching system based on intelligent data analysis. Background Art

[0002] In the modern recruitment process, the person-job matching system has become an important tool to improve recruitment efficiency and matching. As companies' demand for efficient recruitment increases, traditional job matching methods have been unable to meet complex and diverse recruitment needs. Traditional methods usually rely on simplified matching rules, such as keyword searches and basic experience matching, which often fail to fully consider the complex relationship between job seekers' skills, experience, and job requirements.

[0003] With reference to the Chinese patent document with publication number CN118469531A, a resume text parsing method suitable for person-job matching recommendation is proposed to solve the problem of low matching degree of existing person-job matching. The method can quickly and accurately match jobs, making it convenient for individuals to find suitable jobs or companies to find suitable candidates with high matching degree.

[0004] However, the above-mentioned prior art lacks the process of using machine learning models to screen new employees; that is, when training the person-job matching model in the prior art, most of the training data obtained is from the massive job search data on the market, or the job search data of employees who have already been employed. However, due to the authenticity of the above-mentioned job search data, resume fraud, or delayed resume information, the training data of the person-job matching model cannot truly reflect whether the job seeker (employee) can truly match the job position, and ultimately the quality of the training data leads to the problem of low model matching. Summary of the invention

[0005] This embodiment provides a person-job matching system based on intelligent data analysis, and explores how to solve the problem that the existing person-job matching system is difficult to reflect whether the job seeker (employee) can truly match the job position, and ultimately the model matching degree is low due to the quality of training data.

[0006] The present invention provides a person-job matching system based on intelligent data analysis, comprising:

[0007] The matching training set acquisition module is used to acquire the job matching training set D. The training samples of the job matching training set D are N groups of matching data pairs. The matching data pairs are composed of job-seeking feature vectors. and its job matching tags composition;

[0008] ;

[0009] in, , represents the job-seeking feature vector of the i-th second historical employee, represents the set of real numbers with d characteristic dimensions; represents a job matching label of the i-th second historical employee, wherein the job matching label represents different job matching categories;

[0010] The matching degree training set acquisition module has a built-in screening submodule, and the screening submodule is used to obtain the second historical employee from the first historical employee;

[0011] A self-supervised learning module, used to input the job matching training set into a pre-trained model, and the pre-trained model is trained by self-supervised learning to obtain a person-job matching model;

[0012] The recommendation module is used to input the job-seeking feature vector of the current job seeker into the person-job matching model, and the person-job matching model outputs a job with a high matching degree for the current job seeker.

[0013] In some of the embodiments, the acquisition step of the matching degree training set acquisition module includes:

[0014] Obtain a sample training set of the first historical employee, wherein the training samples of the sample training set are the first historical employee and their job-seeking characteristics, wherein the job-seeking characteristics include: skill keyword frequency, skill matching, skill experience years, professional relevance, and commuting distance of the interview text;

[0015] Obtaining second historical job entrants and their predicted job matching degrees from the sample training set of the first historical job entrants;

[0016] Determine the job matching label of the second historical employee according to the predicted job matching degree of the second historical employee;

[0017] sorting the job-seeking features of the second historical job-holder according to a predefined order to construct the job-seeking feature vector;

[0018] Combining the job-seeking feature vector of the second historical employee and the corresponding job matching label into a matching degree data pair of the second historical employee;

[0019] The matching data pairs are collected to obtain the job matching training set of the second historical employee.

[0020] In some embodiments, the job-seeking characteristics of the first historical job seeker are obtained, and the job-seeking characteristics include: skill keyword frequency, skill matching, skill experience years, professional relevance, and commuting distance of the interview text, including:

[0021] Preset skill keywords, wherein the skill keywords include single keywords or combined keywords;

[0022] According to the skill keywords, skill keyword frequencies and skill level values ​​are extracted from the interview text, and a skill feature vector is constructed;

[0023] Calculating the skill matching degree according to the skill keyword frequency and the skill level value;

[0024] The years of skills and experience, professional relevance, and commuting distance were extracted from the interview texts of the first history entrants.

[0025] In some embodiments, obtaining the second historical job applicants and their predicted job matching degrees from the sample training set of the first historical job applicants includes:

[0026] Predefine the job contribution of each job-seeking feature in the job sought by the first historical entrant;

[0027] Preset the weight coefficient of each job-seeking feature according to the job contribution of the first historical job seeker in the job;

[0028] Calculating the predicted job matching degree of the first historical job seeker according to the weight coefficient of each of the job-seeking characteristics;

[0029] Obtain the actual job matching degree of the first historical employees after joining the company;

[0030] Comparing the actual job matching degree with the predicted job matching degree to obtain a matching deviation of the first historical employee;

[0031] The first historical employee corresponding to the matching deviation falling within the first threshold range is defined as the second historical employee.

[0032] In some embodiments, determining the job matching tag of the second historical employee according to the predicted job matching degree of the second historical employee includes:

[0033] In the predicted job matching degree of the second historical employee, K second thresholds of different levels are preset;

[0034] Define job matching tags corresponding to second thresholds of different levels, and create job matching mapping rules according to the second thresholds and their job matching tags;

[0035] According to the job matching mapping rule, a job matching label is determined for the second historical employee.

[0036] In some embodiments, the training step of the self-supervised learning module includes:

[0037] Using the job-seeking feature vector of the second historical employee as a feature variable and the job matching label as a target variable to optimize the model parameters of the pre-trained model;

[0038] Select multi-label classification cross entropy as the loss function for the training process;

[0039] Calculate the total cross entropy loss of the job matching labels according to the loss function;

[0040] Determine whether the total cross entropy loss is lower than the preset total loss threshold. If so, output the current model parameters of the pre-trained model.

[0041] Fine-tune the current model parameters to obtain the person-job matching model with optimal parameters.

[0042] In some embodiments, the loss function The expression is:

[0043] ;

[0044] Among them, K is the total number of job matching label categories, i represents the job matching training set The i-th training sample contains the i-th job-seeking feature vector Match the label with the i-th position , categories are defined as job matching labels Multiple different job types included; It is a pre-trained model For the job-seeking feature vector in the i-th training sample The predicted value of are the parameters of the pre-trained model, is the one-hot encoding of training sample i on the job matching label; It is a pre-trained model The predicted probability of the i-th training sample in the j-th category; Represents the cross entropy loss of the i-th training sample, which calculates the job matching label and predicted values The difference between.

[0045] In some of the embodiments, the fine-tuning expression for fine-tuning the current model parameters is:

[0046] ;

[0047] in, represents the optimal parameters after fine-tuning, Indicates that all possible parameters are found Minimize the average loss of the fine-tuning process, is the number of training samples in the validation set.

[0048] In some of the embodiments, the recommendation step of the recommendation module includes:

[0049] Get the job-seeking feature vector of the current job seeker;

[0050] Input the job-seeking feature vector of the current job seeker into the person-job matching model, and output the job with high matching degree for the current job seeker;

[0051] The person-job matching model has the optimal parameters after fine-tuning , which can recommend highly matching positions to recruiters or job seekers based on the job-seeking feature vector of the current job seeker; the expression is:

[0052] ;

[0053] in, Recommended job tags for current job seekers. is the job-seeking feature vector of the current job seeker.

[0054] Compared with the prior art, the person-job matching system based on intelligent data analysis of the present invention selects the second historical entrants from the first historical entrants, and uses the characteristics of the second historical entrants with high matching degree that have been verified in practice as the training set of the person-job matching model, so that the model can learn excellent job-seeking characteristics in the high-quality training set, thereby providing recruiters or job seekers with higher quality person-job matching. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A flow chart of a matching method of a person-job matching system based on intelligent data analysis in an embodiment provided by the present invention;

[0056] Figure 2 A structural block diagram of a person-job matching system based on intelligent data analysis in an embodiment provided by the present invention;

[0057] Figure 3 A training logic diagram of self-supervised learning in the embodiment provided by the present invention; DETAILED DESCRIPTION

[0058] In order to more clearly understand the purpose, technical solutions and advantages of the present application, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments.

[0059] Unless otherwise defined, the technical terms or scientific terms involved in this application shall have the general meaning understood by people with general skills in the technical field to which this application belongs. The words "one", "a", "the", "these" and the like in this application do not indicate a quantitative limitation, and they may be singular or plural. The terms "include", "comprise", "have" and any variants thereof involved in this application are intended to cover non-exclusive inclusions; for example, a process, method and system, product or device comprising a series of steps or modules (units) is not limited to the listed steps or modules (units), but may include unlisted steps or modules (units), or may include other steps or modules (units) inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that there may be three relationships, for example, "A and / or B" may mean: A exists alone, A and B exist at the same time, and B exists alone. Generally, the character " / " indicates that the objects associated with each other are in an "or" relationship. The terms "first", "second", "third", etc. in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0060] First, the prior art and related concepts involved in the embodiments of the present invention are described:

[0061] First historical entrants: job seekers who are actually hired;

[0062] Second historical entrants: historical entrants with high matching degree among the job seekers actually hired;

[0063] Example 1: Reference Figures 1 to 3 In an embodiment of the present invention, a person-job matching system based on intelligent data analysis is provided.

[0064] Figure 1 This is a flow chart of a matching method of a person-job matching system based on intelligent data analysis of the present invention, and the process includes:

[0065] Step S1: Obtain a job matching training set D, wherein the training samples of the job matching training set D are N sets of matching data pairs, wherein the matching data pairs are composed of job-seeking feature vectors and its job matching tags composition.

[0066] The expression of the job matching training set is:

[0067] ;

[0068] in, , represents the job-seeking feature vector of the i-th second historical employee, represents the set of real numbers with d characteristic dimensions; represents the job matching label of the i-th second historical employee, where the job matching label represents different job matching categories.

[0069] In this embodiment, the job matching training set comes from the job matching data set, and the job matching data set also includes a test set and a validation set.

[0070] In the embodiment of the present invention, step S1 specifically includes:

[0071] S1-1, obtain a sample training set of first-time job seekers, where the training samples of the sample training set are first-time job seekers and their job-seeking characteristics, including: skill keyword frequency, skill matching, skill experience years, professional relevance, and commuting distance in the interview text;

[0072] S1-2, in the sample training set of the first historical entrants, obtain the second historical entrants and their predicted job matching degree;

[0073] S1-3, determining a job matching label of the second historical employee according to the predicted job matching degree of the second historical employee.

[0074] In the embodiment of the present invention, step S1-1 specifically includes:

[0075] S1-1-1, preset skill keywords, wherein the skill keywords include single keywords or combined keywords.

[0076] Exemplarily, first, a skill keyword set is predefined , the set contains m skill keywords. Each skill keyword It can be a single keyword, such as "python", or a combination of keywords, such as "machine learning".

[0077] S1-1-2, based on the skill keywords, extract the skill keyword frequency and skill level value in the interview text, and construct a skill feature vector.

[0078] Next, define the skill keywords The frequency of occurrence in interview text d is , the frequency expression is:

[0079] ;

[0080] in, Indicates skill keywords The total number of times it appears in interview text d, Indicates the total number of words in the text, and the skill level value Is the first historical job entry for skill keywords The self-assessment level can be reflected in the interview text or oral introduction filled in; the skill level value Can be represented as a set ,in, Represents any skill keyword. represents the mth skill keyword; therefore, Represents any skill level value in the set s, Represents the mth skill level value in set s. Each skill level value Represents the proficiency of the first history entrants in mastering the skills; for example, each self-assessment level range can be assumed to be ; It can correspond to no such key skill, 5 corresponds to expert, etc.

[0081] After that, the skill keyword The corresponding skill level value Constructed as a skill feature vector , so that a composite feature that can represent skill association can be defined. The composite feature is in the form of a vector, so the composite feature can be defined as a skill feature vector. The expression is:

[0082] ;

[0083] S1-1-3, calculate the skill matching degree according to the skill keyword frequency and skill level value. The expression of skill matching degree is:

[0084] ;

[0085] in, It represents the skill matching degree, which is the quantitative representation of the skill feature vector in the interview process.

[0086] S1-1-4 extracts the years of skill experience, professional relevance, and commuting distance from the interview texts of the first history employees.

[0087] In the embodiment of the present invention, the interview text refers to the resume text filled out by the first historical employee during the job search process and all text contents answered during the job search process. Therefore, the method of extracting skill experience years, professional relevance and commuting distance can be oral expression, text extraction and the like.

[0088] In the embodiment of the present invention, step S1-2 specifically includes:

[0089] S1-2-1, predefine the contribution of each job-seeking feature to the position sought by the first historical job seeker;

[0090] Define the contribution vector of job-seeking features as ,in, It represents the contribution of the k-th job-seeking feature to the job sought.

[0091] S1-2-2, preset the weight coefficient of each job-seeking feature according to the job contribution of the first historical job seeker in the job;

[0092] Define the weight vector of job-seeking features as , its definition expression is:

[0093] ;

[0094] Among them, the weight coefficient is the weight coefficient, which is defined according to the job contribution of each job-seeking feature.

[0095] S1-2-3, based on the weight coefficient of each job-seeking feature, calculate the predicted job matching degree of the first historical entrant, and the predicted job matching degree The expression is:

[0096] ;

[0097] in, is the kth job-seeking characteristic value of the i-th first historical employee.

[0098] S1-2-4, obtaining the actual job matching degree of the first historical employee after joining the company;

[0099] Actual job matching It is the numerical data of the actual work performance of the first-time employees after joining the company. After being evaluated and scored by the position manager and input into the machine equipment for structure, the numerical actual position matching degree is formed. .

[0100] S1-2-5, comparing the actual job matching degree with the predicted job matching degree to obtain the matching deviation of the first historical entrant;

[0101] ;

[0102] in, Indicates the matching deviation.

[0103] S1-2-6, defining the first historical employee corresponding to the matching deviation falling within the first threshold range as the second historical employee.

[0104] That is, if the matching deviation falls within the predefined first threshold range , then the first historical entrant is defined as the second historical entrant.

[0105] In the embodiment of the present invention, step S1-3 specifically includes:

[0106] S1-3-1, in the predicted job matching degree of the second historical employee, K second thresholds of different levels are preset;

[0107] Exemplarily, the second threshold set is defined as , which is used to divide the different categories of job matching labels. Of course, the categories in this embodiment can be matching fitness, such as very mismatched, relatively matched, very matched, very matched, etc.

[0108] S1-3-2, defining the job matching tags corresponding to the second thresholds at different levels, and creating job matching mapping rules according to the second thresholds and the job matching tags;

[0109] The job matching mapping rules are:

[0110] ;

[0111] in, Indicates the predicted job match of the second historical entrant.

[0112] S1-3-3, determine a job matching label for the second historical employee according to the job matching mapping rule.

[0113] In the embodiment of the present invention, step S1 further includes:

[0114] S1-4, sort the job-seeking features of the second historical job-seekers in a predefined order and construct a job-seeking feature vector; that is, the skill feature vector of the second historical job-seekers is , Skill Matching , Skills and Experience , professional relevance , commuting distance Constructed as a job search feature vector ; Its construction expression is:

[0115] .

[0116] S1-5, combining the job-seeking feature vector of the second historical employee and its corresponding job matching label into a matching degree data pair of the second historical employee;

[0117] S1-6, collect matching data pairs to obtain the job matching training set of the second historical employees.

[0118] Through the above step S1, a training set for the pre-training model to perform job recommendation training is finally obtained.

[0119] Embodiment 2: The technical solution of this embodiment is different from that of Embodiment 1 in that:

[0120] Step S2: input the job matching training set into the pre-training model, and the pre-training model is trained by self-supervised learning to obtain a person-job matching model.

[0121] refer to Figure 3 , training for self-supervised learning, including:

[0122] The job-seeking feature vector of the second historical employee is used as the feature variable and the job matching label is used as the target variable to optimize the model parameters of the pre-trained model;

[0123] Select multi-label classification cross entropy as the loss function for the training process;

[0124] In this embodiment, the job matching training set D is used to pre-train the model Training is performed, and the training process optimizes the model parameters by minimizing the loss function .

[0125] The loss function is:

[0126] ;

[0127] Among them, K is the total number of job matching label categories, i represents the job matching training set The i-th training sample contains the i-th job-seeking feature vector Match the label with the i-th position , categories are defined as job matching labels Contains multiple different types of positions; that is, in the position matching training set In the example, the i-th position matches the label is regarded as the i-th job-seeking feature vector Job matching tags ; It is a pre-trained model For the job-seeking feature vector in the i-th training sample The predicted value of are the parameters of the pre-trained model, is the one-hot encoding of training sample i on the job matching label (i.e., the true label). For example, if a true label = 2 (good match), its one-hot encoding is [0, 0, 1, 0]; It is a pre-trained model The predicted probability of the i-th training sample in the j-th category; Represents the cross entropy loss of the i-th training sample, which calculates the job matching label and predicted values The difference between.

[0128] Calculate the total cross entropy loss of the job matching labels based on the loss function;

[0129] Therefore, the expression of the total cross entropy loss is:

[0130] ;

[0131] in, Represents the total cross entropy loss, which is used to measure the pre-training model is the overall difference between the predicted value and the job matching label, and N is the number of training samples in the training set.

[0132] Determine the total cross entropy loss Is it below the preset total loss threshold? , if it is lower than , then output the current model parameters of the pre-trained model;

[0133] Fine-tune the current model parameters to obtain a person-job matching model with optimal parameters.

[0134] The expression of the fine-tuning process is:

[0135] ;

[0136] in, represents the optimal parameters after fine-tuning, Indicates that all possible parameters are found Minimize the average loss of the fine-tuning process, is the number of training samples in the validation set.

[0137] For example, in this embodiment, a softmax regression model is used as a pre-training model. Therefore, the predicted probability of each category j by the pre-training model is The calculation process includes:

[0138] The pre-trained model calculates the unnormalized score for each input job feature vector in each category j. ;

[0139] ;

[0140] Among them, the unnormalized score represents the original output score of the pre-trained model for category j; represents the weight vector of category j, which defines the input job search feature vector A linear mapping to category j; is the bias term for category j, which represents the score adjustment scalar for category j.

[0141] Use the softmax function to convert the unnormalized scores Convert the input feature vector to the pre-trained model The predicted probability on category j represents the probability that the pre-trained model predicts that the current job applicant belongs to category j; the expression of its conversion process is:

[0142] ;

[0143] in, is an exponential function used to calculate the unnormalized score The index value of Indicates that given a job search feature vector The probability of category j under the condition of , which is equivalent to .

[0144] Embodiment 3: The technical solution of this embodiment is different from that of Embodiment 1 and Embodiment 2 in that:

[0145] Step S3: input the job-seeking feature vector of the current job seeker into the person-job matching model, and the person-job matching model outputs a job with a high matching degree for the current job seeker.

[0146] Step S3 specifically includes:

[0147] S3-1, obtaining the job-seeking feature vector of the current job seeker;

[0148] S3-2, input the job-seeking feature vector of the current job seeker into the person-job matching model, and output the highly matching jobs for the current job seeker.

[0149] The person-job matching model has optimal parameters after fine-tuning , which can recommend highly matching positions to recruiters or job seekers based on the job-seeking feature vector of the current job seeker. Its expression can be defined as:

[0150] Its expression is:

[0151] ;

[0152] in, Recommended job tags for current job seekers. is the job-seeking feature vector of the current job seeker.

[0153] In this embodiment, a person-job matching system based on intelligent data analysis selects a second historical entrant from a first historical entrant, and uses the characteristics of the second historical entrant that have been tested in practice and have a high degree of match as a training set for a person-job matching model, so that the model can learn excellent job-seeking characteristics in the high-quality training set, thereby providing recruiters or job seekers with a higher quality of person-job matching.

[0154] In addition, during the model training process, the model is trained on the second historical entrant data by using self-supervised learning methods, and the model parameters are automatically optimized and adjusted to adapt to the ever-changing job market and job requirements, thereby improving the recommendation effect.

[0155] In addition, the method recommends highly matched positions based on the job-seeking characteristics of job seekers, thereby improving the satisfaction and matching degree of job seekers and recruiters or job seekers.

[0156] In addition, this method reduces manual intervention and simplifies the recruitment process through an automated recommendation process, thereby improving the overall efficiency and quality of recruitment.

[0157] An embodiment of the present invention provides a person-job matching system based on intelligent data analysis, which is used to implement the above-mentioned embodiment, and the descriptions that have been made will not be repeated. The terms "module", "unit", "subunit", etc. used below can be a combination of software and / or hardware that implements predetermined functions. Although the system described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0158] like Figure 2 As shown, Figure 2 This is a structural block diagram of a person-job matching system based on intelligent data analysis of the present invention, the system comprising:

[0159] The matching training set acquisition module is used to obtain the job matching training set D. The training samples of the job matching training set D are N groups of matching data pairs. The matching data pairs are composed of job-seeking feature vectors. and its job matching tags composition;

[0160] ;

[0161] in, , represents the job-seeking feature vector of the i-th second historical employee, represents the set of real numbers with d characteristic dimensions; represents the job matching label of the second historical employee of the i-th position, and the job matching label represents different job matching categories;

[0162] The matching degree training set acquisition module has a built-in screening submodule, and the screening submodule is used to obtain the second historical entrants from the first historical entrants;

[0163] The self-supervised learning module is used to input the job matching training set into the pre-trained model. The pre-trained model is trained through self-supervised learning to obtain the person-job matching model;

[0164] The recommendation module is used to input the job-seeking feature vector of the current job seeker into the person-job matching model, and the person-job matching model outputs the highly matching jobs for the current job seeker.

[0165] In the above system, a high-matching job matching training set is obtained through the matching training set acquisition module; a person-job matching model is trained through the self-supervised learning module; and a high-matching job is output through the recommendation module, thus solving the problem that the existing person-job matching system is difficult to provide higher quality person-job matching.

[0166] The system embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0167] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0168] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A person-job matching system based on intelligent data analysis, characterized in that: include: The matching training set acquisition module is used to acquire the job matching training set D. The training samples of the job matching training set D are N groups of matching data pairs. The matching data pairs are composed of job-seeking feature vectors. and its job matching tags composition; ； in, , represents the job-seeking feature vector of the i-th second historical employee, represents the set of real numbers with d characteristic dimensions; represents a job matching label of the i-th second historical employee, wherein the job matching label represents different job matching categories; The matching degree training set acquisition module has a built-in screening submodule, and the screening submodule is used to obtain the second historical employee from the first historical employee; The screening submodule obtains the second historical job entrants and their predicted job matching degree from the sample training set of the first historical job entrants; The obtaining of the second historical employee includes: Predefine the job contribution of each job-seeking feature in the job sought by the first historical entrant; Preset the weight coefficient of each job-seeking feature according to the job contribution of the first historical job seeker in the job; Calculating the predicted job matching degree of the first historical job seeker according to the weight coefficient of each of the job-seeking characteristics; Obtain the actual job matching degree of the first historical employees after joining the company; Comparing the actual job matching degree with the predicted job matching degree to obtain a matching deviation of the first historical employee; The first historical entrant corresponding to the matching deviation falling within the first threshold range is defined as the second historical entrant; A self-supervised learning module, used to input the job matching training set into a pre-trained model, and the pre-trained model is trained by self-supervised learning to obtain a person-job matching model; The training steps of the self-supervised learning module include: Using the job-seeking feature vector of the second historical employee as a feature variable and the job matching label as a target variable to optimize the model parameters of the pre-trained model; Select multi-label classification cross entropy as the loss function for the training process; Calculate the total cross entropy loss of the job matching labels according to the loss function; Determine whether the total cross entropy loss is lower than the preset total loss threshold. If so, output the current model parameters of the pre-trained model. Fine-tune the current model parameters to obtain the person-job matching model with optimal parameters; The loss function The expression is: ； Among them, K is the total number of job matching label categories, i represents the job matching training set The i-th training sample contains the i-th job-seeking feature vector Match the label with the i-th position , categories are defined as job matching labels Multiple different job types included; It is a pre-trained model For the job-seeking feature vector in the i-th training sample The predicted value of are the parameters of the pre-trained model, is the one-hot encoding of training sample i on the job matching label; It is a pre-trained model The predicted probability of the i-th training sample in the j-th category; Represents the cross entropy loss of the i-th training sample, which calculates the job matching label and predicted values The difference between The fine-tuning expression for fine-tuning the current model parameters is: ； in, represents the optimal parameters after fine-tuning, Indicates that all possible parameters are found Minimize the average loss of the fine-tuning process, is the number of training samples in the validation set; The recommendation module is used to input the job-seeking feature vector of the current job seeker into the person-job matching model, and the person-job matching model outputs a job with a high matching degree for the current job seeker.

2. According to claim 1, a person-job matching system based on intelligent data analysis is characterized in that: The acquisition step of the matching degree training set acquisition module includes: Obtaining a sample training set of the first historical employee, wherein the training samples of the sample training set are the first historical employee and their job-seeking characteristics, wherein the job-seeking characteristics include: skill keyword frequency, skill matching, skill experience years, professional relevance, and commuting distance of the interview text; Obtaining second historical job entrants and their predicted job matching degrees from the sample training set of the first historical job entrants; Determine the job matching label of the second historical employee according to the predicted job matching degree of the second historical employee; sorting the job-seeking features of the second historical job-holder according to a predefined order to construct the job-seeking feature vector; Combining the job-seeking feature vector of the second historical employee and the corresponding job matching label into a matching degree data pair of the second historical employee; The matching data pairs are collected to obtain the job matching training set of the second historical employee.

3. The person-job matching system based on intelligent data analysis according to claim 2 is characterized in that: Obtain the job-seeking characteristics of the first historical employee, including: skill keyword frequency, skill matching, skill experience years, professional relevance, and commuting distance in the interview text, including: Preset skill keywords, wherein the skill keywords include single keywords or combined keywords; According to the skill keywords, skill keyword frequencies and skill level values ​​are extracted from the interview text, and a skill feature vector is constructed; Calculating the skill matching degree according to the skill keyword frequency and the skill level value; The years of skills and experience, professional relevance, and commuting distance were extracted from the interview texts of the first history entrants.

4. The person-job matching system based on intelligent data analysis according to claim 2 is characterized in that: According to the predicted job matching degree of the second historical employee, the job matching label of the second historical employee is determined, including: In the predicted job matching degree of the second historical employee, K second thresholds of different levels are preset; Define job matching tags corresponding to second thresholds of different levels, and create job matching mapping rules according to the second thresholds and their job matching tags; According to the job matching mapping rule, a job matching label is determined for the second historical employee.

5. The person-job matching system based on intelligent data analysis according to claim 1 is characterized in that: The recommendation step of the recommendation module includes: Get the job-seeking feature vector of the current job seeker; Input the job-seeking feature vector of the current job seeker into the person-job matching model, and output the job with high matching degree for the current job seeker; The person-job matching model has the optimal parameters after fine-tuning , which can recommend high-matching positions to recruiters or job seekers based on the job-seeking feature vector of the current job seeker. The expression is: ； in, Recommended job tags for current job seekers. is the job-seeking feature vector of the current job seeker.

Images