WO2021179715A1 - Hidden markov model-based resignation prediction method and related device - Google Patents

Abstract

A hidden Markov model-based resignation prediction method and a related device. Said method comprises: constructing a pre-trained hidden Markov model on the basis of employee static information and influences of a working status on employee resignation intention, and training the pre-trained hidden Markov model by means of sample data of hired employees, so as to acquire a hidden Markov model for predicting a working status and a resignation probability of a job applicant (S101); acquiring application data of the job applicant, and constructing a static information feature data set of the job applicant on the basis of the application data (S102); and inputting the static information feature data set into the hidden Markov model, so as to predict the working status and the resignation probability of the job applicant after being hired (S103). In this way, the present invention provides reference information for a human resource management department to make the decision of whether to hire, avoiding recruiting, to a company, job applicants who are unstable and liable to resign, reducing the costs of human recruitment. In addition, the method has strong expandability on the time scale, for example, resignation warning, etc. can be achieved subsequently.

Description

Resignation prediction method and related device based on hidden Markov model

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on October 21, 2020, the application number is 202011134554.2, and the application name is "Hidden Markov Model-based Resignation Prediction Method and Related Devices". The entire content of the application is approved The reference is incorporated in this application.

Technical field

This application relates to the field of big data technology, and in particular to a method and related devices for predicting turnover based on a hidden Markov model.

Background technique

Human resource management is an important part of company operations, and the prediction of employee turnover is an important problem in the field of human resources. Employee resignation has a great negative impact on the company’s economic costs, efficiency costs, and cultural loss costs. Therefore, the existing technology provides some methods for predicting agent resignation. Most of these methods are logistic regression or random forest for forecasting. Probability and reason for resignation of incumbent employees. As for potential employees, such as job seekers, how to reasonably predict their working status and resignation probability after entry, and provide reference information for the human resources management department to make hiring decisions, there are still obvious reports. The inventor is aware of the evaluation indicators used in the existing methods of predicting agent turnover, such as the effectiveness of interview selection, the pertinence in the training process, the quantitative measurement of the usual high-intensity workload, and the prediction of long-term working conditions, etc. Most are based on human judgment, which inevitably brings subjective arbitrariness and random uncertainty to human resource management.

Summary of the invention

Based on this, it is necessary to provide a resignation prediction method and related devices based on a hidden Markov model to realize a reasonable prediction of the job status and resignation probability of job applicants after entry, and provide reference information for the human resources management department to make hiring decisions. .

A method for predicting turnover based on a hidden Markov model, the method comprising:

Based on the influence of the employee’s static information and work status on the employee’s intention to leave, a pre-trained hidden Markov model is constructed, and the pre-trained hidden Markov model is trained based on the sample data of hired employees to obtain candidates for predicting job applicants Hidden Markov model of job status and turnover probability;

Obtain job applicant's application data, and construct a static information feature data set of the job applicant based on the application data;

The static information feature data set is input into the hidden Markov model to predict the job status and the resignation probability of the job applicant after entering the job.

A resignation prediction device based on a hidden Markov model, the device comprising:

The first acquisition module is used to construct a pre-training hidden Markov model based on the employee’s static information and the influence of the work status on the employee’s intention to leave, and to use the sample data of the hired employee to hide the pre-training The Markov model is trained to obtain the hidden Markov model used to predict the job status and the probability of leaving the job;

A second acquisition module, the second acquisition module is used to acquire job applicants’ application data, and build a static information feature data set of the job applicants based on the application data;

The prediction module is configured to input the static information feature data set into the hidden Markov model to predict the job status and the probability of leaving the job after the job applicant enters the job.

In addition, in order to achieve the above object, the present application also provides a computer device, including a memory and a processor, the memory storing program instructions for implementing the following steps, the steps including:

Based on the influence of the employee’s static information and work status on the employee’s intention to leave, a pre-trained hidden Markov model is constructed, and the pre-trained hidden Markov model is trained based on the sample data of hired employees to obtain candidates for predicting job applicants Hidden Markov model of work status and turnover probability; acquiring job applicant’s application data, and constructing a static information feature data set of the job applicant based on the application data; inputting the static information feature data set into the hidden Markov The husband model predicts the job status and the probability of leaving the job after the job applicant enters the job.

The processor is configured to execute program instructions stored in the memory. In addition, in order to achieve the above object, the present application also provides a storage medium storing computer readable instructions, which when executed by one or more processors, cause the one or more processors to perform the following steps:

Based on the influence of the employee’s static information and work status on the employee’s intention to leave, a pre-trained hidden Markov model is constructed, and the pre-trained hidden Markov model is trained based on the sample data of hired employees to obtain candidates for predicting job applicants Hidden Markov model of work status and turnover probability; acquiring job applicant’s application data, and constructing a static information feature data set of the job applicant based on the application data; inputting the static information feature data set into the hidden Markov The husband model predicts the job status and the probability of leaving the job after the job applicant enters the job.

Compared with the prior art, the Hidden Markov Model-based Resignation Prediction Method and related devices of the present application construct a pre-trained Hidden Markov Model based on the influence of the employee’s static information and work status on the employee’s resignation intention, and The pre-trained hidden Markov model is trained through the sample data of hired employees to obtain a hidden Markov model for predicting the job status and the probability of leaving the job; the application data of the job applicant is obtained, and the construction is based on the application data The static information feature data set of the job applicant; the static information feature data set is input into the hidden Markov model to predict the job status and the resignation probability of the job applicant after entering the job. By predicting the job status and resignation probability of job applicants, after the application process is over, the human resources management department can make a decision on whether to hire or not based on the predicted results, so as to avoid hiring unstable job applicants who are easy to resign into the company. , Thereby reducing the loss caused by employee resignation, while also reducing the cost of manpower recruitment. In addition, the hidden Markov model used in this application is a time series model, and its prediction results also include the time of resignation. Therefore, the resignation prediction method provided by this application has strong scalability on the time scale, and the subsequent sequence can be further realized Respond to resignation warnings in real time.

Description of the drawings

FIG. 1 is a schematic flowchart of a method for predicting turnover based on a hidden Markov model according to an embodiment of the present application;

Fig. 2 is a schematic diagram of a structure of the pre-trained hidden Markov model in the method shown in Fig. 1;

FIG. 3 is a schematic flow chart of a method for resignation prediction based on a hidden Markov model in another embodiment of the present application;

FIG. 4 is a schematic flowchart of a method for predicting turnover based on a hidden Markov model in another embodiment of the present application;

FIG. 5 is a schematic flowchart of a method for predicting turnover based on a hidden Markov model according to another embodiment of the present application;

Fig. 6 is a schematic flow chart of a method for resignation prediction based on a hidden Markov model in another embodiment of the present application;

FIG. 7 is a schematic structural diagram of a resignation prediction device based on a hidden Markov model in an embodiment of the present application;

FIG. 8 is a schematic structural diagram of a computer device according to an embodiment of the present application;

FIG. 9 is a schematic structural diagram of a storage medium according to an embodiment of the present application.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of this application.

In order to reasonably predict the job status and resignation probability of job applicants after entering the job, and to provide reference information for the human resources management department to make a decision on whether to hire or not, this application provides a method and related devices for resignation prediction based on a hidden Markov model.

Please refer to FIG. 1, which is a schematic flowchart of a method for predicting turnover based on a hidden Markov model according to an embodiment of the present application. It should be noted that if there is substantially the same result, the method of the present application is not limited to the sequence of the process shown in FIG. 1. As shown in Figure 1, the method includes:

Step S101: Based on the influence of the employee’s static information and work status on the employee’s intention to leave, construct a pre-trained hidden Markov model, and train the pre-trained hidden Markov model based on the sample data of hired employees to obtain predictions Hidden Markov Model of Job Seeker's Job Status and Resignation Probability.

Optionally, in step 101, the static information of the employee includes some basic personal information of the employee, such as name, age, home address, telephone number, graduate school, education, major, project experience, internship experience, personal evaluation, and professional skills Etc., these basic information are generally not changed in general, and can be extracted and integrated from the application data generated when employees apply for a job. Further, the application data may include any one or a combination of resume data, interview video data, and written examination data.

The work status of employees includes employee performance, employee attendance, employee leave frequency, leadership scoring records, colleague evaluation records, and employee rewards and punishments. The work status data generated after the employee enters the job can be stored in the employee management system, and can be extracted from it when constructing the data set required for training or validating the model.

The employee’s resignation intention, that is, the employee’s likelihood of resignation, can be marked with a numerical value. For example, the value 1 means that the employee resigns, and the value 0 means that the employee does not resign. The closer the value is to 1, the stronger the employee’s resignation intention, and the closer the value is to 0, the weaker the employee’s resignation intention.

It should be noted that when storing data or constructing a data set required for training or validating a model, the employee’s work status and resignation intentions are arranged in a one-to-one correspondence according to time. For incumbent employees, set their resignation intention label value to 0 after entry, that is, for incumbent employees, the resignation intention label value corresponding to their work status at any time can be assumed to be 0. For resigning employees, the resignation intention label value before resignation is set to 0, and the resignation intention label value at the time of resignation is set to 1, that is, the resignation intention label value corresponding to the work status data generated at the time of resignation is required Assume it is 1.

Under normal circumstances, the static information of the employee will have a certain impact on the employee's work status and resignation intention after entry. For example, if an employee has high job expectations for the position he is applying for when applying for a job, these characteristics may be manifested during the interview, and his working status after entering the job may also be better, resulting in a weaker intention to leave. Or, when an employee is a fresh graduate with no work experience, he can predict whether his professional knowledge is solid and his learning ability is strong through the school he attends, the type of activities he participates, etc. These characteristics will generally be directly recorded in the resume data middle. If their professional knowledge is solid and they have strong learning ability, they will be more adaptable to the first job after entering the job, and their working status will be better, and their intention to leave will be weak. On the contrary, their ability to adapt to the first job may be weak after entering the job, resulting in a poor working status and a stronger intention to leave. Or, the size of the city where employees are located will affect their acceptance of commuting distance. For example, in first-tier cities, one-hour commute per day is acceptable, while in third-tier and fourth-tier cities, one-hour commute per day is not easy to accept. That is, when the employee's city and commuting time are within the acceptable range, their intention to leave will be weaker; otherwise, their intention to leave will be stronger.

Therefore, in this application, the employee’s static information is used as input, the employee’s working status is the implicit state, and the employee’s resignation intention is the observation state, and a pre-trained hidden Markov model is constructed λ=((A,B,C ),X,π). Among them, X represents the static information of the employee, which is the input item of the model; A, B, and C are all model parameters, and π is the initial state probability matrix.

For ease of understanding, please refer to FIG. 2. FIG. 2 is a schematic structural diagram of the pre-training hidden Markov model constructed in step S101. As shown in Figure 2, the employee’s work status Y is the implicit status, and the employee’s resignation intention Z is the observation status; A is the first observation probability matrix, and the elements in the first observation probability matrix

Static information indicates when X _i, Y _j operating state of probability; B is the state transition probability matrix, the state transition probability matrix element

Represents the probability of the transition from the working state Y _j to the working state Y _k ; C is the second observation probability matrix, and the elements in the second observation probability matrix

Static information is represented by X _i, Z _j turnover intention of probability, i, j and k are a natural number greater than or equal to 1.

It should be noted that in the pre-trained hidden Markov model λ=((A,B,C),X,π), the model parameters A, B, and C are all unknown and need to be obtained through training. The hidden state sequence {Y ₁ , Y ₂ ,..., Y _n } can be determined by analyzing the work status data of a sample of hired employees, that is, each hidden state Y _j in the hidden state sequence represents the employee’s position. This kind of working status is given. Among them, the sample of hired employees includes samples of current employees and samples of resigned employees. The samples of resigned employees include samples of resigned employees and samples of employees who have submitted resignation applications but have not formally resigned.

Optionally, the hidden state sequence is determined by a statistical method. For example, first count the job status categories of each hired employee sample. Considering that different hired employee samples may have the same or similar job status categories, you can use any appropriate method in the field to de-duplicate the same job status categories. Or directly cluster the job status categories of each sample of hired employees, and finally establish an implicit status sequence based on all the job status categories after deduplication or the clustered job status categories.

By establishing the above-mentioned pre-trained hidden Markov model, and then using the sample data of hired employees to train it, a hidden Markov model can be obtained that can be used to predict the job status of job seekers and the probability of leaving the job. Among them, the hired employee sample data includes application data generated when each employee sample applies for employment, job status data generated after entry, and resignation intention data that corresponds to the job status data in a chronological order.

Optionally, the method for training the pre-trained hidden Markov model using the sample data of hired employees includes: first establishing training set data, the training set data includes static information feature data of a plurality of first hired employee samples Set, work status feature data set sequence, and resignation intention tag value sequence, where the resignation intention tag value in the resignation intention tag value sequence corresponds to the work status feature data set in the work status feature data set sequence in a one-to-one order. A sample of hired employees includes a sample of the first incumbent employee and a sample of the first resigned employee. Then use the training set data to train the pre-trained hidden Markov model.

It should be noted that, in this embodiment, the hired employee sample data is divided into two parts: the first hired employee sample data and the second hired employee sample data. The first hired employee sample data is used to establish training set data, and the training set data is used to train the pre-trained hidden Markov model constructed as described above to obtain model parameters. The second hired employee sample data is used to establish test set data, and the test set data is used to test whether the model parameters obtained after training are better or optimal model parameters.

Specifically, the static information feature data set includes any one or a combination of resume information features, interview video features, and written examination information features.

In order to avoid the dependence of the prediction model on the data form and the over-fitting caused by the excessively high data dimension, preferably, in this embodiment, the deep learning graph neural network model is used for resume text data, interview video data and written test text data. Carry out dimensionality reduction classification to realize the efficient conversion, analysis and integration of unstructured data to improve the accuracy of model prediction.

Specifically, the method for obtaining resume information features includes: inputting resume text into a deep learning resume text graph neural network model, and the deep learning resume text graph neural network model outputs resume information features corresponding to the resume text. Optionally, the resume information features include educational background, age, project experience, internship experience, personal evaluation, professional skills, etc.

The method for acquiring features of the interview video includes: inputting at least one frame of images in the interview video into a deep learning facet neural network model, and the deep learning facet neural network model outputs interview video features corresponding to the image. Optionally, the interview video features include facial expressions, dress, etiquette, interview duration, etc.

The method for acquiring written test information features includes: inputting the written test text into a deep learning written test text graph neural network model, and the deep learning written test text graph neural network model outputs written test information features corresponding to the written test text. Optionally, the written test information features include personality test scores, professional skill scores, and the like.

Specifically, the work status feature data set includes any one or a combination of employee performance, employee attendance, employee leave frequency, leadership scoring, colleague evaluation, employee rewards and punishments.

Specifically, the resignation intention label value sequence is a sequence composed of 0 and 1, such as

Indicates that the employee has not resigned at time t and the work status is Y _j , and then

It means that the employee _resigns when the work status is Y j at time t+1.

Specifically, during training, the employee’s static information feature data set and the employee’s work status data set sequence are used as input, and the employee’s turnover intention value is output by adjusting the model parameters A, B, and C, and then the output intention value of the employee Compare with the corresponding actual resignation intention label value. If the output resignation intention is consistent with the actual resignation intention, determine the model parameters A, B, and C at this time; otherwise, continue to adjust the model parameters A, B, and C.

It should be noted that the employee turnover intention value output by the model ranges from 0 to 1. When the value of the employee's resignation intention output by the model is between 0-0.5, it indicates that the employee's resignation intention is low, and it is judged not to resign. If the corresponding actual resignation intention label value is 0 at this time, it can be judged that the employee resignation intention value output by the model is consistent with the actual resignation intention label value, otherwise the judgment is inconsistent. When the value of the employee's resignation intention output by the model is between 0.5 and 1, it indicates that the employee's resignation intention is high, and the employee is judged to resign. If the corresponding actual resignation intention label value is 1, it can be judged that the employee resignation intention value output by the model is consistent with the actual resignation intention label value, otherwise, the judgment is inconsistent.

In other embodiments of the present application, after the training is completed, the model parameters A, B, and C obtained by training need to be tested and checked. If the test fails, the model parameters A, B, and C obtained after training need to be tested and checked. optimize. Optionally, the test inspection method includes: inputting the static information feature data set in the test set data into the pre-trained hidden Markov model after training for testing, and outputting the work status feature data of each second hired employee sample Set prediction sequence and resignation intention label value prediction sequence. Then the work status feature data set prediction sequence and the resignation intention label value prediction sequence are compared with the corresponding work status feature data set sequence and the resignation intention label value sequence in the test set data respectively, and the prediction accuracy rate is calculated. When the prediction accuracy is greater than or equal to the preset accuracy threshold, the training ends and the model parameters are determined; when the prediction accuracy is less than the preset accuracy threshold, return to re-establish the training set data to repeat the training process to obtain the results after training The model parameters are optimized until a Hidden Markov Model with high prediction accuracy can be used to predict the job status and turnover probability of job applicants.

It should be noted that the method for establishing the test set data is the same as the method for establishing the training set data. For the sake of brevity, it is not repeated here. Specifically, the test set data includes a static information feature data set, a work status feature data set sequence, and a resignation intention label value sequence of a plurality of second hired employee samples, where the resignation intention label value in the resignation intention label value sequence There is a one-to-one correspondence with the work state characteristic data set in the work state characteristic data set sequence according to the time sequence; the second hired employee sample includes a second in-service employee sample and a second resigned employee sample.

It should be noted that in step S101, a Hidden Markov Model for predicting the job status and turnover probability of job applicants can be constructed and trained in advance, and it can be called directly when needed. It can also be constructed and trained to obtain the hidden Markov model operation for predicting the job status and the probability of leaving the job, and then call it.

Step S102: Obtain the job applicant's application data, and construct the job applicant's static information feature data set based on the job applicant's application data.

In step S102, the method of constructing the static information feature data set of job applicants is the same as the method of constructing the static information feature data set of the first hired employee sample in step S101. For the sake of simplicity, it will not be repeated here.

Step S103: Input the static information feature data set into the hidden Markov model for predicting the job seeker's work status and resignation probability, and predict the job seeker's work status and resignation probability after entering the job.

Optionally, in step S103, after the static information feature data set is input into the Hidden Markov Model, the probability of the hidden state sequence is solved by the forward-backward algorithm, and the job status of the job applicant and the probability of leaving the job are predicted.

The specific solution process is as follows:

Is implicitly assumed state sequence _{o 1, o 2, o 3} , ..., o t, ..., o n, the state observation sequence _{s 1, s 2, s 3} , ..., s t,. ..,s _n .

Forward probability:

Given a hidden Markov model λ=((A,B,C),X,π), define the partial observation state sequence o ₁ , o ₂ , o ₃ ,..., o _{t at time t} , and the working state _{The probability of st} is the forward probability, denoted as α _t (i) = P(o ₁ ,o ₂ ,...,o _t ,s _t |λ), according to the recurrence formula

, Solve the forward probability.

Similarly, given hidden Markov model λ = ((A, B, C), X, π), defined at time t and the working state is s _t , the partial observation sequence from t + 1 to T Is the partial observation state sequence o _t+1 ,o _t+2 ,o _t+3 ,...,o _{The probability of T} is the backward probability, denoted as β _t (i) = p(o _t+1 ,o _{t +2} ,o _t+3 ,...,o _T |s _t ,λ), according to the recurrence formula

Solve the backward probability, that is, get the predicted work status of the employee. Due to the relationship P(s _t |o)∝α _t (s _t )β _t (s _t ), the probability of employee leaving can be predicted.

In this application, when forecasting, the data used is the static information feature data set of job applicants, and no one is the evaluation index. Therefore, the subjective arbitrariness and random uncertainty brought about by human resource management can be avoided.

The hidden Markov model-based resignation prediction method of the embodiment shown in Figure 1 constructs a pre-trained hidden Markov model based on the employee’s static information and the influence of work status on the employee’s resignation intention, and uses the sample data of hired employees Train the pre-trained hidden Markov model to obtain a hidden Markov model for predicting the job status and turnover probability of the job applicant; obtain the job applicant’s application data, and construct the job applicant’s static information based on the job applicant’s application data Feature data set: Input the static information feature data set into the hidden Markov model to predict the job status and the probability of leaving the job after the job applicant. By predicting the job status and resignation probability of job applicants, after the application process is over, the human resources management department can make a decision on whether to hire or not based on the predicted results, so as to avoid hiring unstable job applicants who are easy to resign into the company. , Thereby reducing the loss caused by employee resignation, while also reducing the cost of manpower recruitment. In addition, the hidden Markov model used in this embodiment is a time series model, and its prediction result also includes the time of resignation. Therefore, the resignation prediction method provided by this embodiment is highly scalable on the time scale. Further realize real-time response to resignation warning, etc.

In the foregoing embodiment, the hidden status sequence is determined based on the analysis of the work status data of all selected samples of hired employees. Considering that the working status of job seekers with different working years is quite different after entering the job, in another embodiment of this application, the working years are classified, for example, including no working years (ie fresh graduates) category, 1-2 The category of annual working years, the category of working years of 2-5 years, the category of working years of 5-10 years, etc. Then, according to the categories, a hidden Markov model for predicting the future work status and the probability of leaving the job is constructed and trained. In each category, the specific training method is the same as the training method described in step S101, for the sake of brevity, it will not be repeated here.

Please refer to FIG. 3, which is a schematic flowchart of a method for predicting turnover based on a hidden Markov model according to another embodiment of the present application. It should be noted that if there are substantially the same results, the method of the present application is not limited to the sequence of the process shown in FIG. 3. As shown in Figure 3, the method includes:

Step S201: Obtain the application data of the job applicant, construct the static information feature data set of the job applicant based on the job applicant's application data, and determine the working years category of the job applicant.

In this step S201, the application data of the job applicant is obtained, and the step of constructing the static information feature data set of the job applicant based on the application data of the job applicant is similar to the step S102 of the embodiment shown in FIG. Go into details again.

In step S201, any suitable classification model in the art can be used to determine the working years category to which the job applicant belongs. It is also possible to manually classify the categories after identifying the working years.

Step S202: Invoking a Hidden Markov Model corresponding to the working years category for predicting the job status and the probability of leaving the job according to the working years category to which the job applicant belongs.

In this embodiment, prediction models corresponding to different working years categories are constructed and trained in advance, and then in step S202, the corresponding prediction models can be called according to the working years category of the job seeker.

Step S203: Input the static information feature data set into the hidden Markov model for predicting the job-seeker's work status and resignation probability, and predict the job-seeker's work status and resignation probability after entering the job.

In this embodiment, step S203 is similar to step S103 in the embodiment shown in FIG.

The Hidden Markov Model-based Resignation Prediction Method of the embodiment shown in FIG. 3 obtains the job applicant’s application data, constructs the job applicant’s static information feature data set based on the job applicant’s application data and determines the job applicant’s belongings Working life category; according to the working life category of the job applicant, call the hidden Markov model corresponding to the working life category for predicting the job status and turnover probability of the job applicant; input the static information feature data set to the job application prediction The Hidden Markov Model of the job status and the probability of leaving the job predicts the job status and the probability of leaving the job applicant. Compared with the embodiment shown in FIG. 1, through the above method, the accuracy of model prediction can be improved to help the human resource management department make more correct decisions, and further reduce the loss caused by employee resignation and reduce the cost of human recruitment.

Further, considering that employees and job applicants in different regions have different expectations and acceptable scope of work, in another embodiment of this application, the regions where the employees are located are classified, for example, into the south and the north, or According to the city, it is divided into first-tier cities, second-tier cities, third-tier cities, etc., and then constructs and trains hidden Markov models for predicting the future job status and the probability of leaving the job according to the regional category. In each category, the specific training method is the same as the training method described in step S101, for the sake of simplicity, it will not be repeated here.

Please refer to FIG. 4, which is a schematic flow chart of a method for resignation prediction based on a hidden Markov model in another embodiment of the present application. It should be noted that if there is substantially the same result, the method of the present application is not limited to the sequence of the process shown in FIG. 4. As shown in Figure 4, the method includes:

Step S301: Obtain application data of the job applicant, construct a static information feature data set of the job applicant based on the job applicant's application data, and determine the area category to which the job applicant belongs.

In this step S301, the application data of the job applicant is obtained, and the step of constructing the static information feature data set of the job applicant based on the application data of the job applicant is similar to the step S102 of the embodiment shown in FIG. Go into details again.

Optionally, in step S301, the administrative region to which the job seeker belongs is determined according to the resume text information of the job seeker, and then the category of the region to which the job seeker belongs is determined according to the administrative region to which the job seeker belongs.

Step S302: According to the category of the area to which the job seeker belongs, a hidden Markov model corresponding to the category of the job seeker is called for predicting the job status of the job seeker and the probability of leaving the job.

In this embodiment, prediction models corresponding to different area categories are constructed and trained in advance, and then in step S302, the corresponding prediction models can be called according to the area category to which the job applicant belongs.

Step S303: Input the static information feature data set into the hidden Markov model for predicting the job-seeker's work status and resignation probability, and predict the job-seeker’s work status and resignation probability after entering the job.

In this embodiment, step S303 is similar to step S103 of the embodiment shown in FIG.

The Hidden Markov Model-based Resignation Prediction Method of the embodiment shown in FIG. 4 obtains the job applicant’s application data, constructs the job applicant’s static information feature data set based on the job applicant’s application data and determines the region where the job applicant belongs Category; call the hidden Markov model corresponding to the area category for predicting the job applicant’s working status and the probability of resignation according to the area category to which the job applicant belongs; input the static information feature data set into the job applicant’s working status and The Hidden Markov Model of Resignation Probability, which predicts the job status and resignation probability of the job applicant after entering the job. Compared with the embodiment shown in FIG. 1, through the above method, the accuracy of model prediction can be improved to help the human resource management department make more correct decisions, and further reduce the loss caused by employee resignation and reduce the cost of human recruitment.

Please refer to FIG. 5, which is a schematic flowchart of a method for predicting turnover based on a hidden Markov model according to another embodiment of the present application. It should be noted that if there is substantially the same result, the method of the present application is not limited to the sequence of the process shown in FIG. 5. As shown in Figure 5, the method includes:

Step S401: Obtain application data of the job applicant, construct a static information feature data set of the job applicant based on the job applicant's application data, and determine the area category and the working years category of the job applicant.

In this step S401, the application data of the job applicant is obtained, and the step of constructing the static information characteristic data set of the job applicant based on the application data of the job applicant is similar to the step S102 of the embodiment shown in FIG. The steps are similar to step S201 of the embodiment shown in FIG. 3, and the step of determining the working years category of the job applicant is similar to step S301 of the embodiment shown in FIG.

Step S402: According to the region category and the working years category of the job seeker, the hidden Markov model corresponding to the region category and the working years category is called for predicting the job seeker's working status and the probability of leaving the job.

In this embodiment, according to the regional classification, in each regional category, the prediction models corresponding to different working years categories are pre-built and trained, and then in step S402, the job applicant can be based on the regional category and the working years category. Call the corresponding prediction model.

Step S403: Input the static information feature data set into the hidden Markov model for predicting the job-seeker's work status and resignation probability, and predict the job-seeker’s work status and resignation probability after entering the job.

In this embodiment, step S403 is similar to step S103 of the embodiment shown in FIG.

The Hidden Markov Model-based Resignation Prediction Method of the embodiment shown in FIG. 5 obtains the job applicant’s application data, constructs the job applicant’s static information feature data set based on the job applicant’s application data and determines the region where the job applicant belongs Category and working years category; according to the region category and working years category of the job seeker, call the hidden Markov model corresponding to the region category to predict the job seeker’s working status and the probability of leaving the job; input the static information feature data set The Hidden Markov Model for predicting the job status and the probability of leaving the job seeker predicts the job status and the probability of leaving the job after the job seeker. Compared with the embodiment shown in FIG. 1, through the above method, the accuracy of model prediction can be improved to help the human resource management department make more correct decisions, and further reduce the loss caused by employee resignation and reduce the cost of human recruitment.

Please refer to FIG. 6, which is a schematic flowchart of a method for predicting turnover based on a hidden Markov model according to another embodiment of the present application. It should be noted that if there is substantially the same result, the method of the present application is not limited to the sequence of the process shown in FIG. 6. As shown in Figure 6, the method includes:

Step S501: Based on the influence of the employee’s static information and work status on the employee’s intention to leave, construct a pre-trained hidden Markov model, and train the pre-trained hidden Markov model based on the sample data of hired employees to obtain predictions Hidden Markov Model of Job Seeker's Job Status and Resignation Probability.

Optionally, in this embodiment, step S501 is similar to step S101 of the embodiment shown in FIG.

Step S502: Obtain application data of the job applicant, construct a static information characteristic data set of the job applicant based on the application data of the job applicant, and create a job application portrait based on the static information characteristic data set of the job applicant.

Optionally, in this embodiment, the step of acquiring the job applicant’s application data in step S502, and constructing the job applicant’s static information feature data set based on the job applicant’s application data is similar to step S102 in the embodiment shown in FIG. 1 , For the sake of simplicity, I won’t repeat it here.

Optionally, when the static information feature data set of the job applicant is obtained, a personal file (ie job application portrait) dedicated to the job applicant can be created for the job applicant, and the personal file can be matched with the job applicant’s photo and application number , Easy to find, all these personal files created can be stored in the database, because the applicant's application number is unique within the enterprise, so the files stored in the database can be searched according to the unique identification of the application number.

Step S503: Input the static information feature data set into the Hidden Markov Model for predicting the job status and resignation probability of the job applicant, and predict the job status and resignation probability of the job applicant after entering the job.

Optionally, in this embodiment, step S503 is similar to step S503 of the embodiment shown in FIG.

Step S504: Determine whether the corresponding entry time when the job applicant's resignation probability is greater than the preset probability threshold is greater than or equal to the preset time threshold.

Optionally, in step S504, the preset probability threshold and the preset time threshold may be set according to actual needs. For example, the preset probability threshold is set to 0.9, and the preset time threshold is set to 2 years. Of course, the preset time threshold can also be set separately according to the job position. For example, the preset time threshold for important or core project positions can be set slightly longer Some, such as 5 years, etc.

The aforementioned Hidden Markov Model has been introduced as a time series model. When it is predicted that the job applicant will leave the job, the corresponding entry time at the time of resignation can be correspondingly known. If the entry time is greater than or equal to the preset time threshold, the job search is determined If the job seeker is a stable job seeker who is not easy to resign, step S505 is executed. Otherwise, it is determined that the job seeker is an unstable job seeker who is easy to resign, and step S506 is executed.

Step S505: Set an admission mark on the job-seeking portrait of the job-seeker.

Step S506: Set a non-admission flag on the job-seeking portrait of the job-seeker.

The hidden Markov model-based resignation prediction method of the embodiment shown in Figure 6 constructs a pre-trained hidden Markov model based on the employee’s static information and the influence of the work status on the employee’s resignation intention, and uses the sample data of hired employees Train the pre-trained hidden Markov model to obtain a hidden Markov model for predicting the job status and turnover probability of the job applicant; obtain the job applicant’s application data, and build the job applicant’s static state based on the job applicant’s application data Information feature data set, and create a job application portrait based on the job applicant’s static information feature data set; input the static information feature data set into the hidden Markov model used to predict the job applicant’s work status and the probability of leaving the job to predict the job applicant The job status and resignation probability after entering the job; judge whether the entry time corresponding to the job applicant’s resignation probability is greater than the preset probability threshold is greater than or equal to the preset time threshold; if so, set an admission mark on the job applicant’s portrait ; Otherwise, set a non-admission mark on the job-seeking portrait of the job-seeker. By creating job search portraits for job applicants, judging whether they can be accepted according to preset conditions, and setting corresponding signs on the job search portraits, it is more convenient for the human resources management department to intuitively obtain the information of candidates that can be admitted, and it is also for human resources management The department will search for subsequent employees to provide reference information, which makes the work of the human resources management department more convenient, can improve the work management efficiency of the human resources management department, and reduce the cost of recruitment.

FIG. 7 is a schematic structural diagram of a resignation prediction device based on a hidden Markov model in an embodiment of the present application. As shown in FIG. 7, the resignation prediction device 60 includes a first acquisition module 61, a second acquisition module 62 and a prediction module 63.

Among them, the first acquisition module 61 is used to construct a pre-training hidden Markov model based on the employee’s static information and work status’s influence on the employee’s intention to leave, and to perform the pre-training hidden Markov model on the sample data of hired employees. Train to obtain a hidden Markov model for predicting the job status and turnover probability of job applicants. The second acquiring module 62 is used to acquire the job applicant's application data, and build the job applicant's static information feature data set based on the job applicant data. The prediction module 63 is coupled with the first acquisition module 61 and the second acquisition module 62, and is used to input the static information feature data set into the hidden Markov model to predict the job status and the probability of leaving the job after the job applicant.

Optionally, the first obtaining module 61 obtains the hidden Markov model for predicting the job status and the probability of leaving the job. The intention to leave is the observation state, and a pre-training hidden Markov model is constructed. The model parameters of the pre-training hidden Markov model include the first observation probability matrix A, the state transition probability matrix B, and the second observation probability matrix C; An element of the observation probability matrix A

Static information indicates when X _i, Y _j operating state probability; the state transition probability matrix element B is

Indicates the probability of the transition from the working state Y _j to the working state Y _k ; the element in the second observation probability matrix C

Static information is represented by X _i, Z _j turnover intention of probability, i, j and k are a natural number greater than or equal to 1; the establishment of training data, the training data set comprises a first plurality of samples has been hired employee Static information feature data set, work status feature data set sequence, and resignation intention label value sequence. The resignation intention label value in the resignation intention label value sequence corresponds to the work status feature data set in the work status feature data set sequence according to the time sequence. ; Among them, the first hired employee sample includes the first in-service employee sample and the first resigned employee sample; the training set data is used to train the pre-training hidden Markov model.

Optionally, the operation of the first obtaining module 61 to obtain the hidden Markov model for predicting the job status and the probability of leaving the job further includes: establishing a test set data, the test set data includes a plurality of second hired employee samples The static information feature data set, the work status feature data set sequence and the resignation intention label value sequence, the resignation intention label value in the resignation intention label value sequence and the work status feature data set in the work status feature data set sequence are sorted by time one by one Correspondence; among them, the second sample of hired employees includes the second sample of incumbent employees and the second sample of resigned employees; the static information feature data set in the test set data is input into the pre-trained hidden Markov model after training Test, output the prediction sequence of the work status feature data set and the expected sequence of the resignation intention label value of each second hired employee sample; compare the prediction sequence of the work status feature data set with the corresponding work status feature data set sequence in the test set data Compare and compare the resignation intention label value prediction sequence with the corresponding resignation intention label value sequence in the test set data to calculate the prediction accuracy; when the prediction accuracy is greater than or equal to the preset accuracy threshold, the training ends and the model is determined Parameters; when the prediction accuracy is less than the preset accuracy threshold, return to the step of establishing the training set data to optimize the model parameters obtained after training.

Optionally, the static information feature data set includes any one or a combination of resume information features, interview video features, and written examination information features.

Optionally, the method for obtaining resume information features includes: inputting resume text into a deep learning resume text graph neural network model, and the deep learning resume text graph neural network model outputs the resume information feature.

Optionally, the method for acquiring features of the interview video includes: inputting at least one frame of images in the interview video into a deep-learning facial-attempt neural network model, and the deep-learning facial-attempt neural network model outputs the interview video features.

Optionally, the method for acquiring written test information features includes: inputting the written test text into a deep learning written test text graph neural network model, and the deep learning written test text graph neural network model outputs the written test information feature.

Optionally, the resignation prediction device 60 further includes a determination model 64 coupled to the second acquisition module 62 for determining the area category to which the job applicant belongs based on the job applicant's application data. The first acquiring module 61 is also coupled to the determining module 64, and is configured to call a Hidden Markov Model corresponding to the region category for predicting the job status and the probability of leaving the job according to the region category to which the job seeker belongs.

Optionally, the determination model 64 is also used to determine the type of working years to which the job applicant belongs based on the job application data of the job applicant. The first acquisition module 61 is also configured to call a Hidden Markov Model corresponding to the working years category for predicting the job seeker's working status and the probability of leaving the job according to the working years category to which the job seeker belongs.

Optionally, the resignation prediction device 60 further includes a creation module 65 coupled to the second acquisition module 62, and is configured to create a job-seeking portrait according to the static information feature data set of the job-seeker.

Optionally, the prediction module 63 is also coupled with the creation module 65, and is used to display the job applicant’s job search portrait when it is predicted that the job applicant’s resignation probability is greater than the preset probability threshold and the corresponding entry time is less than the preset time threshold. Set the non-admission flag; when it is predicted that the job applicant’s resignation probability is greater than the preset probability threshold and the corresponding entry time is greater than or equal to the preset time threshold, set an admission flag on the job applicant’s portrait.

Please refer to FIG. 8, which is a schematic structural diagram of a computer device according to an embodiment of the application. As shown in FIG. 8, the computer device 70 includes a processor 71 and a memory 72 coupled to the processor 71. The memory 72 stores computer-readable instructions, and when the computer-readable instructions are executed by the processor 71, the processor 71 executes the steps of the above-mentioned hidden Markov model-based resignation prediction method.

The processor 71 may also be referred to as a CPU (Central Processing Unit, central processing unit). The processor 71 may be an integrated circuit chip with signal processing capabilities. The processor 71 may also be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component . The general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like.

Refer to FIG. 9, which is a schematic structural diagram of a storage medium according to an embodiment of the application. The storage medium 80 stores computer-readable instructions 81, which when executed by one or more processors, cause the one or more processors to execute the steps of the above-mentioned Hidden Markov Model-based Resignation Prediction Method . Wherein, the computer-readable instruction 81 may be stored in the above-mentioned storage medium in the form of a software product, including several instructions for enabling a computer device (may be a personal computer, a server, or a network device, etc.) or a processor (processor) Perform all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium 80 includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disks or optical disks, etc., which can store program codes. Media, or terminal devices such as computers, servers, mobile phones, tablets, etc. The storage medium may be non-volatile or volatile.

In the several embodiments provided in this application, it should be understood that the disclosed system, device, and method can be implemented in other ways. For example, the device embodiments described above are only illustrative, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation.

In addition, the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or software functional unit. The above are only implementations of this application, and do not limit the scope of this application. Any equivalent structure or equivalent process transformation made using the content of the description and drawings of this application, or directly or indirectly applied to other related technical fields, The same reasoning is included in the scope of patent protection of this application.

Claims (20)

1. A method for predicting turnover based on a hidden Markov model, wherein the method includes:

   Based on the influence of the employee’s static information and work status on the employee’s intention to leave, a pre-trained hidden Markov model is constructed, and the pre-trained hidden Markov model is trained based on the sample data of hired employees to obtain candidates for predicting job applicants Hidden Markov model of job status and turnover probability;

   Obtain job applicant's application data, and construct a static information feature data set of the job applicant based on the application data;

   The static information feature data set is input into the hidden Markov model to predict the job status and the resignation probability of the job applicant after entering the job.
2. The method according to claim 1, wherein the pre-training hidden Markov model is constructed based on the static information of the employee and the influence of the work status on the employee’s intention to leave, and the pre-training hidden Markov model is constructed based on the sample data of hired employees. The Markov model is trained to obtain the hidden Markov model used to predict the job status and turnover probability of job applicants, including:

   Taking the employee’s static information as input, taking the employee’s working status as the hidden state and the employee’s turnover intention as the observation state, a pre-training hidden Markov model is constructed. The model parameters of the pre-training hidden Markov model include the first The observation probability matrix A, the state transition probability matrix B and the second observation probability matrix C; wherein, the elements in the first observation probability matrix A

   

   Static information indicates when X _i, Y _j probability working state; and the state transition probability matrix element B is

   

   Indicates the probability of the transition from the working state Y _j to the working state Y _k ; the element in the second observation probability matrix C

   

   Represents the static information is X _i, intention to quit the probability of Z _j, i, j and k are positive integers;

   The training set data is established, and the training set data includes a static information feature data set, a work status feature data set sequence, and a resignation intention label value sequence of a plurality of first hired employee samples, and resignation in the resignation intention label value sequence The intention label value corresponds to the work state feature data set in the work state feature data set sequence in a time sequence; wherein, the first hired employee sample includes the first in-service employee sample and the first resigned employee sample;

   Using the training set data to train the pre-training hidden Markov model.
3. The method according to claim 2, wherein after said constructing a pre-trained hidden Markov model, the method further comprises:

   The test set data is established, and the test set data includes a static information feature data set, a work status feature data set sequence, and a resignation intention label value sequence of a plurality of second hired employee samples, and resignation in the resignation intention label value sequence The intention label value corresponds to the work status feature data set in the work status feature data set sequence in a chronological order; wherein, the second hired employee sample includes a second in-service employee sample and a second resigned employee sample;

   After the training of the pre-training hidden Markov model using the training set data, the method further includes:

   Input the static information feature data set in the test set data into the pre-trained hidden Markov model after training for testing, and output the work status feature data set prediction sequence and resignation of each second hired employee sample Intent label value prediction sequence;

   Compare the predicted sequence of the work state characteristic data set with the corresponding work state characteristic data set sequence in the test set data, and compare the resignation intention label value prediction sequence with the corresponding resignation intention in the test set data Align the labeled value sequence to calculate the prediction accuracy;

   When the prediction accuracy rate is greater than or equal to the preset accuracy rate threshold, the training is ended and the model parameters are determined;

   When the prediction accuracy is less than the preset accuracy threshold, return to the step of establishing training set data to optimize the model parameters obtained after training.
4. The method according to any one of claims 1 to 3, wherein the static information feature data set includes any one or a combination of resume information features, interview video features, and written examination information features.
5. The method of claim 4, wherein:

   The method for acquiring the resume information feature includes: inputting resume text into a deep learning resume text graph neural network model, and the deep learning resume text graph neural network model outputs the resume information feature; and/or

   The method for acquiring features of the interview video includes: inputting at least one frame of the image in the interview video into a deep learning facet neural network model, and the deep learning facet neural network model outputs the interview video feature; and/ or

   The method for acquiring the written test information feature includes: inputting the written test text into a deep learning written test text graph neural network model, and the deep learning written test text graph neural network model outputting the written test information feature.
6. The method of claim 1, wherein:

   The method further includes: determining the area category to which the job seeker belongs based on the application data of the job seeker;

   The acquisition of the hidden Markov model used to predict the job status and turnover probability of job applicants includes:

   According to the region category to which the job seeker belongs, call the hidden Markov model corresponding to the region category for predicting the job status and the probability of leaving the job; and/or

   The method further includes: determining the type of working years to which the job applicant belongs based on the application data of the job applicant;

   The acquisition of the hidden Markov model used to predict the job status and turnover probability of job applicants includes:

   According to the working years category of the job applicant, a hidden Markov model corresponding to the working years category for predicting the job status and the probability of leaving the job is invoked.
7. The method according to claim 1, wherein the method further comprises: creating a job application portrait according to the static information feature data set of the job applicant;

   After predicting the job status and resignation probability of the job seeker after entering the job, it further includes:

   When it is predicted that the resignation probability of the job seeker is greater than the preset probability threshold, the corresponding entry time is less than the preset time threshold, setting a non-admission flag on the job seeker portrait of the job seeker;

   When it is predicted that the resignation probability of the job seeker is greater than the preset probability threshold, and the corresponding entry time is greater than or equal to the preset time threshold, an admission mark is set on the job seeker portrait of the job seeker.
8. A resignation prediction device based on a hidden Markov model, wherein the device includes:

   The first acquisition module is used to construct a pre-training hidden Markov model based on the employee’s static information and the influence of the work status on the employee’s intention to leave, and to use the sample data of the hired employee to hide the pre-training The Markov model is trained to obtain the hidden Markov model used to predict the job status and the probability of leaving the job;

   A second acquisition module, the second acquisition module is used to acquire job applicants’ application data, and build a static information feature data set of the job applicants based on the application data;

   A prediction module, which is used to input the static information feature data set into the hidden Markov model to predict the job status of the job applicant and the probability of leaving the job.
9. A computer device includes a memory and a processor, the memory stores program instructions for implementing the following steps, the steps including:

   Based on the influence of the employee’s static information and work status on the employee’s intention to leave, a pre-trained hidden Markov model is constructed, and the pre-trained hidden Markov model is trained based on the sample data of hired employees to obtain candidates for predicting job applicants Hidden Markov model of job status and turnover probability;

   Obtain job applicant's application data, and construct a static information feature data set of the job applicant based on the application data;

   The static information feature data set is input into the hidden Markov model to predict the job status and the resignation probability of the job applicant after entering the job.
10. The computer device according to claim 9, wherein the pre-training hidden Markov model is constructed based on the employee’s static information and the influence of the work status on the employee’s intention to leave, and the pre-training The hidden Markov model is trained to obtain the hidden Markov model used to predict the job status and the probability of leaving the job, including:

    Taking the employee’s static information as input, taking the employee’s working status as the hidden state and the employee’s turnover intention as the observation state, a pre-training hidden Markov model is constructed. The model parameters of the pre-training hidden Markov model include the first The observation probability matrix A, the state transition probability matrix B and the second observation probability matrix C; wherein, the elements in the first observation probability matrix A

    

    Static information indicates when X _i, Y _j probability working state; and the state transition probability matrix element B is

    

    Indicates the probability of the transition from the working state Y _j to the working state Y _k ; the element in the second observation probability matrix C

    

    Represents the static information is X _i, intention to quit the probability of Z _j, i, j and k are positive integers;

    The training set data is established, and the training set data includes a static information feature data set, a work status feature data set sequence, and a resignation intention label value sequence of a plurality of first hired employee samples, and resignation in the resignation intention label value sequence The intention label value corresponds to the work state feature data set in the work state feature data set sequence in a time sequence; wherein, the first hired employee sample includes the first in-service employee sample and the first resigned employee sample;

    Using the training set data to train the pre-training hidden Markov model.
11. The computer device according to claim 10, wherein after said constructing a pre-trained hidden Markov model, it further comprises:

    The test set data is established, and the test set data includes a static information feature data set, a work status feature data set sequence, and a resignation intention label value sequence of a plurality of second hired employee samples, and resignation in the resignation intention label value sequence The intention label value corresponds to the work status feature data set in the work status feature data set sequence in a chronological order; wherein, the second hired employee sample includes a second in-service employee sample and a second resigned employee sample;

    After the training of the pre-training hidden Markov model using the training set data, the method further includes:

    Input the static information feature data set in the test set data into the pre-trained hidden Markov model after training for testing, and output the work status feature data set prediction sequence and resignation of each second hired employee sample Intent label value prediction sequence;

    Compare the predicted sequence of the work state characteristic data set with the corresponding work state characteristic data set sequence in the test set data, and compare the resignation intention label value prediction sequence with the corresponding resignation intention in the test set data Align the labeled value sequence to calculate the prediction accuracy;

    When the prediction accuracy rate is greater than or equal to the preset accuracy rate threshold, the training is ended and the model parameters are determined;

    When the prediction accuracy is less than the preset accuracy threshold, return to the step of establishing training set data to optimize the model parameters obtained after training.
12. The computer device according to any one of claims 9-11, wherein the static information feature data set includes any one or a combination of resume information features, interview video features, and written examination information features.
13. The computer device according to claim 12, wherein:

    The method for obtaining the resume information feature includes: inputting resume text into a deep learning resume text graph neural network model, and the deep learning resume text graph neural network model outputs the resume information feature; and/or

    The method for acquiring features of the interview video includes: inputting at least one frame of the image in the interview video into a deep learning facet neural network model, and the deep learning facet neural network model outputs the interview video feature; and/ or

    The method for acquiring the written test information feature includes: inputting the written test text into a deep learning written test text graph neural network model, and the deep learning written test text graph neural network model outputting the written test information feature.
14. The computer device according to claim 9, wherein:

    The method further includes: determining the area category to which the job seeker belongs based on the application data of the job seeker;

    The acquisition of the hidden Markov model used to predict the job status and turnover probability of job applicants includes:

    According to the region category to which the job seeker belongs, call the hidden Markov model corresponding to the region category for predicting the job status and the probability of leaving the job; and/or

    The method further includes: determining the type of working years to which the job applicant belongs based on the application data of the job applicant;

    The acquiring a hidden Markov model used to predict the job status and turnover probability of job applicants includes:

    According to the working years category of the job applicant, a hidden Markov model corresponding to the working years category for predicting the job status and the probability of leaving the job is invoked.
15. The computer device according to claim 9, wherein the steps performed by the computer device further comprise: creating a job application portrait according to the static information feature data set of the job applicant;

    After predicting the job status and resignation probability of the job seeker after entering the job, it further includes:

    When it is predicted that the resignation probability of the job seeker is greater than the preset probability threshold, the corresponding entry time is less than the preset time threshold, setting a non-admission flag on the job seeker portrait of the job seeker;

    When it is predicted that the resignation probability of the job seeker is greater than the preset probability threshold, and the corresponding entry time is greater than or equal to the preset time threshold, an admission mark is set on the job seeker portrait of the job seeker.
16. A storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the computer-readable instructions implement the following steps, the steps including:

    Based on the influence of the employee’s static information and work status on the employee’s intention to leave, a pre-trained hidden Markov model is constructed, and the pre-trained hidden Markov model is trained based on the sample data of hired employees to obtain candidates for predicting job applicants Hidden Markov model of job status and turnover probability;

    Obtain job applicant's application data, and construct a static information feature data set of the job applicant based on the application data;

    The static information feature data set is input into the hidden Markov model to predict the job status and the resignation probability of the job applicant after entering the job.
17. The storage medium according to claim 16, wherein the pre-training hidden Markov model is constructed based on the static information of the employee and the influence of the work status on the employee's intention to leave, and the pre-training is performed based on the sample data of hired employees The hidden Markov model is trained to obtain the hidden Markov model used to predict the job status and turnover probability of job applicants, including:

    Taking the employee’s static information as input, taking the employee’s working status as the hidden state and the employee’s turnover intention as the observation state, a pre-training hidden Markov model is constructed. The model parameters of the pre-training hidden Markov model include the first The observation probability matrix A, the state transition probability matrix B and the second observation probability matrix C; wherein, the elements in the first observation probability matrix A

    

    Static information indicates when X _i, Y _j probability working state; and the state transition probability matrix element B is

    

    Indicates the probability of the transition from the working state Y _j to the working state Y _k ; the element in the second observation probability matrix C

    

    Represents the static information is X _i, intention to quit the probability of Z _j, i, j and k are positive integers;

    The training set data is established, and the training set data includes a static information feature data set, a work status feature data set sequence, and a resignation intention label value sequence of a plurality of first hired employee samples, and resignation in the resignation intention label value sequence The intention label value corresponds to the work state feature data set in the work state feature data set sequence in a time sequence; wherein, the first hired employee sample includes the first in-service employee sample and the first resigned employee sample;

    Using the training set data to train the pre-training hidden Markov model.
18. The storage medium according to claim 17, wherein, after the construction of the pre-training hidden Markov model, the method further comprises:

    The test set data is established, and the test set data includes a static information feature data set, a work status feature data set sequence, and a resignation intention label value sequence of a plurality of second hired employee samples, and resignation in the resignation intention label value sequence The intention label value corresponds to the work status feature data set in the work status feature data set sequence in a chronological order; wherein, the second hired employee sample includes a second in-service employee sample and a second resigned employee sample;

    After the training of the pre-training hidden Markov model using the training set data, the method further includes:

    Input the static information feature data set in the test set data into the pre-trained hidden Markov model after training for testing, and output the work status feature data set prediction sequence and resignation of each second hired employee sample Intent label value prediction sequence;

    Compare the predicted sequence of the work state characteristic data set with the corresponding work state characteristic data set sequence in the test set data, and compare the resignation intention label value prediction sequence with the corresponding resignation intention in the test set data Align the labeled value sequence to calculate the prediction accuracy;

    When the prediction accuracy rate is greater than or equal to the preset accuracy rate threshold, the training is ended and the model parameters are determined;

    When the prediction accuracy rate is less than the preset accuracy rate threshold, return to the step of establishing training set data to optimize the model parameters obtained after training.
19. The computer device according to any one of claims 16-18, wherein the static information feature data set includes any one or a combination of resume information features, interview video features, and written examination information features.
20. The computer device according to claim 19, wherein:

    The method for acquiring the resume information feature includes: inputting resume text into a deep learning resume text graph neural network model, and the deep learning resume text graph neural network model outputs the resume information feature; and/or

    The method for acquiring features of the interview video includes: inputting at least one frame of the image in the interview video into a deep learning facet neural network model, and the deep learning facet neural network model outputs the interview video feature; and/ or

    The method for acquiring the written test information feature includes: inputting the written test text into a deep learning written test text graph neural network model, and the deep learning written test text graph neural network model outputting the written test information feature.

Images