A method and device for scheduling human resources based on dynamic optimization

Abstract

The present invention discloses a method and device for scheduling human resources based on dynamic optimization, the method comprising: obtaining a job requirement information set and a human resource description information set; the job requirement information set comprises job requirement information; preprocessing the job requirement information set and the human resource description information set to obtain the preprocessed job requirement information set and the human resource description information set; performing job matching processing on the preprocessed job requirement information set and the human resource description information set to obtain a job matching information set; the job matching information set comprises job matching information; the job matching information is used to describe information about matching personnel for a job.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing and artificial intelligence, and in particular to a method and device for scheduling human resources based on dynamic optimization. Background Art

[0002] With the development of science and technology, labor positions are becoming more diversified, and the relevant human resource data of the incumbents' job skills, work experience, ability and quality are also becoming more diversified. Each job position has different requirements for the incumbent's quality, ability, experience, etc. Only when the incumbent has more qualities than these requirements can he be better qualified for the job, thus achieving a win-win situation for the unit and the incumbent.

[0003] For companies that have recruitment needs, they need to screen talents that match the positions from massive and complex data. At present, in order to discover talents, resumes are usually screened manually, especially for important positions in the company, such as managers, directors and other positions. The human resources department will spend a lot of time sorting and selecting personnel. For the matching of existing incumbents and positions, it is necessary to match them based on the similarity between the positions that the incumbents have worked in and the positions to be matched. There is a lot of subjectivity and it cannot achieve the best match between the incumbents and the positions, thereby reducing the company's operating capacity. This method is highly subjective and does not have a unified standard, which reduces the accuracy and efficiency of discovering talents. Summary of the invention

[0004] The present invention mainly solves the problem that the existing human resource scheduling method is highly subjective, has no unified standard, and reduces the accuracy and efficiency of finding talents. The present invention discloses a human resource scheduling method and device based on dynamic optimization.

[0005] In a first aspect of the embodiments of the present application, a method for scheduling human resources based on dynamic optimization is disclosed, comprising:

[0006] S1, obtaining a job requirement information set and a human resource description information set; the job requirement information set includes job requirement information; the human resource description information set includes human resource description information; the job requirement information includes job title, job number, job region, job responsibility description information, strong constraint conditions, and weak constraint conditions; the human resource description information includes name, serial number, professional skill information sequence, personal expertise, course training participated in, work experience information sequence, and award and excellence evaluation information sequence;

[0007] S2, preprocessing the job requirement information set and the human resource description information set to obtain a preprocessed job requirement information set and a human resource description information set;

[0008] S3, performing job matching processing on the pre-processed job requirement information set and human resource description information set to obtain a job matching information set; the job matching information set includes job matching information; the job matching information is used to describe information about matching personnel for a job.

[0009] The preprocessing of the job requirement information set and the human resource description information set to obtain the preprocessed job requirement information set and the human resource description information set includes:

[0010] S21, performing data cleaning processing on the job requirement information set and the human resource description information set to obtain an updated job requirement information set and human resource description information set;

[0011] S22, performing category consistency check processing on the updated job requirement information set and human resource description information set respectively to obtain a pre-processed job requirement information set and human resource description information set.

[0012] The performing job matching processing on the preprocessed job requirement information set and the human resource description information set to obtain a job matching information set includes:

[0013] S31, performing data enhancement processing on the preprocessed job requirement information set to obtain an enhanced job requirement information set;

[0014] S32, using the first correlation model, processing the enhanced job requirement information set and the human resource description information set to obtain a candidate human resource description information set;

[0015] S33, using the trained job matching model, processing the enhanced job requirement information set and the candidate human resource description information set to obtain a job matching information set.

[0016] The step of performing data enhancement processing on the preprocessed job requirement information set to obtain an enhanced job requirement information set includes:

[0017] Using a large language model, rewriting the job responsibility description information in each piece of job requirement information in the preprocessed job requirement information set to obtain rewritten job responsibility description information;

[0018] Using the rewritten job responsibility description information, the job responsibility description information in the corresponding job requirement information is replaced to obtain the added job requirement information;

[0019] All the added job requirement information is added to the preprocessed job requirement information set to obtain an enhanced job requirement information set.

[0020] The method of using the first correlation model to process the enhanced job requirement information set and the human resource description information set to obtain a candidate human resource description information set includes:

[0021] The initialization judgment sequence number value is 1; confirming that the initialized candidate human resource description information set is a human resource description information set;

[0022] Confirm that the job number in the enhanced job requirement information set is the job requirement information with the judgment sequence value, and is the job requirement information to be processed;

[0023] For each human resource description information in the candidate human resource description information set, determine whether its job experience contains the job title of the to-be-processed job requirement information, and obtain a first determination result; for human resource description information for which the first determination result is negative, delete it from the candidate human resource description information set;

[0024] For each human resource description information in the candidate human resource description information set, determine whether its professional skills include the job responsibilities description information of the to-be-processed job requirement information, and obtain a second determination result; for human resource description information for which the second determination result is negative, delete it from the candidate human resource description information set;

[0025] For each human resource description information in the candidate human resource description information set, determine whether it satisfies the weak constraint condition of the to-be-processed job requirement information to obtain a third determination result; for human resource description information for which the third determination result is negative, delete it from the candidate human resource description information set;

[0026] Performing boundary calculation processing on each human resource description information in the candidate human resource description information set and the boundary range of each constraint item of the strong constraint condition of the to-be-processed job requirement information to obtain a boundary distance value of the human resource description information;

[0027] The boundary calculation process, its calculation expression is:

[0028]

[0029] Among them, ds represents the calculated boundary distance value, a i Indicates the value of the corresponding data of the i-th constraint item of the strong constraint condition in the human resources description information, lmin i and lmax iThey respectively represent the lower bound and upper bound of the value of the ith constraint item of the strong constraint condition, and N1 is the number of constraint items of the strong constraint condition;

[0030] For each human resource description information in the candidate human resource description information set, determine whether the boundary distance value is greater than a preset distance threshold value to obtain a fourth determination result; for human resource description information for which the fourth determination result is negative, delete the human resource description information from the candidate human resource description information set;

[0031] Increasing the judgment sequence number value by 1;

[0032] Determine whether the judgment sequence number value is greater than the number of job requirement information in the enhanced job requirement information set to obtain a fifth judgment result; when the fifth judgment result is no, trigger execution to confirm that the job requirement information with the job number in the enhanced job requirement information set as the judgment sequence number value is the job requirement information to be processed; when the fifth judgment result is yes, obtain a candidate human resource description information set.

[0033] The job matching model includes: a feature encoding module, a feature fusion module, and a prediction module;

[0034] The feature encoding module includes a first residual multi-head self-attention module, a second residual multi-head self-attention module, a third residual multi-head self-attention module, a fourth residual multi-head self-attention module and a multi-head mutual attention module;

[0035] The first residual multi-head self-attention module, the second residual multi-head self-attention module, the third residual multi-head self-attention module, and the fourth residual multi-head self-attention module are connected to the multi-head mutual attention module; the first residual multi-head self-attention module, the second residual multi-head self-attention module, the third residual multi-head self-attention module, and the fourth residual multi-head self-attention module are respectively used to receive the information sequence of the strong constraint conditions in the job requirement information, the professional skills information sequence, the personal expertise information sequence, and the award and excellence information sequence in the human resources description information, and perform self-attention feature extraction on the received information sequence to obtain corresponding feature information; the information sequence of the strong constraint conditions is to represent all constraint items of the strong constraint conditions as an information sequence;

[0036] The multi-head mutual attention module is used to extract multi-head mutual attention features from the feature information corresponding to each information sequence to obtain mutual attention features;

[0037] The feature encoding module is used to extract features from various input information sequences to obtain mutual attention features;

[0038] The feature fusion module is used to fuse the mutual attention features to obtain fused features; the output end of the feature fusion module is connected to the input end of the prediction module;

[0039] The feature fusion module includes a first input module, a first convolution module, a depth-separable convolution module, a first dimension-raising convolution module, a second dimension-raising convolution module, a third dimension-raising convolution module, a fourth dimension-raising convolution module, a second convolution module, a first pooling module, a third convolution module and a first fully connected module;

[0040] The input end of the first input module is used to receive the mutual attention feature; the output end of the first input module of the feature fusion module is connected to the input end of the first convolution module of the feature fusion module; the output end of the first convolution module of the feature fusion module is connected to the input end of the depthwise separable convolution module of the feature fusion module; the output end of the depthwise separable convolution module of the feature fusion module is connected to the input end of the first dimensionality-raising convolution module of the feature fusion module;

[0041] The output end of the first dimensionality-raising convolution module of the feature fusion module is connected to the input end of the second dimensionality-raising convolution module of the feature fusion module; the output end of the second dimensionality-raising convolution module of the feature fusion module is connected to the input end of the third dimensionality-raising convolution module of the feature fusion module; the output end of the third dimensionality-raising convolution module of the feature fusion module is connected to the input end of the fourth dimensionality-raising convolution module of the feature fusion module; the output end of the fourth dimensionality-raising convolution module of the feature fusion module is connected to the input end of the second convolution module of the feature fusion module; the output end of the second convolution module of the feature fusion module is connected to the input end of the first pooling module of the feature fusion module; the output end of the first pooling module of the feature fusion module is connected to the input end of the third convolution module of the feature fusion module; the output end of the third convolution module of the feature fusion module is connected to the input end of the first fully connected module of the feature fusion module;

[0042] The output end of the feature encoding module is connected to the input end of the feature fusion module;

[0043] The prediction module includes: a second input module, a fourth convolution module, a fifth convolution module, a sixth convolution module, a second pooling module, a seventh convolution module, a second fully connected module and a third fully connected module;

[0044] The input end of the second input module of the prediction module is connected to the output end of the second dimensionality-raising convolution module of the feature fusion module; the output end of the second input module of the prediction module is connected to the input end of the fourth convolution module of the prediction module; the output end of the fourth convolution module of the prediction module is connected to the input end of the fifth convolution module of the prediction module; the output end of the fifth convolution module of the prediction module is connected to the input end of the sixth convolution module of the prediction module; the output end of the sixth convolution module of the prediction module is connected to the input end of the second pooling module of the prediction module; the output end of the second pooling module of the prediction module is connected to the input end of the seventh convolution module of the prediction module; the output end of the seventh convolution module of the prediction module is connected to the input end of the second fully connected module of the prediction module; the output end of the second fully connected module of the prediction module is connected to the input end of the third fully connected module of the prediction module;

[0045] The prediction module is used to perform prediction processing on the fusion feature to obtain the human resource description information matched by the input job requirement information; the output end of the third fully connected module of the prediction module is used to output the predicted human resource description information matched by the input job requirement information;

[0046] The human resource description information matched by all the job requirement information in the enhanced job requirement information set is the job matching information set.

[0047] The training process of the job matching model includes:

[0048] S331, initializing the number of iterations; presetting the training number threshold;

[0049] S332, obtaining job requirement information and human resource description information from the job matching training data set; determining the job requirement information and human resource description information as input data; determining matching human resource information corresponding to the job requirement information in the job matching training data set as label data;

[0050] S333, using the job matching model, processing the input data to obtain predicted human resource description information;

[0051] S334, calculating and obtaining a difference information value between the predicted human resource description information and the label data;

[0052] S335, judging whether the difference information value satisfies a convergence condition, and obtaining a sixth judgment result;

[0053] When the sixth judgment result is no, judging whether the number of iterations is equal to the training number threshold, and obtaining a seventh judgment result;

[0054] When the seventh judgment result is no, determining that the model training state does not meet the training termination condition;

[0055] When the seventh judgment result is yes, determining that the model training state satisfies the training termination condition;

[0056] When the sixth judgment result is yes, determining that the model training state satisfies the training termination condition;

[0057] When the model training state does not meet the termination training condition, the parameter updating model is used to update the parameters of the feature fusion module and the prediction module, so that the number of iterations is increased by 1, and S332 is executed;

[0058] When the model training state satisfies the training termination condition, the training process of the job matching model is completed to obtain a trained job matching model.

[0059] The parameter updating model is:

[0060]

[0061] θ←θ+v;

[0062] In the formula, x (i) is the input data corresponding to the i-th training sample in the job matching training data set, y (i) is the label information corresponding to the i-th training sample in the job matching training data set, v is the parameter update value, θ is the parameter of the feature fusion module and the prediction module, η is the initial parameter learning rate, α is the momentum angle parameter, 0≤α≤π / 4, ▽ θ It means to find the partial derivative of variable θ, f(x (i) ; θ) represents the predicted human resource description information obtained by the job matching model for the input data corresponding to the i-th training sample in the job matching training data set, and f(·) is the calculation function corresponding to the job matching model.

[0063] In a second aspect of the embodiment of the present application, a scheduling device for human resources based on dynamic optimization is disclosed, the device comprising:

[0064] A memory storing executable program code;

[0065] a processor coupled to the memory;

[0066] The processor calls the executable program code stored in the memory to execute the human resource scheduling method based on dynamic optimization.

[0067] In a third aspect of the embodiments of the present application, a computer storable medium is disclosed, wherein the computer storable medium stores computer instructions, and when the computer instructions are called, they are used to execute the human resource scheduling method based on dynamic optimization.

[0068] The beneficial effects of the present invention are:

[0069] The present invention utilizes a dynamic optimization method to achieve the best match between job requirements and human resources, solves the problems of strong subjectivity and low matching accuracy in traditional matching methods, and improves the accuracy of talent matching.

[0070] Before performing job matching, the present invention first uses strong optimization conditions or weak optimization conditions to perform multiple rounds of screening and processing on human resource demand data, filters out data with low matching degrees, and improves the work efficiency of human resource matching. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a flow chart for implementing the method of the present invention. DETAILED DESCRIPTION

[0072] In order to better understand the content of the present invention, an embodiment is given here.

[0073] Figure 1 It is a flow chart for implementing the method of the present invention.

[0074] In a first aspect of the embodiments of the present application, a method for scheduling human resources based on dynamic optimization is disclosed, comprising:

[0075] S1, obtaining a job requirement information set and a human resource description information set; the job requirement information set includes job requirement information; the human resource description information set includes human resource description information; the job requirement information includes job title, job number, job region, job responsibility description information, strong constraint conditions, and weak constraint conditions; the human resource description information includes name, serial number, height, weight, vision, professional skill information sequence, personal expertise, courses and trainings participated in, work experience information sequence, award and excellence information sequence, and colleague evaluation;

[0076] S2, preprocessing the job requirement information set and the human resource description information set to obtain a preprocessed job requirement information set and a human resource description information set;

[0077] S3, performing job matching processing on the pre-processed job requirement information set and the human resource description information set to obtain a job matching information set; the job matching information set includes job matching information; the job matching information is used to describe information about matching personnel for the job;

[0078] The job requirement information set and the human resource description information set can be obtained by searching a database, or by collecting the job requirement information set through sensors installed in the workplace; the human resource description information set can be obtained by digitally converting the human resource information. The digital conversion can be achieved through a scanner and digital text recognition software.

[0079] The preprocessing of the job requirement information set and the human resource description information set to obtain the preprocessed job requirement information set and the human resource description information set includes:

[0080] S21, performing data cleaning processing on the job requirement information set and the human resource description information set to obtain an updated job requirement information set and human resource description information set;

[0081] S22, performing category consistency check processing on the updated job requirement information set and human resource description information set, respectively, to obtain a pre-processed job requirement information set and human resource description information set;

[0082] The data cleaning process includes filling missing values, smoothing noise data, and smoothing or deleting outliers; the smoothing of noise data is to first identify the noise data, and then smooth the noise data according to the data before and after the noise data; the outlier point can be identified by using a preset data value range, and the data that is not within the value range is an outlier point. The filling value of the missing value can be determined by averaging the measured values ​​within a certain sampling interval before and after the missing value.

[0083] The category consistency check process is to determine whether the data attributes of each category of data in the updated job requirement information set and human resource description information set are within the data attribute range of the category data, and obtain a judgment result; the data with a judgment result of no will be deleted from the updated job requirement information set and human resource description information set.

[0084] The data attributes include nominal attributes, text attributes, numeric attributes, sequential attributes, interval attributes, exhaustive attributes, etc.;

[0085] For example, the data attributes of job title and name are nominal attributes; the data attributes of job number and serial number are ordinal attributes; the data attributes of job region and course training attended are exhaustive attributes; the data attributes of professional skill information sequence, personal specialty information sequence, and award and excellence information sequence are digital attributes; by quantifying professional skills, personal specialty, and award and excellence, the corresponding digital sequence is obtained as the relevant information in the human resource description information, and this information is a digital attribute;

[0086] The data attributes of job description information and colleagues' evaluation of work experience are text attributes; the data attributes of strong constraints and weak constraints are interval attributes, and the corresponding data attributes in the human resources description information are data attributes; the data attributes of height, weight, and vision are numeric attributes.

[0087] The performing job matching processing on the preprocessed job requirement information set and the human resource description information set to obtain a job matching information set includes:

[0088] S31, performing data enhancement processing on the preprocessed job requirement information set to obtain an enhanced job requirement information set;

[0089] S32, using the first correlation model, processing the enhanced job requirement information set and the human resource description information set to obtain a candidate human resource description information set;

[0090] S33, using the trained job matching model, processing the enhanced job requirement information set and the candidate human resource description information set to obtain a job matching information set;

[0091] The method of using the trained job matching model to process the enhanced job requirement information set and the candidate human resource description information set is to use the information sequence of strong constraints in the job requirement information, the professional skill information sequence, personal expertise information sequence and award and excellence information sequence in the human resource description information as input data and input them into the trained job matching model.

[0092] The step of performing data enhancement processing on the preprocessed job requirement information set to obtain an enhanced job requirement information set includes:

[0093] Using a large language model, rewriting the job responsibility description information in each piece of job requirement information in the preprocessed job requirement information set to obtain rewritten job responsibility description information;

[0094] Using the rewritten job responsibility description information, the job responsibility description information in the corresponding job requirement information is replaced to obtain the added job requirement information;

[0095] All the added job requirement information is added to the preprocessed job requirement information set to obtain an enhanced job requirement information set.

[0096] The use of the large language model to rewrite the job responsibility description information in each piece of job requirement information in the preprocessed job requirement information set to obtain the rewritten job responsibility description information may be:

[0097] Speak the following sentence and input it into the big language model to get the rewritten job description information:

[0098] Please rewrite the following sentences and give three ways of describing them: + [Job description information]

[0099] Please reorganize the following sentences and provide three different descriptions: + [job description information];

[0100] Please rewrite the following statement and provide three different ways of describing it: + [Job description information];

[0101] Please rephrase the following sentence and list three different ways of describing it: + [job description information];

[0102] Please reorganize the following description and provide three different narrative modes: + [Job description information].

[0103] The large language model can be the Wenxinyiyan large model, the Tongyi large model, and the Zhipu large model.

[0104] The method of using the first correlation model to process the enhanced job requirement information set and the human resource description information set to obtain a candidate human resource description information set includes:

[0105] The initialization judgment sequence number value is 1; confirming that the initialized candidate human resource description information set is a human resource description information set;

[0106] Confirm that the job number in the enhanced job requirement information set is the job requirement information with the judgment sequence value, and is the job requirement information to be processed;

[0107] For each human resource description information in the candidate human resource description information set, determine whether its job experience contains the job title of the to-be-processed job requirement information, and obtain a first determination result; for human resource description information for which the first determination result is negative, delete it from the candidate human resource description information set;

[0108] For each human resource description information in the candidate human resource description information set, determine whether its professional skills include the job responsibilities description information of the to-be-processed job requirement information, and obtain a second determination result; for human resource description information for which the second determination result is negative, delete it from the candidate human resource description information set;

[0109] For each human resource description information in the candidate human resource description information set, determine whether it satisfies the weak constraint condition of the to-be-processed job requirement information to obtain a third determination result; for human resource description information for which the third determination result is negative, delete it from the candidate human resource description information set;

[0110] Performing boundary calculation processing on each human resource description information in the candidate human resource description information set and the boundary range of each constraint item of the strong constraint condition of the to-be-processed job requirement information to obtain a boundary distance value of the human resource description information;

[0111] The boundary calculation process, its calculation expression is:

[0112]

[0113] Among them, ds represents the calculated boundary distance value, a i Indicates the value of the corresponding data of the i-th constraint item of the strong constraint condition in the human resources description information, lmin i and lmax i They respectively represent the lower bound and upper bound of the value of the ith constraint item of the strong constraint condition, and N1 is the number of constraint items of the strong constraint condition;

[0114] For each human resource description information in the candidate human resource description information set, determine whether the boundary distance value is greater than a preset distance threshold value to obtain a fourth determination result; for human resource description information for which the fourth determination result is negative, delete the human resource description information from the candidate human resource description information set;

[0115] Increasing the judgment sequence number value by 1;

[0116] Determine whether the judgment sequence number value is greater than the number of job requirement information in the enhanced job requirement information set to obtain a fifth judgment result; when the fifth judgment result is no, trigger execution to confirm that the job requirement information in the enhanced job requirement information set whose job number is the judgment sequence number value is the job requirement information to be processed; when the fifth judgment result is yes, obtain a candidate human resource description information set;

[0117] The job matching model includes: a feature encoding module, a feature fusion module, and a prediction module;

[0118] The feature encoding module includes a first residual multi-head self-attention module, a second residual multi-head self-attention module, a third residual multi-head self-attention module, a fourth residual multi-head self-attention module and a multi-head mutual attention module;

[0119] The first residual multi-head self-attention module, the second residual multi-head self-attention module, the third residual multi-head self-attention module, and the fourth residual multi-head self-attention module are connected to the multi-head mutual attention module; the first residual multi-head self-attention module, the second residual multi-head self-attention module, the third residual multi-head self-attention module, and the fourth residual multi-head self-attention module are respectively used to receive the information sequence of the strong constraint conditions in the job requirement information, the professional skills information sequence, the personal expertise information sequence, and the award and excellence information sequence in the human resources description information, and perform self-attention feature extraction on the received information sequence to obtain corresponding feature information; the information sequence of the strong constraint conditions is to represent all constraint items of the strong constraint conditions as an information sequence;

[0120] The multi-head mutual attention module is used to extract multi-head mutual attention features from the feature information corresponding to each information sequence to obtain mutual attention features;

[0121] The feature encoding module is used to extract features from various input information sequences to obtain mutual attention features;

[0122] The feature fusion module is used to fuse the mutual attention features to obtain fused features; the output end of the feature fusion module is connected to the input end of the prediction module;

[0123] The feature fusion module includes a first input module, a first convolution module, a depth-separable convolution module, a first dimension-raising convolution module, a second dimension-raising convolution module, a third dimension-raising convolution module, a fourth dimension-raising convolution module, a second convolution module, a first pooling module, a third convolution module and a first fully connected module;

[0124] The input end of the first input module is used to receive the mutual attention feature; the output end of the first input module of the feature fusion module is connected to the input end of the first convolution module of the feature fusion module; the output end of the first convolution module of the feature fusion module is connected to the input end of the depthwise separable convolution module of the feature fusion module; the output end of the depthwise separable convolution module of the feature fusion module is connected to the input end of the first dimensionality-raising convolution module of the feature fusion module;

[0125] The output end of the first dimensionality-raising convolution module of the feature fusion module is connected to the input end of the second dimensionality-raising convolution module of the feature fusion module; the output end of the second dimensionality-raising convolution module of the feature fusion module is connected to the input end of the third dimensionality-raising convolution module of the feature fusion module; the output end of the third dimensionality-raising convolution module of the feature fusion module is connected to the input end of the fourth dimensionality-raising convolution module of the feature fusion module; the output end of the fourth dimensionality-raising convolution module of the feature fusion module is connected to the input end of the second convolution module of the feature fusion module; the output end of the second convolution module of the feature fusion module is connected to the input end of the first pooling module of the feature fusion module; the output end of the first pooling module of the feature fusion module is connected to the input end of the third convolution module of the feature fusion module; the output end of the third convolution module of the feature fusion module is connected to the input end of the first fully connected module of the feature fusion module;

[0126] The output end of the feature encoding module is connected to the input end of the feature fusion module;

[0127] The prediction module includes: a second input module, a fourth convolution module, a fifth convolution module, a sixth convolution module, a second pooling module, a seventh convolution module, a second fully connected module and a third fully connected module;

[0128] The input end of the second input module of the prediction module is connected to the output end of the second dimensionality-raising convolution module of the feature fusion module; the output end of the second input module of the prediction module is connected to the input end of the fourth convolution module of the prediction module; the output end of the fourth convolution module of the prediction module is connected to the input end of the fifth convolution module of the prediction module; the output end of the fifth convolution module of the prediction module is connected to the input end of the sixth convolution module of the prediction module; the output end of the sixth convolution module of the prediction module is connected to the input end of the second pooling module of the prediction module; the output end of the second pooling module of the prediction module is connected to the input end of the seventh convolution module of the prediction module; the output end of the seventh convolution module of the prediction module is connected to the input end of the second fully connected module of the prediction module; the output end of the second fully connected module of the prediction module is connected to the input end of the third fully connected module of the prediction module.

[0129] The prediction module is used to perform prediction processing on the fusion feature to obtain the human resource description information matched by the input job requirement information; the output end of the third fully connected module of the prediction module is used to output the predicted human resource description information matched by the input job requirement information;

[0130] The human resource description information matched by all the job requirement information in the enhanced job requirement information set is the job matching information set;

[0131] The training process of the job matching model includes:

[0132] S331, initializing the number of iterations; presetting the training number threshold;

[0133] S332, obtaining job requirement information and human resource description information from the job matching training data set; determining the job requirement information and human resource description information as input data; determining matching human resource information corresponding to the job requirement information in the job matching training data set as label data;

[0134] S333, using the job matching model, processing the input data to obtain predicted human resource description information;

[0135] S334, calculating and obtaining a difference information value between the predicted human resource description information and the label data;

[0136] S303, judging whether the difference information value satisfies a convergence condition, and obtaining a sixth judgment result;

[0137] When the sixth judgment result is no, judging whether the number of iterations is equal to the training number threshold, and obtaining a seventh judgment result;

[0138] When the seventh judgment result is no, determining that the model training state does not meet the training termination condition;

[0139] When the seventh judgment result is yes, determining that the model training state satisfies the training termination condition;

[0140] When the sixth judgment result is yes, determining that the model training state satisfies the training termination condition;

[0141] When the model training state does not meet the termination training condition, the parameter updating model is used to update the parameters of the feature fusion module and the prediction module, so that the number of iterations increases by 1, and S332 is executed;

[0142] When the model training state satisfies the training termination condition, the training process of the job matching model is completed to obtain a trained job matching model.

[0143] The job matching training data set may be an open source job matching data set on the Internet.

[0144] The calculation of the difference information value may adopt a cross entropy loss function.

[0145] The convergence condition refers to that the difference information value is less than a preset judgment value.

[0146] The first residual multi-head self-attention module, the second residual multi-head self-attention module, the third residual multi-head self-attention module, and the fourth residual multi-head self-attention module can be a Transformer model.

[0147] The multi-head mutual attention module is implemented using the Mutli Self-Attention module.

[0148] The job matching training data set may adopt a historical data set of job matching, or obtain an open source human resources text data set on the Internet, and obtain the job matching training data set by adding job tags to the human resources text data set.

[0149] The parameter updating model is:

[0150]

[0151] θ←θ+v;

[0152] In the formula, x (i) is the input data corresponding to the i-th training sample in the job matching training data set, y (i) is the label information corresponding to the i-th training sample in the job matching training data set, v is the parameter update value, θ is the parameter of the feature fusion module and the prediction module, η is the initial parameter learning rate, α is the momentum angle parameter, 0≤α≤π / 4, ▽ θ It means to find the partial derivative of variable θ, f(x (i) ; θ) represents the predicted human resource description information obtained by the job matching model for the input data corresponding to the i-th training sample in the job matching training data set, and f(·) is the calculation function corresponding to the job matching model;

[0153] The training samples of the job matching training data set include: job requirement information, human resource description information, and matching human resource information corresponding to the job requirement information.

[0154] In a second aspect of the embodiment of the present application, a scheduling device for human resources based on dynamic optimization is disclosed, the device comprising:

[0155] A memory storing executable program code;

[0156] a processor coupled to the memory;

[0157] The processor calls the executable program code stored in the memory to execute the human resource scheduling method based on dynamic optimization.

[0158] In a third aspect of the embodiments of the present application, a computer storable medium is disclosed, wherein the computer storable medium stores computer instructions, and when the computer instructions are called, they are used to execute the human resource scheduling method based on dynamic optimization.

[0159] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for scheduling human resources based on dynamic optimization, characterized in that: include: S1, obtaining a job requirement information set and a human resource description information set; the job requirement information set includes job requirement information; the human resource description information set includes human resource description information; the job requirement information includes job title, job number, job region, job responsibility description information, strong constraint conditions, and weak constraint conditions; the human resource description information includes name, serial number, professional skill information sequence, personal expertise, course training participated in, work experience information sequence, and award and excellence evaluation information sequence; S2, preprocessing the job requirement information set and the human resource description information set to obtain a preprocessed job requirement information set and a human resource description information set; S3, performing job matching processing on the pre-processed job requirement information set and the human resource description information set to obtain a job matching information set; the job matching information set includes job matching information; the job matching information is used to describe information about matching personnel for the job; The performing job matching processing on the preprocessed job requirement information set and the human resource description information set to obtain a job matching information set includes: S31, performing data enhancement processing on the preprocessed job requirement information set to obtain an enhanced job requirement information set; S32, using the first correlation model, processing the enhanced job requirement information set and the human resource description information set to obtain a candidate human resource description information set; S33, using the trained job matching model, processing the enhanced job requirement information set and the candidate human resource description information set to obtain a job matching information set; The method of using the first correlation model to process the enhanced job requirement information set and the human resource description information set to obtain a candidate human resource description information set includes: The initialization judgment sequence number value is 1; confirming that the initialized candidate human resource description information set is a human resource description information set; Confirm that the job number in the enhanced job requirement information set is the job requirement information with the judgment sequence value, and is the job requirement information to be processed; For each human resource description information in the candidate human resource description information set, determine whether its job experience contains the job title of the to-be-processed job requirement information, and obtain a first determination result; for human resource description information for which the first determination result is negative, delete it from the candidate human resource description information set; For each human resource description information in the candidate human resource description information set, determine whether its professional skills include the job responsibilities description information of the to-be-processed job requirement information, and obtain a second determination result; for human resource description information for which the second determination result is negative, delete it from the candidate human resource description information set; For each human resource description information in the candidate human resource description information set, determine whether it satisfies the weak constraint condition of the to-be-processed job requirement information to obtain a third determination result; for human resource description information for which the third determination result is negative, delete it from the candidate human resource description information set; Performing boundary calculation processing on each human resource description information in the candidate human resource description information set and the boundary range of each constraint item of the strong constraint condition of the to-be-processed job requirement information to obtain a boundary distance value of the human resource description information; The boundary calculation process, its calculation expression is: Among them, ds represents the calculated boundary distance value, a i Indicates the value of the corresponding data of the i-th constraint item of the strong constraint condition in the human resources description information, lmin i and lmax i They respectively represent the lower bound and upper bound of the value of the ith constraint item of the strong constraint condition, and N1 is the number of constraint items of the strong constraint condition; For each human resource description information in the candidate human resource description information set, determine whether the boundary distance value is greater than a preset distance threshold value to obtain a fourth determination result; for human resource description information for which the fourth determination result is negative, delete the human resource description information from the candidate human resource description information set; Increasing the judgment sequence number value by 1; Determine whether the judgment sequence number value is greater than the number of job requirement information in the enhanced job requirement information set to obtain a fifth judgment result; when the fifth judgment result is no, trigger execution to confirm that the job requirement information with the job number in the enhanced job requirement information set as the judgment sequence number value is the job requirement information to be processed; when the fifth judgment result is yes, obtain a candidate human resource description information set.

2. The method for scheduling human resources based on dynamic optimization according to claim 1, characterized in that: The preprocessing of the job requirement information set and the human resource description information set to obtain the preprocessed job requirement information set and the human resource description information set includes: S21, performing data cleaning processing on the job requirement information set and the human resource description information set to obtain an updated job requirement information set and human resource description information set; S22, performing category consistency check processing on the updated job requirement information set and human resource description information set respectively to obtain a pre-processed job requirement information set and human resource description information set.

3. The method for scheduling human resources based on dynamic optimization according to claim 1, characterized in that: The step of performing data enhancement processing on the preprocessed job requirement information set to obtain an enhanced job requirement information set includes: Using a large language model, rewriting the job responsibility description information in each piece of job requirement information in the preprocessed job requirement information set to obtain rewritten job responsibility description information; Using the rewritten job responsibility description information, the job responsibility description information in the corresponding job requirement information is replaced to obtain the added job requirement information; All the added job requirement information is added to the preprocessed job requirement information set to obtain an enhanced job requirement information set.

4. The method for scheduling human resources based on dynamic optimization according to claim 1, characterized in that: The job matching model includes: a feature encoding module, a feature fusion module, and a prediction module; The feature encoding module includes a first residual multi-head self-attention module, a second residual multi-head self-attention module, a third residual multi-head self-attention module, a fourth residual multi-head self-attention module and a multi-head mutual attention module; The first residual multi-head self-attention module, the second residual multi-head self-attention module, the third residual multi-head self-attention module, and the fourth residual multi-head self-attention module are connected to the multi-head mutual attention module; the first residual multi-head self-attention module, the second residual multi-head self-attention module, the third residual multi-head self-attention module, and the fourth residual multi-head self-attention module are respectively used to receive the information sequence of the strong constraint conditions in the job requirement information, the professional skills information sequence, the personal expertise information sequence, and the award and excellence information sequence in the human resources description information, and perform self-attention feature extraction on the received information sequence to obtain corresponding feature information; the information sequence of the strong constraint conditions is to represent all constraint items of the strong constraint conditions as an information sequence; The multi-head mutual attention module is used to extract multi-head mutual attention features from feature information corresponding to each information sequence to obtain mutual attention features; The feature encoding module is used to extract features from various input information sequences to obtain mutual attention features; The feature fusion module is used to fuse the mutual attention features to obtain fused features; the output end of the feature fusion module is connected to the input end of the prediction module; The feature fusion module includes a first input module, a first convolution module, a depth-separable convolution module, a first dimension-raising convolution module, a second dimension-raising convolution module, a third dimension-raising convolution module, a fourth dimension-raising convolution module, a second convolution module, a first pooling module, a third convolution module and a first fully connected module; The input end of the first input module is used to receive the mutual attention feature; the output end of the first input module of the feature fusion module is connected to the input end of the first convolution module of the feature fusion module; the output end of the first convolution module of the feature fusion module is connected to the input end of the depthwise separable convolution module of the feature fusion module; the output end of the depthwise separable convolution module of the feature fusion module is connected to the input end of the first dimensionality-raising convolution module of the feature fusion module; The output end of the first dimensionality-raising convolution module of the feature fusion module is connected to the input end of the second dimensionality-raising convolution module of the feature fusion module; the output end of the second dimensionality-raising convolution module of the feature fusion module is connected to the input end of the third dimensionality-raising convolution module of the feature fusion module; the output end of the third dimensionality-raising convolution module of the feature fusion module is connected to the input end of the fourth dimensionality-raising convolution module of the feature fusion module; the output end of the fourth dimensionality-raising convolution module of the feature fusion module is connected to the input end of the second convolution module of the feature fusion module; the output end of the second convolution module of the feature fusion module is connected to the input end of the first pooling module of the feature fusion module; the output end of the first pooling module of the feature fusion module is connected to the input end of the third convolution module of the feature fusion module; the output end of the third convolution module of the feature fusion module is connected to the input end of the first fully connected module of the feature fusion module; The output end of the feature encoding module is connected to the input end of the feature fusion module; The prediction module includes: a second input module, a fourth convolution module, a fifth convolution module, a sixth convolution module, a second pooling module, a seventh convolution module, a second fully connected module and a third fully connected module; The input end of the second input module of the prediction module is connected to the output end of the second dimensionality-raising convolution module of the feature fusion module; the output end of the second input module of the prediction module is connected to the input end of the fourth convolution module of the prediction module; the output end of the fourth convolution module of the prediction module is connected to the input end of the fifth convolution module of the prediction module; the output end of the fifth convolution module of the prediction module is connected to the input end of the sixth convolution module of the prediction module; the output end of the sixth convolution module of the prediction module is connected to the input end of the second pooling module of the prediction module; the output end of the second pooling module of the prediction module is connected to the input end of the seventh convolution module of the prediction module; the output end of the seventh convolution module of the prediction module is connected to the input end of the second fully connected module of the prediction module; the output end of the second fully connected module of the prediction module is connected to the input end of the third fully connected module of the prediction module; The prediction module is used to perform prediction processing on the fusion feature to obtain the human resource description information matched by the input job requirement information; the output end of the third fully connected module of the prediction module is used to output the predicted human resource description information matched by the input job requirement information; The human resource description information matched by all the job requirement information in the enhanced job requirement information set is the job matching information set.

5. The method for scheduling human resources based on dynamic optimization as claimed in claim 4, characterized in that: The training process of the job matching model includes: S331, initializing the number of iterations; presetting the training number threshold; S332, obtaining job requirement information and human resource description information from the job matching training data set; determining the job requirement information and human resource description information as input data; determining matching human resource information corresponding to the job requirement information in the job matching training data set as label data; S333, using the job matching model, processing the input data to obtain predicted human resource description information; S334, calculating and obtaining a difference information value between the predicted human resource description information and the label data; S335, judging whether the difference information value satisfies a convergence condition, and obtaining a sixth judgment result; When the sixth judgment result is no, judging whether the number of iterations is equal to the training number threshold, and obtaining a seventh judgment result; When the seventh judgment result is no, determining that the model training state does not meet the training termination condition; When the seventh judgment result is yes, determining that the model training state satisfies the training termination condition; When the sixth judgment result is yes, determining that the model training state satisfies the training termination condition; When the model training state does not meet the termination training condition, the parameter updating model is used to update the parameters of the feature fusion module and the prediction module, so that the number of iterations is increased by 1, and S332 is executed; When the model training state satisfies the training termination condition, the training process of the job matching model is completed to obtain a trained job matching model.

6. The method for scheduling human resources based on dynamic optimization as claimed in claim 5, characterized in that: The parameter updating model is: θ←θ+v; In the formula, x (i) is the input data corresponding to the i-th training sample in the job matching training data set, y (i) is the label information corresponding to the i-th training sample in the job matching training data set, v is the parameter update value, θ is the parameter of the feature fusion module and the prediction module, η is the initial parameter learning rate, α is the momentum angle parameter, 0≤α≤π / 4, It means to find the partial derivative of variable θ, f(x (i) ; θ) represents the predicted human resource description information obtained by the job matching model for the input data corresponding to the i-th training sample in the job matching training data set, and f(·) is the calculation function corresponding to the job matching model.

7. A human resource scheduling device based on dynamic optimization, characterized in that: The device comprises: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the human resource scheduling method based on dynamic optimization according to any one of claims 1 to 6.

8. A computer storable medium, characterized in that: The computer storable medium stores computer instructions, and when the computer instructions are called, they are used to execute the human resource scheduling method based on dynamic optimization as described in any one of claims 1 to 6.

Images