Intelligent modeling and predicting method suitable for human resource supply and demand relationship management

Abstract

The invention discloses an intelligent modeling and prediction method suitable for human resource supply-demand relationship management. The method comprises the steps of determining human resource information to be predicted; constructing a self-organizing deep neural network supply-demand relationship prediction model, inputting the to-be-predicted human resource information into the self-organizing deep neural network supply-demand relationship prediction model for processing, and outputting a prediction result of a human resource supply-demand matching state; wherein the self-organizing deep neural network supply-demand relation prediction model is obtained after training is carried out through a training set, cross validation is carried out through a validation set and testing is carried out through a testing set, and the training set, the validation set and the testing set are all human resource information feature variables and matching states representing the human resource relation. According to the method, a deep neural network model is designed and trained to fit a human resource supply and demand relationship state, and approximation of human resource supply and demand relationship dynamic characteristics is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence empowered human resource supply - demand relationship management and analysis, and particularly relates to an intelligent modeling and prediction method applicable to human resource supply - demand relationship management. Background Art

[0002] As the country enters a critical period of development, along with the changes in the macro - economic growth mode and the enterprise employment structure, new challenges are posed to enterprise human resource managers. How to accurately judge the impact of the external environment on the enterprise's human resource supply - demand relationship and then effectively respond to the situation faced is a problem that human resource managers and participants need to face together, and it is also an important topic that needs to be studied when enterprises formulate human resource plans.

[0003] In the process of human resource planning and management, the human resource supply - demand relationship is an important research content, which directly affects the operation and development of the human resource market. Essentially, the human resource supply - demand relationship is a multi - factor complex dynamic system, involving the interaction of dynamic processes such as economic operation, population quantity change, and job - person change, and containing human behavior and decision - making, with characteristics such as strong non - linearity and strong coupling. Therefore, realizing intelligent modeling and prediction of the human resource supply - demand relationship is a challenging problem. Currently, the most widely used research method for human resource supply - demand relationship is mainly statistical analysis combined with empirical judgment. With the continuous promotion of the construction of human resource intelligent management by the country, artificial intelligence technology has become an important means to improve efficiency and reduce costs in the field of human resource supply - demand relationship management.

[0004] At the same time, the national intelligent demand for human resource supply - demand relationship management brings new challenges to existing methods: (1) The statistical analysis method combined with empirical judgment relies too much on the subjective judgment and experience of staff and cannot give the objective characteristics and operation rules based on the human resource supply - demand state; (2) The non - linear characteristics of the human resource supply - demand relationship are obvious and the states are highly coupled, and a single mechanism model cannot achieve accurate modeling and prediction. The data - driven modeling idea breaks through the bottleneck faced by the mechanism model and avoids the problem of "unclear mechanism" of the object by studying the characteristics contained in the data. On the other hand, the continuous development of artificial neural networks and their self - learning ability provides a practical carrier and methodological basis for the data - driven modeling idea. Therefore, the data - driven artificial neural network modeling method can realize the state approximation and prediction analysis of the target object by learning the characteristics contained in the data, and has become one of the most active research directions in the field of intelligent representation, modeling, and prediction analysis of human resource supply - demand relationship. Summary of the Invention

[0005] Facing the major demand for the intelligent management of enterprise human resources, and aiming at the problems of low accuracy and low efficiency of the existing human resource supply-demand relationship prediction models based on empirical judgment and mechanism models, the present invention proposes an intelligent modeling and prediction method applicable to the management of human resource supply-demand relationships, aiming to provide technical support for the intelligent management of enterprise human resources by studying data-driven neural network modeling, optimization, prediction and other methods, an intelligent modeling and prediction method applicable to the management of human resource supply-demand relationships.

[0006] On the one hand, to achieve the above object, the present invention provides an intelligent modeling and prediction method applicable to the management of human resource supply-demand relationships, including:

[0007] Determine the human resource information to be predicted;

[0008] Construct a self-organizing deep neural network supply-demand relationship prediction model, input the to-be-predicted human resource information into the self-organizing deep neural network supply-demand relationship prediction model for processing, and output the prediction result of the human resource supply-demand matching state;

[0009] Among them, the self-organizing deep neural network supply-demand relationship prediction model is obtained after being trained by a training set, cross-validated by a validation set, and tested by a test set. The training set, the validation set, and the test set are all human resource information feature variables and matching states representing human resource relationships.

[0010] Preferably, obtaining the training set includes:

[0011] Obtain the original data from the human resource management platform, mark the data, and determine the input-output state data representing the human resource demand relationship;

[0012] Among them, the input-output state data representing the human resource demand relationship includes human resource information feature variables and matching state marker feature variables.

[0013] Preferably, constructing the self-organizing deep neural network supply-demand relationship prediction model includes:

[0014] Initialize the deep belief network DBN with a single hidden layer, pre-train the DBN using the human resource information to generate an initial weight parameter matrix;

[0015] Expand the network structure of the DBN through an incremental structure growth strategy, and update the weight parameter matrix based on the transfer learning rule;

[0016] Set the stop criterion for structure growth according to the reconstruction error to generate an incremental deep pre-training model IDPTM;

[0017] The IDPTM is trained and optimized using the training set, and cross-validated in combination with the validation set to obtain the self-organizing deep neural network supply-demand relationship prediction model.

[0018] Preferably, the incremental structure growth strategy includes:

[0019] The number of neurons in the hidden layer of the DBN network is expanded in a multiplicative pattern, and the migration rules of the weight parameters are defined through a mapping function;

[0020] Neurons with the same number as the initial hidden layer in the DBN network are added, and knowledge transfer is achieved through a mapping function.

[0021] Preferably, the stopping criterion is determined by the change in the reconstruction error of the validation set, specifically: when the reduction in the reconstruction error in consecutive p-step iterations is less than a preset threshold, the growth of the network structure stops.

[0022] Preferably, the reconstruction error is:

[0023]

[0024] In the formula, N s and N d are the number and dimension of the training samples respectively, v ij and are the initial input data and the corresponding reconstructed data respectively, and e is the reconstruction error.

[0025] Preferably, the change in the reconstruction error is:

[0026]

[0027] In the formula, p is the number of steps of the stopping criterion, i is the number of iteration steps and i >> p, e is the reconstruction error, is the change in the reconstruction error.

[0028] Preferably, constructing the self-organizing deep neural network supply-demand relationship prediction model further includes:

[0029] The accuracy of the model is evaluated through preset performance indicators, where the preset performance indicators include the mean absolute percentage error MAPE and the coefficient of determination R 2 .

[0030] On the other hand, to achieve the above object, the present invention also provides a computer-readable storage medium, which stores program instructions for executing the intelligent modeling and prediction method applicable to human resource supply-demand relationship management, and the intelligent modeling and prediction function of the human resource supply-demand relationship is realized when the program instructions are executed by a processor.

[0031] Compared with the prior art, the present invention has the following advantages and technical effects:

[0032] (1) In view of the problems of low accuracy and low efficiency of the existing human resource supply - demand relationship prediction models based on empirical judgment and mechanism models, the present invention proposes an intelligent modeling and prediction method applicable to the management of human resource supply - demand relationships. By designing and training a deep neural network model to fit the state of human resource supply - demand relationships, the approximation of the dynamic characteristics of human resource supply - demand relationships is realized;

[0033] (2) In view of the possible over - fitting problem in the training process of the deep neural network, the present invention divides the sample data into a training set, a validation set and a test set, and combines a parameter optimization algorithm and a cross - validation method to improve the stability and generalization of the deep neural network model, and further realizes the accurate prediction of human resource supply - demand relationships. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The drawings constituting a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation of this application. In the drawings:

[0035] Figure 1 Schematic diagram of the human resource supply - demand relationship prediction model based on self - organizing deep neural network according to the embodiment of the present invention;

[0036] Figure 2 Schematic diagram of the one - step growth of the incremental deep pre - training model according to the embodiment of the present invention;

[0037] Figure 3 Schematic diagram of the verification result of the prediction model according to the embodiment of the present invention;

[0038] Figure 4 Schematic diagram of the test result of the prediction model according to the embodiment of the present invention;

[0039] Figure 5 Schematic diagram of the test error of the prediction model according to the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will refer to the drawings and combine the embodiments to detail this application.

[0041] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer - executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0042] An intelligent modeling and prediction method applicable to the management of human resource supply and demand relationships, comprising:

[0043] Determine the human resource information to be predicted;

[0044] Construct a self-organizing deep neural network supply and demand relationship prediction model, input the to-be-predicted human resource information into the self-organizing deep neural network supply and demand relationship prediction model for processing, and output the prediction result of the human resource supply and demand matching status;

[0045] Among them, the self-organizing deep neural network supply and demand relationship prediction model is obtained after being trained by a training set, cross-validated by a validation set, and tested by a test set. The training set, the validation set, and the test set are all human resource information feature variables and matching statuses that characterize human resource relationships.

[0046] This embodiment uses a deep neural network to learn, model, and optimize the actual data of the human resource supply and demand status, and uses a new data set to test the prediction model, thereby realizing the intelligent modeling and prediction of the human resource supply and demand relationship.

[0047] Furthermore, obtaining the training set includes:

[0048] Obtain the original data from the human resource management platform, mark the data, and determine the input-output status data representing the human resource demand relationship;

[0049] Among them, the input-output status data representing the human resource demand relationship includes human resource information feature variables and matching status marking feature variables.

[0050] Specifically, the human resource supply and demand relationship is a multi-factor complex dynamic system, involving dynamic processes such as economic operation, population quantity change, and job-person change occurring interactively, and containing human behaviors and decisions, with characteristics such as strong nonlinearity and strong coupling. Therefore, the data directly obtained from the human resource company needs to be marked and used to train the prediction model. The acquisition of human resource information feature variables, matching status, and their related markings is shown in Table 1 below.

[0051] Table 1

[0052]

[0053] In this embodiment, according to the information feature variables and matching status represented by Table 1, 6000 groups of data (data pairs of information feature variables and matching status) are obtained from a certain human resource company, among which 4000 groups are used as training samples, 1500 groups are used as validation samples, and 500 groups are used as test samples.

[0054] Furthermore, as Figure 1, constructing a self-organizing deep neural network supply-demand relationship prediction model includes:

[0055] Initialize a deep belief network (DBN) with a single hidden layer, and use the input-output state data to pre-train the DBN to generate an initial weight parameter matrix;

[0056] Expand the network structure of the DBN through an incremental structure growth strategy, and update the weight parameter matrix based on the transfer learning rule;

[0057] Set the stop criterion for structure growth according to the reconstruction error to generate an incremental deep pre-training model (IDPTM);

[0058] Use the training set to train and optimize the IDPTM, and perform cross-validation in combination with the validation set to obtain the self-organizing deep neural network supply-demand relationship prediction model.

[0059] Specifically, first initialize a deep belief network (DBN) with a single hidden layer, and then pre-train it. Assume that the number of inputs to the initialized DBN and the number of neurons in its hidden layer are m and n respectively. Then, the knowledge (weight parameter matrix) learned after pre-training using the human resource data samples is saved in the knowledge source domain.

[0060] Furthermore, the incremental structure growth strategy includes:

[0061] Expand the number of neurons in the hidden layer of the DBN network in a multiplicative pattern, and define the transfer rule of the weight parameters through a mapping function;

[0062] Add the same number of neurons as the initial hidden layer in the DBN network, and achieve knowledge transfer through a mapping function.

[0063] Specifically, add neurons. In this embodiment, add neurons that are twice the number of neurons in the initial DBN hidden layer. The new weight parameter matrix becomes

[0064] Define the following mapping f:

[0065] f:{1,2,…,n…,3n}→{1,2,…,n} (1);

[0066] Make it satisfy:

[0067]

[0068] In the formula, j is the serial number of the neuron, x is the serial number of the input vector element, and n is the number of neural networks in the hidden layer.

[0069] Based on equations (1) and (2), a knowledge transfer rule based on transfer learning can be constructed, and the new weight parameter matrix can be expressed as:

[0070]

[0071] Add a hidden layer. In this embodiment, a hidden layer with the same number of neurons as the initial DBN is added, and the corresponding newly added weight parameter matrix is

[0072] Similarly, define the following mapping g:

[0073] g:{1,2,…,n…,3n}→{1,2,…,n} (4);

[0074] such that it satisfies:

[0075]

[0076] Based on equations (4) and (5), a knowledge transfer rule based on transfer learning can be constructed, and the new weight parameter matrix can be expressed as:

[0077]

[0078] where k = 1, 2, …, n. Thus, one-step growth of the deep pre-training model has been achieved, as Figure 2 shown.

[0079] Furthermore, the stopping criterion is determined by the change in the reconstruction error of the validation set. Specifically: when the decrease in the reconstruction error in consecutive p steps of iteration is less than a preset threshold, stop the growth of the network structure.

[0080] Specifically, calculate the pre-training error, taking the reconstruction error as the pre-training error, and define it as follows:

[0081]

[0082] where N s and N d are the number and dimension of training samples respectively, v ij and are the initial input data and the corresponding reconstructed data respectively, and e is the reconstruction error.

[0083] Set the stopping criterion for structure growth, calculate the reconstruction error e v using the validation data, and v take the decrease in e over consecutive p steps as the stopping criterion for structure growth, as shown in formula (8):

[0084]

[0085] In the formula, p is the number of steps of the stopping criterion, i is the number of iteration steps and i >> p, is the change amount of the reconstruction error.

[0086] In this embodiment, when is equal to the threshold of the stopping criterion the incremental DBN structure stops growing and enters the next step; otherwise, continue with the structure growth step. Thus, an incremental deep pre-training model (Incremental deep pre-training model, IDPTM) is constructed to achieve effective feature extraction of the original data.

[0087] Furthermore, constructing the self-organizing deep neural network supply-demand relationship prediction model further includes:

[0088] Evaluating the model accuracy through preset performance metrics, where the preset performance metrics include the mean absolute percentage error MAPE and the coefficient of determination R 2 .

[0089]

[0090] where N is the number of training samples, y(t), and are the target output, the actual output of the model, and the average output respectively.

[0091] This embodiment also provides a computer-readable storage medium, which stores program instructions for executing an intelligent modeling and prediction method applicable to human resource supply-demand relationship management. When the program instructions are executed by a processor, the intelligent modeling and prediction function of the human resource supply-demand relationship is realized.

[0092] This embodiment first trains the designed model, and determines the optimal structure scale of the prediction model as 8 - 30 - 20 - 15 - 6 - 1 through the trial-and-error method, with 4000 groups of training samples, 1500 groups of validation samples, and 500 groups of test samples. Figure 3 is the verification result of the model, Figure 4 is the test result of the model, Figure 5 is the test error schematic diagram. In order to fully demonstrate the performance of the human resource supply-demand relationship prediction model based on the self-organizing deep neural network, 20 independent repeated experiments are carried out and compared with similar methods. The average value of the comparison results is shown in Table 2. From Figures 3 - 5 and Table 2, it can be seen that the human resource supply-demand relationship prediction model based on the self-organizing deep neural network is superior to other models in terms of prediction accuracy, computational complexity, etc.

[0093] Table 2

[0094] Soft sensor model Predicted APE value <![CDATA[Predict R 2 value]]> Convergence time (s) Running time (s) Self - organizing deep neural network 0.9327 0.9265 6.94 16.27 Deep belief network 1.6762 0.7348 14.87 28.35 BP neural network 1.7019 0.7175 12.47 25.63 Dynamic fuzzy neural network 1.2143 0.8826 10.89 24.18 Self - organizing fuzzy neural network based on sensitivity analysis 1.4089 0.8671 8.72 19.46 Radial basis neural network 1.5383 0.8573 13.29 26.16

[0095] The above are only the preferred specific implementation manners of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An intelligent modeling and prediction method applicable to the management of the supply-demand relationship of human resources, characterized in that, Including: Determine the human resource information to be predicted; Construct a self-organizing deep neural network supply-demand relationship prediction model, input the human resource information to be predicted into the self-organizing deep neural network supply-demand relationship prediction model for processing, and output the prediction result of the human resource supply-demand matching status; Among them, the self-organizing deep neural network supply-demand relationship prediction model is obtained after being trained by a training set, cross-validated by a validation set, and tested by a test set. The training set, the validation set, and the test set are all human resource information feature variables and matching statuses that characterize human resource relationships.

2. The intelligent modeling and prediction method applicable to the management of the supply-demand relationship of human resources according to claim 1, wherein Obtaining the training set includes: Obtain the original data from the human resource management platform, mark the data, and determine the input-output status data that characterizes the human resource demand relationship; Among them, the input-output status data that characterizes the human resource demand relationship includes human resource information feature variables and matching status marking feature variables.

3. The intelligent modeling and prediction method applicable to the management of the supply-demand relationship of human resources according to claim 1, characterized in that, Constructing the self-organizing deep neural network supply-demand relationship prediction model includes: Initialize the deep belief network (DBN) with a single hidden layer, and use the human resource information to pre-train the DBN to generate an initial weight parameter matrix; Expand the network structure of the DBN through an incremental structure growth strategy, and update the weight parameter matrix based on the transfer learning rule; Set the stop criterion for structure growth according to the reconstruction error to generate an incremental deep pre-training model (IDPTM); Use the training set to train and optimize the IDPTM, and perform cross-validation in combination with the validation set to obtain the self-organizing deep neural network supply-demand relationship prediction model.

4. The intelligent modeling and prediction method for human resource supply and demand relationship management according to claim 3, characterized in that The incremental structure growth strategy includes: Expand the number of neurons in the hidden layer of the DBN network according to a multiple rule, and define the transfer rule of the weight parameters through a mapping function; Add the same number of neurons as the initial hidden layer in the DBN network, and realize knowledge transfer through a mapping function.

5. The intelligent modeling and prediction method applicable to the management of the supply-demand relationship of human resources according to claim 3, characterized in that, The stop criterion is determined by the change amount of the reconstruction error of the validation set. Specifically, when the reduction amount of the reconstruction error in continuous p-step iterations is less than the preset threshold, stop the growth of the network structure.

6. The intelligent modeling and prediction method applicable to the management of human resource supply and demand relationships according to claim 5, characterized in that, The reconstruction error is: where N s and N d are the number and dimension of the training samples respectively, v ij and are the initial input data and the corresponding reconstructed data respectively, and e is the reconstruction error.

7. The intelligent modeling and prediction method applicable to the management of the supply-demand relationship of human resources according to claim 6, characterized in that The change amount of the reconstruction error is: Where p is the number of steps of the stopping criterion, i is the number of iteration steps and i >> p, e is the reconstruction error, and is the change amount of the reconstruction error.

8. The intelligent modeling and prediction method for human resource supply-demand relationship management according to claim 3, characterized in that Constructing the self-organizing deep neural network supply-demand relationship prediction model further includes: Evaluate the model accuracy through preset performance metrics, where the preset performance metrics include the Mean Absolute Percentage Error (MAPE) and the Coefficient of Determination (R). 2 .

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program instructions for implementing the intelligent modeling and prediction method for human resource supply-demand relationship management according to any one of claims 1-8. When the program instructions are executed by a processor, the intelligent modeling and prediction function of the human resource supply-demand relationship is realized.

Images