Design and interactive performance of human resource management system based on artificial intelligence.

The purpose is to strengthen Human Resources Management (HRM) through information management using Artificial Intelligence (AI) technology. First, the selection criteria of the applicant's resume during recruitment and the formulation standards of the contract salary are analyzed. Then, the resume information is extracted and converted into the data-type format. Besides, the salary forecast model in the HRM system (HRMS) is designed based on the Back Propagation Neural Network (BPNN), and network structure, parameter initialization, and activation function of the BPNN are selected and optimized. The experimental results demonstrate that the algorithm optimized by the Nadm has shown improved convergence speed and forecast effect, with 187 iterations. Moreover, compared with other regression algorithms, the designed algorithm achieves the best test scores. The above results can provide references for designing the AI-based HRMS.

Introduction

With the socio-economic development, talents have become the mainstay of enterprise development, so more enterprises are competing for human resources rather than the traditionally narrower sense of labor resources. Furthermore, with structural informatization, enterprise data generation is increasing exponentially, which can no longer be efficiently handled using the manual Human Resource Management System (HRMS) [1]. Accordingly, the Information Management System (IMS) is employed to collect, store, manage, and analyze the exploding numbers of human resource data to improve the traditional enterprise Human Resource Management (HRM) [2]. In other words, HRM and Information Technology (IT) are combined in an attempt to establish an integrated HRMS to standardize the business process of Human Resource (HR) departments, centralize HR information, and enhance HRM transparency [3]. The quality of the HRMS will determine the enterprise performance and sustainable development. Therefore, it is urgent to establish a high-performance HRMS.

Numerous studies have shown that HRM efficiency can be uplifted by integrating IT. Therefore, this paper utilizes big data technology to manage and analyze HR data by designing a salary forecast model based on Artificial Intelligence (AI). The proposed model forecasts the salary by analyzing job applicants’ resumes from basic information, school, society, and enterprise factors. Then, the relevant parameters of the Back Propagation Neural Network (BPNN) are set and optimized to improve the forecast accuracy of the proposed model, and the model performance is tested through experiment. This paper innovatively integrates the BPNN into HRM to forecast employees’ salaries by analyzing their resumes to provide salary references for the HR department for the applicants and to promote enterprise development.

Literature review

Abdullah et al. (2020) designed an enterprise cloud-based HRMS with 16 standard modules to solve HR problems using the CodeIgniter Web framework, which was then launched and deployed on the Amazon Web Service elastic computing cloud and used for an efficient enterprise HRM [4]. Necula and Strîmbei (2019) developed an architecture to semantically enrich data through data science and semantic web technology for talent training. The experimental results suggested that the classification effect of the proposed architecture was better than the commonly used regression analysis, Random Forest (RF), and Support Vector Machine (SVM), and the proposed architecture could effectively mark the resume data and use the semantic web to extract data information from the resume [5]. Jawad (2020) proposed a website-based HRMS to manage employee activity information, such as salary, registration, and promotion. The HRMS consisted of two parts: website design and database. Experimental results showed that the designed HRMS presented high performance and efficiency in employee information storage and management [6]. Qin et al. (2020) put forward a Recurrent Neural Network (RNN)-based applicant-job matching framework using job applicants’ perception ability, word-level semantic representation, and experience. This method could reduce the dependence on manual labor and improve employability. The information matching degree indicator was used to measure the importance of semantic representation and the contribution of job experience to job requirements [7]. Serje et al. (2018) used the occupational wage data from the International Labour Organization (ILO) to study econometric models of health workers’ income in different countries. They employed the selection model to analyze the skill and income data of health workers. The income of health workers varied in different countries and was negatively correlated with the country’s Gross National Product (GNP). The results could predict the cost of health care intervention and resource needs during sustainable development [8]. He et al. (2016) researched the cross-level relationship between salary differences and individual turnover intention. They investigated and analyzed employees’ annual objective salaries and self-reporting attitudes through the Questionnaire Survey (QS). Results demonstrated that the employee’s turnover intention and salary were positively correlated: the lower the salary was, the stronger the turnover intention was [9]. Shen et al. (2019) explored the organizational self-esteem, supervisory behaviors of communication between leaders, and restrictions on employee feedback behavior. They analyzed the employment behavior of different managers and employees through the hierarchical regression analysis and path analysis strategy. Results suggested that abuse supervision would directly affect employees’ turnover intention [10].

Briefly, the current research on HRMS mostly focuses on employee recruitment, registration, and management. Therefore, the salary forecast model is proposed for the BPNN-based HRMS to predict the salary of employees by analyzing the resume of candidates, thereby providing a check and balance mechanism for applicants and the enterprise interests.

Method

Salary forecast model in the BPNN-based HRM system

Usually, during the enterprise talent recruitment, qualified resumes are first picked out according to applicants’ age, educational background, work experience, and personal skills, which, traditionally, relies on manual selection. Thus, HR expertise knowledge, as well as some industrial common senses, are often required from the relevant personnel [11], where subjective human errors or prejudices might shut some outstanding talents out of the enterprise threshold. Then, the applicants are often informed of their possible salary range at the face-to-face interview stage, which might vary significantly per department and position [12]. Yet, the final salary is mostly determined by the department head after series of possible interviews and discussions, which, again, involves tremendous manual works that might, in turn, lead to deviations. There is an increasing voice calling for the elimination of such non-objective factors under the current competitive market environment.

Meanwhile, under the traditional HRMS, given limited HR personnel and countless applicants, resume selection is more a qualitative and speculative analysis process than an objective and scientific evaluation procedure. AI technology can well lend itself to such a predicament to help enterprises implement a salary forecast model, which can predict applicants’ salary based on their resumes from as early as the resume-selection stage and compares the forecast with their expected salary to provide further references [13]. The model-forecasted salary can be used as a benchmark salary for applicants, based upon which the actual salary can be reasonably adjusted considering the specific departmental and positional standards. An advantage of model-forecasted salaries is objectivity due to massive amounts of data calculation; on the other hand, the salary forecast model substantially reduces the workload of HR personnel and improves overall work efficiency [14].

The first step for salary forecast is the preprocessing of applicants’ resumes, after which the extracted structured information is used to train the salary forecast model against various data formats. The resume data can be classified as essential information and supplementary information, as shown in Fig 1, in which the resume is illustrated through hierarchically structured content. Common methods to extract information reasonably include the rule-based extraction, the Cascaded Hybrid Model, and the Conditional Random Field (CRF) extraction method.

Fig 1

Information level of resumes.

Employee salary determination is an intricate business that involves employee job specialty, as well as some subjective and objective factors, such as the nature of the enterprise. So is the salary forecast process that depends on the completeness of the resume information and the model performance to extract data features accurately and efficiently. The resume features division based on its content and formats read:

Personal factors: name, age, gender, phone number, and other information, divided according to the degree of association with salary forecast. For example, residence and birthplace will affect employment tenure; name, phone number, and email address are unique identifiers of the applicants; there may be special requirements on age and gender for specific positions [15].

School-related factors including school-time, alma mater, educational background, specialty, and awards. The alma mater and specialty can reflect an applicant’s learning ability and expertise skills, which also affect the salary forecast; awards are the manifestation of the applicant’s school performance and learning attitudes.

Social factors including served companies, positions, work hours, and length of service. Served companies and positions can reflect the level of personal abilities and skills. Length of service can reflect the mastery of relevant skills and personal adaptability. The number of served companies also affects the result of the salary forecast [16].

Enterprise factors are more closely related to the target enterprises. They are not included in the resumes, including the applying positions and occupation levels. Different enterprises in the same industry have different salary plans. The position applied to is the core of the salary forecast, which has a significant impact on the salary level. The occupational level is the requirement of salary level and the key to salary forecast [17].

Therefore, the salary demands of applicants should be forecasted from multiple aspects. Afterward, the character data extracted from the resume are converted into numerical data for subsequent data preprocessing and model training.

The salary forecast is a regression analysis process. The Neural Network (NN) in Machine Learning (ML) can be used as a salary regression forecast model, which is a parallel interconnected network composed of adaptive neural units and simulates the interaction process between the biological nervous system with the outside world. BPNN is the most common NN model. Theoretically, a 3-layer BPNN can approach a continuous function of arbitrary precision with a definite learning ability [18]. Yet, the network structure, parameter settings, and optimization algorithms will affect the NN training results and the salary model’s forecast effect. The learning process of the NN is the adjustment process of neuron parameters. BPNN’s error backpropagation process can adjust parameters. BPNN adopts a Gradient Descent (GD) strategy, which uses the target’s negative gradient direction as the search direction. During continuous iterations, the network parameters are updated, the model converges, and finally, the network parameters are adjusted [19].

BPNN contains a multi-layer signal feedforward NN, and the learning process is divided into two processes: forward propagation and directional propagation. The sample data enters the network from the input layer during the forward propagation, processed by the hidden layer and output from the output layer [20]. The error backward propagation process starts from the output layer, returns the results by layers in the reverse direction and the connection of neurons, and adjusts the neuron parameters in the path [21]. The processes of BPNN’s forward and backward propagation are illustrated in Fig 2.

Fig 2

The forward propagation and backward propagation processes in BPNN.

In the signal forward propagation process of the 3-layer BPNN, x represents the output of the input layer neuron, l denotes the output of the hidden layer neuron, and z indicates the output of the output layer neuron; α refers to the threshold of the hidden layer, and β stands for the threshold of the output layer; w indicates the weight of the input layer to the hidden layer, and v denotes the weight of the hidden layer to the output layer. The sample data are not processed in the input layer, and the output l of the j-th node in the hidden layer can be expressed as Eq (1).

The output z of the k-th node in the output layer reads:

The output z of the output layer is the calculation of forward propagation. However, a gap exists between the generated forecast value and the actual value. The loss function E can serve as a standard to measure the error. The expression of loss function E reads:

In Eq (3), θ = θ(α, β, w, v) represents all the network parameters, and z = (z1,z2,…z) denotes the output vector [22]. BPNN uses a GD method to minimize the loss function in the continuous iteration process [23]. The calculation process is described as follows.

First, the connection weight gradient of the output layer is calculated according to:

The output error δ, weight gradient , and threshold gradient of the output node are calculated as in Eqs (6)–(8), respectively:

The connection weight gradient of the hidden layer can be presented as:

The calculations of the output error γ, weight gradient , and threshold gradient of the hidden layer read:

Meanwhile, the learning rate η of each network layer can be adjusted to optimize the update step size and convergence speed of the iterative algorithm [24]. In BPNN, parameters are adjusted according to the negative gradient direction of the target. Assuming that the learning rate of each network layer is the same, then:

In Eqs (19) and (20), f() represents the activation function, and c() refers to the error function. Therefore, the gradient of the hidden layer parameters and the output layer parameters can be calculated by layers. This process is called backpropagation [25].

Network structure. Before BPNN can approach non-linear functions with arbitrary precision through its non-linear mapping, generalization, and fault tolerance capabilities, the network structure, activation function, and parameter initialization must be solved [26]. Since the salary forecast model processes multiple inputs while producing only one output, the number of input neurons must be determined. Based on the experimental analysis, the number of input and output neurons is set to 14 and 1, respectively. Since the hidden layer does not have a definite calculation equation, the hidden layer number is estimated first then determined through trial and error. The number of hidden layer neurons n can be determined according to:

In Eqs (21) and (22), n represents the number of nodes in the input layer, and n denotes the number of nodes in the output layer [27].

Activation function. The activation function contains linear or non-linear functions used to solve linear problems and uncertain problems, respectively. The commonly used activation function Sigmoid function is expressed as Eq (23).

The output result of this function is between (0, 1), and the function has differentiability and saturated nonlinearity, which can enhance the non-linear mapping ability of the network [28].

Parameter initialization. Parameter initialization will affect the training results and the convergence degree of the model. If the parameter setting is unreasonable, the model will fall near the local minimum during training and fails to converge. In the initial situation, the initial connection weight is accumulated, and the state of each neuron is 0. Neurons in the same layer should not be assigned the same weight; otherwise, the calculation result will be the same output [29]. Therefore, it is necessary to ensure that the connection weight is a random decimal number, and the value difference between each other is small. Hence, a good convergence speed will be obtained. This paper selects a Gaussian random number with a mean value of 0, and a standard deviation of (n represents the number of neuron input connection weights), as well as a Gaussian random number with a mean value of 0, and a standard deviation of 1 as the weight and threshold parameters, respectively [30]. During algorithm training, the weight of the corresponding NN is reduced to minimize the recruitment discrimination caused by gender, age, and other factors.

Loss function. It can measure the error between the NN forecasted value and the actual value, which is the important criterion for training the learning model. Here, the objective function is calculated by quadratic Mean Square Error (MSE), as shown in Eq (24).

In Eq (24), y represents the objective function, and z denotes the output vector [31].

Optimization methods. BPNN has a simple structure and strong learning ability, but its complexity and performance are affected by the parameter initialization and network structure. Besides, given a flat error surface, oscillation will occur and affect the convergence speed. During the GD, the model easily falls into a local minimum and affects its forecast result. In the actual application, the convergence speed of BPNN needs to be optimized to avoid falling into a local minimum [32]. Common optimization methods include:

In BPNN, the GD algorithm is applied to update the parameters in the backward propagation. According to the data bulk, the GD algorithm can be divided into Batch Gradient Descent (BGD) [33], Stochastic Gradient Descent (SGD) [34], and Mini-batch Gradient Descent (MBGD) [35].

The last parameter change and the current parameter increment are connected by adding the former to the latter. This method of adding momentum to the parameter increment is called the additional momentum method, including the Momentum method and Nesterov Accelerated Gradient (NAG) method [36].

The adaptive learning rate method dynamically adjusts the learning rate during the model training process to speed up convergence and reduce oscillations. Common adaptive learning rate optimization methods are Adagrad and RMSprop [37].

Combining the additional momentum method and the adaptive learning rate method can obtain a hybrid optimization method that can optimize the GD. Such hybrid method has a fast convergence speed and fewer oscillations, including Adaptive Moment Estimation (Adam) and Nesterov-accelerated Adaptive Moment Estimation (Nadam) [38].

Determining parameters of the salary forecast model

In summary, the ratio of the number of neurons in the input layer, the hidden layer, and the output layer is 14:15:1 in the designed 3-layer BPNN. The activation function is a Sigmoid function, while the error function is a quadratic MSE function. MBGD and Nadam optimization algorithms are used to optimize the model, and the BPNN parameters and experimental environment are displayed in Tables 1 and 2. Therefore, the output errors δ and γ of the output layer and the hidden layer are calculated according to:

Table 1

BPNN parameter setting.

Parameter	Value	Parameter	Value
Learning rate	0.1	Number of network layers	4
Maximum number of training	10	Number of input layer neurons	14
Accuracy required by the training	0	Number of hidden layer neurons	15
Minimum gradient requirement	1.00E-10	Number of output layer neurons	1
Training iteration process displayed	25	GD algorithm	MBGD algorithm
Maximum training time	inf		Nadam algorithm
Momentum factor	0.9	Loss function	Quadratic mean square function
		Activation function	Sigmoid function

Table 2

Experimental hardware and software environment.

Hardware environment	CPU	Intel Xeon E5-2640 v4 @2.40GHz	Software environment	Operating system	Ubuntu 14.04
	Graphics Processing Unit (GPU)	NVIDIA TITANX (pascal)		Software	Python 2.7
	RAM +Internal memory	16GB+256GB

In Eqs (25) and (26), y and z represent the target value and output value of the k-th node. The salary forecast model’s training process based on BPNN is shown in Fig 3 [39].

Fig 3

The training process for the salary forecast model.

Introduction to experimental samples

Because salary forecast belongs to a multiple-input-single-output mapping process, effective salary forecast results can be obtained by considering different influencing factors of salary like the model’s output and optimizing the influence weight of each factor in the model through training. The information in the resume library of an enterprise is analyzed to verify the effect of the designed model. The number of samples is 1,000, of which the number of recruits is 298, the number of rejections is 702, and each resume has a corresponding salary status. Data characteristics of the resume include age, salary, gender, the highest academic credential, major, marital status, job position, position applied, work experience, and length of service. Since the salary forecast model structure is affected by the training samples and model parameters, the most suitable parameter settings are obtained by continuously adjusting the model parameters. Recruitment bias caused by factors such as gender and age may arise during employee recruitment. However, in the present work, the resume data of the enterprise after successful recruitment is used so that the work does not directly consider the possible moral hazard. Among the risks, the generated salary forecast model also meets the enterprise’s recruitment requirements. Therefore, the research results do not consider the possible moral hazard. To better compare the classification effect of the classifier, the following indicators are employed to evaluate the classification effect of the classifier. Among them, TP refers to the proportion of resumes correctly classified as entry, FP denotes the proportion of resumes incorrectly classified as entry, TN indicates the proportion of resumes correctly classified as non-entry, and FN represents the proportion of resumes incorrectly classified as the entry.

Precision refers to the total proportion of positive samples correctly classified by the classifier.

Recall in Eq (28) indicates the proportion of positive samples that are correctly predicted.

Results and discussion

Parameter selection

According to the empirical equation, the number of neurons is in the range of [5, 14]. Hence, the model with several neurons between [3, 16] is trained and tested. The number of samples used is 1,000, with 500 in the training set and 500 in the validation set. The hidden layer neurons are trained 30 times to obtain the average result. The number of iterations is set at 1,000, with the training loss accuracy E<0.01. The training results are displayed in Fig 4.

Fig 4

Comparison of training results of different numbers of neurons in the hidden layer.

According to Fig 4, when the number of neurons in the hidden layer is 14, the model’s training loss is the smallest, which is 0.009914. When the number of neurons is 15, the number of iterations and the model’s verification loss are the smallest, of 64 and 0.010511, respectively. Therefore, the number of nodes in the input layer, the hidden layer, and the output layer of the designed BPNN is 14, 15, and 1, respectively. Such a structure can accelerate the model’s running speed and reduce the verification loss and training loss.

The effect of the optimization algorithm on the model is verified through the Leave-One-Out (LOO) method. The sample-set with 1,000 samples is divided into a training set with 900 samples and a test set with 100 samples. The number of cycles is set to 10,000, or the training is stopped when the training loss accuracy E is less than 0.005. The convergence speed of the model optimized by different algorithms is shown in Fig 5; the comparison of training results is displayed in Fig 6.

Fig 5

Comparison of the convergence speed of different optimization methods.

Fig 6

Comparison of training results of different optimization methods.

As shown in Fig 5, among various optimization algorithms, Adam and Nadm have faster convergence speeds than other optimization algorithms, which can quickly converge to the minimum and reduce the model’s running time. Furthermore, the convergence effect of Nadm is better than that of Adam. According to the training results comparison of six algorithms, the test scores and training losses of various optimization algorithms are close. However, there are noticeable differences in the number of training cycles. SGD has the largest number of training cycles, of 3,228 times. In contrast, Nadm has the minimal training cycles of 187 times. Therefore, among the various NN GD algorithms, the hybrid optimization algorithm Nadm presents the best optimization effect and convergence speed. Hence, Nadm is chosen as the GD optimization algorithm for the salary forecast model.

Analysis of salary relevance

It has been argued that salary level might somehow reflect the relevance between talent supply and recruitment positions. This section selects 10,000 pieces of enterprise recruitment data, including the lowest salary, highest salary, and average salary, location, company size, financing, education level, work experience, and job type. First, the relevance among different influencing factors is analyzed using the Pearson correlation coefficient method, and the results are shown in Fig 7. Fig 7 reveals that the relevance between salary is the strongest, and the relevance among education level, job type, and work experience and salary is also stronger than other influencing factors. Therefore, salary is taken as the dependent variable with the influencing factors including job type, work experience, and education level as the independent variable, and the analysis results are shown in Fig 8.

Fig 7

Analysis of salary relevance.

Fig 8

The relationship between dependent variables and salary.

a) Job type; b) Work experience; c) Education level.

Fig 8A suggests that there is a certain relationship between job type and salary. The salary of different jobs varies much. The salary range of data risk jobs and products is 15-25k/month. The development of the Internet in recent years has made the income of related industries continue to rise. Operation positions and design positions are affected by personnel, and the salary will also change. Fig 8B indicates that the salary will continue to improve with the accumulation of work experience. The salary will not change much within three years, but will increase after three to five years, and will increase significantly after five to ten years. The maximum monthly salary can reach more than 40 k/month. Hence, after several years of work experience, employees can get a higher salary. Fig 8C shows that with the improvement of education level, the salary also rises. The salary of undergraduates is 50% higher than that of college students. The salary of masters and doctors is much higher than that of other education levels, and the salary ceiling of doctors is higher. To sum up, job type, work experience, and education level are the main factors that affect salary, and the different influencing factors will also affect salary. But in general, the higher the education level is, the richer the work experience is, and the more likely the employees in popular positions are to get a high salary.

Performance test

The resume information status of the experimental sample is summarized in Fig 9. The salary forecast model based on BPNN is applied to fit the salary data to verify its actual forecast effect, and the results are shown in Fig 10. Common regression algorithms are adopted for comparative simulations, including Linear, Polynomial, Ridge, Lasso, and ElasticNet. The ratio of the correct number of samples output by different models to all samples is taken as the model’s score. The scores of the training process and testing process are presented in Fig 11. The number of samples in the training set used for model training is 900, and the number of samples in the test set is 100. Six experiments are carried out on each model and the average value is taken as the final result.

Fig 9

Resume information distribution.

a) Applicant department status; b) Resume information status.

Fig 10

The fitting effect of Nadm-optimized salary forecast model.

Fig 11

Comparison of salary forecast results of different algorithms.

Fig 9 presents that the recruitment data involves different job types, and the data content includes salary level, gender, education level, and other information. According to Fig 10, the BPNN-based salary forecast model optimized by the Nadm gradient has an excellent fitting performance. However, some errors still exist. The reason may be that the data are normalized during the model training process, resulting in errors in the calculation results. However, the overall error is acceptable. As shown in Fig 11, the Nadm gradient optimization model has the highest test score than other algorithms.

Performance comparison

BPNN is employed to analyze the establishment information to obtain the forecasted salary and compare the salary level of the sample database from precision and recall rate. The forecast results of the LSTM network, GRU network, Bi-LSTM network, and Bi-GRU network are compared, as illustrated in Fig 12.

Fig 12

Comparison of salary forecast results of different NN algorithms.

As shown in Fig 12, the precision and recall rate of BPNN optimized by the GD algorithm for salary forecast is above 0.9, which is better than the forecast results of the LSTM network, GRU network, Bi-LSTM network, and Bi-GRU network. Hence, the actual effect of the designed salary forecast model can provide better accuracy than similar models. Thus, during the actual application, a desirable salary can be provided to applicants by analyzing their resumes. The results can provide a theoretical reference for establishing a reasonable employee salary system.

In summary, the designed salary forecast algorithm based on the BPNN model uses a network structure with 14:15:1 (input layer: hidden layer: output layer) of the number of neurons. Besides, the Nadm gradient optimization algorithm is used to optimize the model to get a faster convergence speed and excellent forecast effect. Compared with other regression algorithms, the algorithm used here has the best forecast effect and test score. The actual forecast results also show that the salary forecast model can provide desirable salaries for applicants. Therefore, the designed salary forecast algorithm can apply to forecast the salary in the HRM system.

Conclusion

This work aims to build an intelligent HRMS and improve the efficiency of HRM. A Human-Computer Interaction (HCI)-based HRMS is designed using AI technology to strengthen enterprises’ management and development capabilities. First, J2EE is employed to design a modularized HRMS. Second, the Artificial Neural Network (ANN) is adopted to optimize the HRMS, and the salary forecast module is implemented to effectively judge the applicants’ ability according to the resume and industry information, thereby offering a reasonable salary. Eventually, the performance of the designed salary prediction model of the HRMS is tested and analyzed. The experimental results demonstrate that the network structure, parameter settings, and gradient optimization algorithm will affect the model’s forecast structure. Tests prove that the Nadm gradient optimization algorithm can effectively improve the model’s convergence speed and actual fitting effect. Compared with other algorithms, the model optimized by Nadm has optimal test scores. Therefore, the proposed algorithm can be applied to the salary forecast of the HRM system. However, there are still some shortcomings. The salary of applicants might be affected by many factors in real life, but this paper conducts the correlation analysis only on several main influencing factors, so the proposed salary forecast model is relatively simple, with a weak data feature processing ability. In the research process, the model has been optimized to further reduce the impact of potential recruitment discrimination on the output of the model. In the follow-up research, it is worth considering more possible influencing factors to enhance the model forecast ability and accuracy.

(ZIP)

Click here for additional data file.

19 Jan 2021

PONE-D-21-00932

Designing a Human Resource Management System and Analyzing Its Human-Computer Interaction Performance Under the Background of Artificial Intelligence

PLOS ONE

Dear Dr. Zhao,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Before review process, some comments should be addressed for manuscript revision.

Therefore, I invite the authors to resubmit the revised manuscript for further reviews.

1. The contributions of this study should be highlight in the first section.

2. A literature review section should be given and discussed.

3. The authors only used a neural network to solve the research questions. However, the authors should propose an original model or improve the neural network to solve the research questions.

4. The authors should present the structure of the used neural networks.

5. The authors should present the loss function of the used neural networks.

6. The authors should present the activation functions of the used neural networks.

7. The authors should compare neural networks with LSTM networks, GRU networks, Bi-LSTM networks, and Bi-GRU networks.

8. The authors should give practical experimental results to compare the used method with other methods.

9. The authors should present the limitation of this study.

10. Conclusions and future work should be given and discuss and given in the last section.

Please submit your revised manuscript by Mar 05 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Chi-Hua Chen, Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments:

Before review process, some comments should be addressed for manuscript revision.

Therefore, I invite the authors to resubmit the revised manuscript for further reviews.

1. The contributions of this study should be highlight in the first section.

2. A literature review section should be given and discussed.

3. The authors only used a neural network to solve the research questions. However, the authors should propose an original model or improve the neural network to solve the research questions.

4. The authors should present the structure of the used neural networks.

5. The authors should present the loss function of the used neural networks.

6. The authors should present the activation functions of the used neural networks.

7. The authors should compare neural networks with LSTM networks, GRU networks, Bi-LSTM networks, and Bi-GRU networks.

8. The authors should give practical experimental results to compare the used method with other methods.

9. The authors should present the limitation of this study.

10. Conclusions and future work should be given and discuss and given in the last section.

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We suggest you thoroughly copyedit your manuscript for language usage, spelling, and grammar. If you do not know anyone who can help you do this, you may wish to consider employing a professional scientific editing service.

Whilst you may use any professional scientific editing service of your choice, PLOS has partnered with both American Journal Experts (AJE) and Editage to provide discounted services to PLOS authors. Both organizations have experience helping authors meet PLOS guidelines and can provide language editing, translation, manuscript formatting, and figure formatting to ensure your manuscript meets our submission guidelines. To take advantage of our partnership with AJE, visit the AJE website (http://learn.aje.com/plos/) for a 15% discount off AJE services. To take advantage of our partnership with Editage, visit the Editage website (www.editage.com) and enter referral code PLOSEDIT for a 15% discount off Editage services.  If the PLOS editorial team finds any language issues in text that either AJE or Editage has edited, the service provider will re-edit the text for free.

Upon resubmission, please provide the following:

The name of the colleague or the details of the professional service that edited your manuscript

A copy of your manuscript showing your changes by either highlighting them or using track changes (uploaded as a *supporting information* file)

A clean copy of the edited manuscript (uploaded as the new *manuscript* file)

3. PLOS requires an ORCID iD for the corresponding author in Editorial Manager on papers submitted after December 6th, 2016. Please ensure that you have an ORCID iD and that it is validated in Editorial Manager. To do this, go to ‘Update my Information’ (in the upper left-hand corner of the main menu), and click on the Fetch/Validate link next to the ORCID field. This will take you to the ORCID site and allow you to create a new iD or authenticate a pre-existing iD in Editorial Manager. Please see the following video for instructions on linking an ORCID iD to your Editorial Manager account: https://www.youtube.com/watch?v=_xcclfuvtxQ

4. Thank you for stating the following in the Competing Interests section:

"The authors have declared that no competing interests exist"

We note that one or more of the authors are employed by a commercial company: Jiangsu branch of China Mobile Group.

4.1. Please provide an amended Funding Statement declaring this commercial affiliation, as well as a statement regarding the Role of Funders in your study. If the funding organization did not play a role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript and only provided financial support in the form of authors' salaries and/or research materials, please review your statements relating to the author contributions, and ensure you have specifically and accurately indicated the role(s) that these authors had in your study. You can update author roles in the Author Contributions section of the online submission form.

Please also include the following statement within your amended Funding Statement.

“The funder provided support in the form of salaries for authors [insert relevant initials], but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the ‘author contributions’ section.”

If your commercial affiliation did play a role in your study, please state and explain this role within your updated Funding Statement.

4.2. Please also provide an updated Competing Interests Statement declaring this commercial affiliation along with any other relevant declarations relating to employment, consultancy, patents, products in development, or marketed products, etc.

Within your Competing Interests Statement, please confirm that this commercial affiliation does not alter your adherence to all PLOS ONE policies on sharing data and materials by including the following statement: "This does not alter our adherence to  PLOS ONE policies on sharing data and materials.” (as detailed online in our guide for authors http://journals.plos.org/plosone/s/competing-interests) . If this adherence statement is not accurate and  there are restrictions on sharing of data and/or materials, please state these. Please note that we cannot proceed with consideration of your article until this information has been declared.

Please include both an updated Funding Statement and Competing Interests Statement in your cover letter. We will change the online submission form on your behalf.

Please know it is PLOS ONE policy for corresponding authors to declare, on behalf of all authors, all potential competing interests for the purposes of transparency. PLOS defines a competing interest as anything that interferes with, or could reasonably be perceived as interfering with, the full and objective presentation, peer review, editorial decision-making, or publication of research or non-research articles submitted to one of the journals. Competing interests can be financial or non-financial, professional, or personal. Competing interests can arise in relationship to an organization or another person. Please follow this link to our website for more details on competing interests: http://journals.plos.org/plosone/s/competing-interests

5. Please upload a new copy of Figure 3 as the detail is not clear. Please follow the link for more information: https://blogs.plos.org/plos/2019/06/looking-good-tips-for-creating-your-plos-figures-graphics/

6. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

[Note: HTML markup is below. Please do not edit.]

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

18 Mar 2021

Additional Editor Comments:

Before review process, some comments should be addressed for manuscript revision.

Therefore, I invite the authors to resubmit the revised manuscript for further reviews.

1. The contributions of this study should be highlight in the first section.

Response: Thanks for reviewing this manuscript. The innovative points have been supplemented in the introduction section.

2. A literature review section should be given and discussed.

Response: Thanks for your comment. A literature review has been given to summarize the previous research works.

3. The authors only used a neural network to solve the research questions. However, the authors should propose an original model or improve the neural network to solve the research questions.

Response: Thanks for reviewing this manuscript. In Section 2.2, the defects of BPNN are introduced. Therefore, in “(5) Optimization methods,” the gradient optimization algorithm is proposed to optimize the BPNN to improve its convergence speed; besides, the gradient descent optimization algorithm is introduced.

4. The authors should present the structure of the used neural networks.

5. The authors should present the loss function of the used neural networks.

6. The authors should present the activation functions of the used neural networks.

Response: Thanks for commenting. Neural network parameters have been supplemented in Section 2.2.

7. The authors should compare neural networks with LSTM networks, GRU networks, Bi-LSTM networks, and Bi-GRU networks.

Response: Thanks for your suggestion. Comparison results of actual research works have been supplemented.

8. The authors should give practical experimental results to compare the used method with other methods.

Response: Thanks for your suggestion. Comparison results of actual research works have been supplemented.

9. The authors should present the limitation of this study.

Response: Thanks for your comment. In the conclusion section, the limitations of the manuscript are introduced.

10. Conclusions and future work should be given and discuss and given in the last section.

Response: Thanks for commenting. In the conclusion section, the research content is summarized, and the deficiencies in the manuscript at the current stage and the prospects for future works are proposed.

Submitted filename: Response.docx

Click here for additional data file.

26 Apr 2021

PONE-D-21-00932R1

Designing a Human Resource Management System and Analyzing Its Human-Computer Interaction Performance Under the Background of Artificial Intelligence

PLOS ONE

Dear Dr. Gong,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Jun 10 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see:  http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at  https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols .

We look forward to receiving your revised manuscript.

Kind regards,

Chi-Hua Chen, Ph.D.

Academic Editor

PLOS ONE

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: I Don't Know

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: (I am reviewing this paper from the point of view of Machine Learning / Natural Language Processing - I had not reviewed the previous version of the paper)

I appreciate the authors have addressed the previous comments by the reviewers. However there is one fundamental problem and a couple of related problems with this paper that need to be addressed in my opinion before it can be published.

As the authors note, the resumes that they propose need to be input to a salary forecasting algorithm, present complex, structured information (line 75-124, fig 1). Extracting this information is not trivial. The authors only mention some methods that are "commonly used" (line 97), and then, (line 123) "since the info extracted from resumes is character data, it needs to be transformed to numeric data". This makes it sound like a resume is simply processed as a string of characters. The results they obtain are not interpretable in any way then - ie, we have no idea which features of the resume may actually affect the salary, which from fig. 7 appears underestimated; in fact in the end it may only depend on the type of position applied. We are not told what the performance of a baseline that simply predicts based on type of positions would be.

This brings up questions of ethical responsibility as well: how do we know which features may affect this computation? what if it is correlated with gender, or age, say? do we run the risk of a forecast model that is inherently biased? the authors don't discuss any ethical implications of their work.

Also, it is not clear what P/R in Fig 9 and scores in fig 8 are about. P/R are in general applied to classification, when we can say what are TPs (true positives), etc. what are TP/FP etc here? I can't imagine that it's computed on exact salary forecast. Rather, is it computed on the hiring/not hiring classification (line 319)?

I thought I'd find the answer in the data, that the authors say is available. But what is available is data on running on their experiments, not the original dataset of the 1000 resumes; or the kind of information that may have been extracted, like education, previous work experience etc. That data may not be available since it comes from a company, but the authors should be clear about that. As a minimum, they should add information about what is contained in those resumes: which sort of positions were applied to; what sort of educational experience; gender distributions; age ranges etc.

The other problem with this paper is that it is not clear what the contributions of the authors are. Lines 136-240 (approximately) discuss how BPNN's work and can be trained; but this is not the authors' contribution, ie, they did not invented BPNN's or how to train them. I am not sure describing how BPNN's have been applied to the salary forecasting problem (without any description of the data and the features that affect the model, as noted above), warrants a journal publication.

The literature review (which has been added) is still only half a page, and only mentions papers from 2019/2020. I assume trying to forecast salary existed before 2019?

line 346 "Finally, the human-computer interaction performance ...": there is no human computer interaction whatsoever in this paper; there is no design of an interface, there's no evaluation of how users would use these results.

English: the authors note a professional has revised the paper but there are still few infelicities:

Abstract: "The purposes are " --> " The purpose of this paper is ..."

line 84 "on spectacular judgments" --> "on speculative judgments" (Spectacular is for sure the wrong adjective: I presume the authors mean speculative)

line 119 " Applying position": "the position applied to"

line 331 "analyzing the applicant’s resume, which has some reference value": what is "the reference value"?

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

1 Jun 2021

Reviewer #1: (I am reviewing this paper from the point of view of Machine Learning / Natural Language Processing - I had not reviewed the previous version of the paper)

I appreciate the authors have addressed the previous comments by the reviewers. However there is one fundamental problem and a couple of related problems with this paper that need to be addressed in my opinion before it can be published.

Response: Thanks for reviewing this manuscript. We have revised this manuscript based on the comments of yours.

As the authors note, the resumes that they propose need to be input to a salary forecasting algorithm, present complex, structured information (line 75-124, fig 1). Extracting this information is not trivial. The authors only mention some methods that are "commonly used" (line 97), and then, (line 123) "since the info extracted from resumes is character data, it needs to be transformed to numeric data". This makes it sound like a resume is simply processed as a string of characters. The results they obtain are not interpretable in any way then - ie, we have no idea which features of the resume may actually affect the salary, which from fig. 7 appears underestimated; in fact in the end it may only depend on the type of position applied. We are not told what the performance of a baseline that simply predicts based on type of positions would be.

Response: Resumes received during recruitment are not in a fixed format, so that the resume information shall be processed and converted into character data that can be processed uniformly. In lines 114-129, the resume information that affects the salary level of employees is introduced in detail. Descriptions of Figure 7 suggest that the model error may be caused by normalizing the resume data. Nonetheless, the overall prediction trend is in line with the actual results so that the obtained results are considered to meet the actual requirements. During training, the dataset after successful recruitment is used; thus, the internal parameters of the model have been considered for the salary level caused by the employment position during training. The salary of employees depends on a variety of factors. Hence, we do not study salary changes caused by a single factor such as job position.

This brings up questions of ethical responsibility as well: how do we know which features may affect this computation? what if it is correlated with gender, or age, say? do we run the risk of a forecast model that is inherently biased? the authors don't discuss any ethical implications of their work.

Response: In the “Introduction to Experimental Samples” section, the data characteristics and current moral hazards of the experimental samples are introduced. In lines 114-129, factors that affect the salary of employees are introduced. The information in these resumes will impact the results of salary predictions. Besides, the setting of the BPNN parameters used will also affect the prediction results. Since the training of the prediction model uses the data after successful recruitment, the moral hazard caused by gender and age is not directly involved in the research process, and the prediction results obtained by the model are also in line with the enterprise’s recruitment requirements. Therefore, the risk of recruitment bias that may exist is not considered in the research process.

Also, it is not clear what P/R in Fig 9 and scores in fig 8 are about. P/R are in general applied to classification, when we can say what are TPs (true positives), etc. what are TP/FP etc here? I can't imagine that it's computed on exact salary forecast. Rather, is it computed on the hiring/not hiring classification (line 319)?

Response: In the “Introduction to Experimental Samples” section, the introduction to experimental samples and test indicators have been introduced.

I thought I'd find the answer in the data, that the authors say is available. But what is available is data on running on their experiments, not the original dataset of the 1000 resumes; or the kind of information that may have been extracted, like education, previous work experience etc. That data may not be available since it comes from a company, but the authors should be clear about that. As a minimum, they should add information about what is contained in those resumes: which sort of positions were applied to; what sort of educational experience; gender distributions; age ranges etc.

Response: In the “Introduction to Experimental Samples” section, an introduction to experimental data has been supplemented to facilitate readers’ understanding of the dataset.

The other problem with this paper is that it is not clear what the contributions of the authors are. Lines 136-240 (approximately) discuss how BPNN's work and can be trained; but this is not the authors' contribution, ie, they did not invented BPNN's or how to train them. I am not sure describing how BPNN's have been applied to the salary forecasting problem (without any description of the data and the features that affect the model, as noted above), warrants a journal publication.

Response: The prediction effect of the model is affected by the training samples and model parameters. To get a more accurate prediction effect, it is necessary to continuously test to get the best parameter settings of the model. Therefore, it is considered that the process of solving the optimal parameters also belongs to the research contribution of the manuscript. Since salary prediction is a multiple-input single-output mapping process, effective salary prediction results can be obtained by taking different salary influencing factors as the output of the model and optimizing the influence weight of each factor in the model through training.

The literature review (which has been added) is still only half a page, and only mentions papers from 2019/2020. I assume trying to forecast salary existed before 2019?

Response: Research content related to employee salary prediction before 2019 has been added.

line 346 "Finally, the human-computer interaction performance ...": there is no human computer interaction whatsoever in this paper; there is no design of an interface, there's no evaluation of how users would use these results.

Response: This is a good suggestion. Revisions have been made correspondingly.

English: the authors note a professional has revised the paper but there are still few infelicities:

Abstract: "The purposes are " --> " The purpose of this paper is ..."

line 84 "on spectacular judgments" --> "on speculative judgments" (Spectacular is for sure the wrong adjective: I presume the authors mean speculative)

line 119 " Applying position": "the position applied to"

line 331 "analyzing the applicant’s resume, which has some reference value": what is "the reference value"?

Response: Thanks for your comments. This manuscript has been revised.

Submitted filename: comments.docx

Click here for additional data file.

21 Jun 2021

PONE-D-21-00932R2

Designing a Human Resource Management System and Analyzing Its Human-Computer Interaction Performance Under the Background of Artificial Intelligence

PLOS ONE

Dear Dr. Gong,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Aug 05 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see:  http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at  https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols .

We look forward to receiving your revised manuscript.

Kind regards,

Chi-Hua Chen, Ph.D.

Academic Editor

PLOS ONE

Journal Requirements:

Additional Editor Comments (if provided):

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: I appreciate the authors have attempted to address my comments, added to the literature, added a short section "introduction to the sample" which addresses some of my questions on the data and the evaluation metrics; and eliminated the reference to human computer interaction. However, my major concerns remain unaddressed.

Specifically:

1. They have not addressed the issues I was raising as concerns which attributes affect salary predictions, the baseline, and importantly, whether the model is biased.

a. I had asked for a baseline, as is common in ML experiments. None was provided, and the justification was "The salary of employees depends on a variety of factors. Hence, we do not study salary changes caused by a single factor such as job position. " Sure, but if there is no baseline there is the possibility that a much simpler model could be as if not more effective, and explanatory. I used "job position" as a simple example that intuitively makes sense, but many other baselines would be possible. I don't consider the GRU results a baseline in this sense.

b. I had asked for some elaboration on which features may affect the model, but no insight was provided. The authors themselves say "The salary of employees depends on a variety of factors. " but we have no idea which factors matter after these experiments. The authors could have run ablation studies, although not sure their encoding of resumes allows for that. But if that's the case, they should find a way of addressing this point.

c. If I understand the authors' argument re bias, they basically say "this is a model about previous recruitment, so there's no moral hazard", to wit "Since the training of the prediction model uses the data after successful recruitment, the moral hazard caused by gender and age is not directly involved in the research process, and the prediction results obtained by the model are also in line with the enterprise’s recruitment requirements".

This is not satisfactory, and in fact, troublesome. Previous successful recruitment may have been affected by bias, and the authors' model may just embody that bias. Additionally, obviously a model is not built just for the sake of building it, but for some future application; that's the motivation the authors themselves provide in the introduction and elsewhere, like in lines 88 to 104. We don't know what the predictions are based on, there is no explainability, which is really a concern.

2. Contribution to BPNN. I had raised questions on their contribution from this point of view and their response is that "[...] it is considered that the process of solving the optimal parameters also belongs to the research contribution of the manuscript. "

I assume the authors refer to section "Determining parameters of the salary forecast model" starting line 204. I don't see why this goes beyond the standard training of a deep learning model. Sure it is important to experiment with different parameter settings and report them, but it's an application of already known methods, it does not introduce novel methods.

3. Data: contrary to the authors' claim, the data is not available. The data which is available refers to their experimental parameters, but it does not include the original dataset with the resumes on which they run their experiments.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

5 Jul 2021

Reviewer #1: I appreciate the authors have attempted to address my comments, added to the literature, added a short section "introduction to the sample" which addresses some of my questions on the data and the evaluation metrics; and eliminated the reference to human computer interaction. However, my major concerns remain unaddressed.

Reply: Thanks for your suggestions. We have further adjusted the content of the article to solve the problems. And the equation of the article is optimized and supplemented to better show the problems to be studied.

Specifically:

1. They have not addressed the issues I was raising as concerns which attributes affect salary predictions, the baseline, and importantly, whether the model is biased.

Reply: we appreciate your careful reading and comment. A number of related analysis and experiments about the impact on salary are added, and the potential recruitment bias of the model is analyzed.

a. I had asked for a baseline, as is common in ML experiments. None was provided, and the justification was "The salary of employees depends on a variety of factors. Hence, we do not study salary changes caused by a single factor such as job position. " Sure, but if there is no baseline there is the possibility that a much simpler model could be as if not more effective, and explanatory. I used "job position" as a simple example that intuitively makes sense, but many other baselines would be possible. I don't consider the GRU results a baseline in this sense.

Reply: In order to reflect the baseline of wages, we have added the wage changes based on "job type", "education level" and "work experience".

b. I had asked for some elaboration on which features may affect the model, but no insight was provided. The authors themselves say "The salary of employees depends on a variety of factors. " but we have no idea which factors matter after these experiments. The authors could have run ablation studies, although not sure their encoding of resumes allows for that. But if that's the case, they should find a way of addressing this point.

Reply: thanks for pointing out this. We have added the analysis content of influencing factors of salary, and selected three influencing factors of "job type", "work experience" and "education level" for analysis.

c. If I understand the authors' argument re bias, they basically say "this is a model about previous recruitment, so there's no moral hazard", to wit "Since the training of the prediction model uses the data after successful recruitment, the moral hazard caused by gender and age is not directly involved in the research process, and the prediction results obtained by the model are also in line with the enterprise’s recruitment requirements".

This is not satisfactory, and in fact, troublesome. Previous successful recruitment may have been affected by bias, and the authors' model may just embody that bias. Additionally, obviously a model is not built just for the sake of building it, but for some future application; that's the motivation the authors themselves provide in the introduction and elsewhere, like in lines 88 to 104. We don't know what the predictions are based on, there is no explainability, which is really a concern.

Reply: According to the reviewers, recruitment discrimination always exists in the recruitment process, so the data set adopted also inherits this possible bias. In the research, the salary forecast based on the resume information of recruitment is obtained by combining many aspects, so the recruitment will not be terminated for a certain reason. In order to reduce the possible risks, the weight of neural network of factors that may cause recruitment discrimination such as gender and age will be reduced, to eliminate this potential problem as far as possible. A description of this part has been added in the section "Determining parameters of the salary forecast model".

2. Contribution to BPNN. I had raised questions on their contribution from this point of view and their response is that "[...] it is considered that the process of solving the optimal parameters also belongs to the research contribution of the manuscript. "

I assume the authors refer to section "Determining parameters of the salary forecast model" starting line 204. I don't see why this goes beyond the standard training of a deep learning model. Sure it is important to experiment with different parameter settings and report them, but it's an application of already known methods, it does not introduce novel methods.

Reply: “Determining parameters of the salary forecast model” presents the parameter validation process for the designed model. This part can help readers know the setting of model parameters, so that readers can repeat the experiment. The model designed will be used for salary prediction, so this part can help readers better realize this model and apply it to practice.

3. Data: contrary to the authors' claim, the data is not available. The data which is available refers to their experimental parameters, but it does not include the original dataset with the resumes on which they run their experiments.

Reply: In the results and discussion, we have supplemented the content of factor level analysis and the related introduction of data set.

Submitted filename: Response to Reviewers.docx

Click here for additional data file.

27 Jul 2021

PONE-D-21-00932R3

Designing a Human Resource Management System and Analyzing Its Human-Computer Interaction Performance Under the Background of Artificial Intelligence

PLOS ONE

Dear Dr. Gong,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Sep 10 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see:  http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at  https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols .

We look forward to receiving your revised manuscript.

Kind regards,

Chi-Hua Chen, Ph.D.

Academic Editor

PLOS ONE

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #2: (No Response)

Reviewer #3: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #2: Partly

Reviewer #3: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: N/A

Reviewer #3: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #2: No

Reviewer #3: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #2: No

Reviewer #3: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #2: 1) The title is misleading and not specific:

"Designing a Human Resource Management System and Analyzing Its Human-

Computer Interaction Performance Under the Background of Artificial Intelligence"

- it is not clear what the above means?

2) Abstract is not factual. The authors did not investigate all the aspects of HR as they claim here:

"The purpose of this paper is to enhance human resources’ information management

level and strengthen enterprises’ competitiveness and development capabilities. The

Artificial Intelligence (AI) technology is adopted to optimize the Human Resource

Management (HRM) process, reduce the workload, and improve office efficiency."

3) The study and model limitations must be discussed.

4) The quality of Figures 1-11 is unacceptable.

5) The are several grammatical and syntax errors. Professional English language editing is required.

Reviewer #3: The Authors were addressed all the comments raised by the previous reviewers.

They have to elaborate how the model was affect recruitment bias and salary bias, how they will overcome this issue, data validation need be show.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: Yes: Waldemar Karwowski

Reviewer #3: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

19 Aug 2021

Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #2: 1) The title is misleading and not specific:

"Designing a Human Resource Management System and Analyzing Its Human-

Computer Interaction Performance Under the Background of Artificial Intelligence"

- it is not clear what the above means?

Reply: The title of the article has been adjusted to show the research content of the article more clearly.

2) Abstract is not factual. The authors did not investigate all the aspects of HR as they claim here:

"The purpose of this paper is to enhance human resources’ information management

level and strengthen enterprises’ competitiveness and development capabilities. The

Artificial Intelligence (AI) technology is adopted to optimize the Human Resource

Management (HRM) process, reduce the workload, and improve office efficiency."

Reply: the abstract of the article has been adjusted, briefly introducing the research content of the article.

3) The study and model limitations must be discussed.

Reply: In the conclusion of the article, the limitations of the research and model are discussed, and future work prospects are introduced.

4) The quality of Figures 1-11 is unacceptable.

Reply: we have adjusted the quality of the figures in the article, and enhanced the beauty and clarity.

5) The are several grammatical and syntax errors. Professional English language editing is required.

Reply: thanks for pointing out this. We have invited a native speaker to revise the grammatical and syntax errors.

Reviewer #3: The Authors were addressed all the comments raised by the previous reviewers.

They have to elaborate how the model was affect recruitment bias and salary bias, how they will overcome this issue, data validation need be show.

Reply: thanks for your approval of our effort. In line 230-232, we have introduced methods to reduce the employment discrimination caused by gender and age in the research process, and analyzed the correlation of salary influencing factors in "Analysis of salary relevance", the model focuses on judging criteria based on job type, work experience, and academic level, which further reduces potential recruitment issues.

Submitted filename: comments.docx

Click here for additional data file.

3 Sep 2021

PONE-D-21-00932R4

The Design and Interactive Performance of Human Resource Management System Based on Artificial Intelligence

PLOS ONE

Dear Dr. Gong,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Oct 18 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see:  http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at  https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols .

We look forward to receiving your revised manuscript.

Kind regards,

Chi-Hua Chen, Ph.D.

Academic Editor

PLOS ONE

Journal Requirements:

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: N/A

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #2: The authors made satisfactory revisions. No further comments are provided. However, the final editing for the use of the English language is recommended.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

12 Oct 2021

Date: Sep 03 2021 04:27AM

To: "Yangda Gong" gongyangda@hhu.edu.cn

From: "PLOS ONE" plosone@plos.org

Subject: PLOS ONE Decision: Revision required [PONE-D-21-00932R4]

PONE-D-21-00932R4

The Design and Interactive Performance of Human Resource Management System Based on Artificial Intelligence

PLOS ONE

Dear Dr. Gong,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Oct 18 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

�  A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

�  A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

�  An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Chi-Hua Chen, Ph.D.

Academic Editor

PLOS ONE

Journal Requirements:

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #2: All comments have been addressed

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #2: Yes

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: N/A

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #2: Yes

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #2: Yes

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #2: The authors made satisfactory revisions. No further comments are provided. However, the final editing for the use of the English language is recommended.

Reply: thanks for your careful reading and suggestion. The use of the English language has been edited.

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

In compliance with data protection regulations, you may request that we remove your personal registration details at any time. (Remove my information/details). Please contact the publication office if you have any questions.

Submitted filename: Response to Reviewers.doc

Click here for additional data file.

23 Dec 2021

Design and Interactive Performance of Human Resource Management System Based on Artificial Intelligence

PONE-D-21-00932R5

Dear Dr. Gong,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Chi-Hua Chen, Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

19 Jan 2022

PONE-D-21-00932R5

Design and Interactive Performance of Human Resource Management System Based on Artificial Intelligence

Dear Dr. Gong:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Professor Chi-Hua Chen

Academic Editor

PLOS ONE