HRM Risk Early Warning Based on a Hybrid Solution of Decision Tree and Support Vector Machine

Abstract

Human resource management is the core support and strong guarantee for the survival and innovative development of enterprises, which is directly related to the realization of the strategic development goals of enterprises. However, in today’s rapidly changing internal and external environment, HRM (Human Resource Management) activities of enterprises may face various risks at any time. Therefore, this paper designs a HRM risk warning scheme based on decision tree and support vector machine, which combines the features of support vector machine binary classification and recombines each category of multicategory classification according to decision tree multi-sub-category classification, and can identify all categories more accurately. Simulation experiments show that the convergence of the network designed in this way tends to be stable, and the performance of the early warning for enterprises meets the standard.

This work is licensed under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.

1. Introduction

With the deepening of the globalized market economy and the advent of the knowledge economy era, the competition among enterprises has become more intense, human resources have become an indispensable and critical resource for enterprise development, and enterprises have gradually realized its role and importance and raised it to an unprecedented level [1]: such as the risk of not being able to attract the right talent to match with the enterprise due to the wrong choice of recruitment channels, the risk of ignoring the needs of employees and blindly organizing training leading to unsatisfactory results, and the risk of key employees leaving unexpectedly due to unreasonable performance appraisal or unfair salary management [2, 3]. These risks will bring unpredictable losses to the enterprise if not handled properly and can even be serious enough to affect the survival of the enterprise [4]. Therefore, every enterprise should increase the monitoring of HR management risks, make accurate analysis and early warning of early risks, and take effective control programs as early as possible [5].

However, in terms of the current status of enterprise risk control, most enterprises only limit their attention to financial and operational risks when preventing and avoiding risks, but neglect human resource management risks, and have not paid enough attention to them [6]. At present, the early warning ability and control level of human resource management risks of domestic enterprises are generally low; a survey shows that 18.2% of enterprises do not pay enough attention to the unexpected departure of senior managers, while 82.8% of enterprises do not reserve the “replacement” of senior managers in advance, once the sudden departure of senior managers will make. In the event of a sudden departure of a senior manager, the enterprise will be in danger of interrupting its normal operation due to the lack of a suitable leader [7, 8]. Although these data reflect only a part of the current HR management risks of Chinese enterprises, namely, the risk of leaving and reserving senior positions, although it is only one of the perspectives, but a glimpse to see the whole picture, it makes us realize the current situation of HR (Human Resources) risk management in China; that is, HR management risks have become one of the key risks faced by enterprises and the level of risk management is generally low [9, 10]. Therefore, the implementation of HR risk management has become an inevitable thing for modern enterprises. If HRM risks are not handled properly, it will seriously affect the healthy development of the enterprise and also seriously damage the public image of the enterprise [11, 12]. Therefore, the study of HRM risk early warning and its control is crucial to the long-term development of enterprises [13].

Human resource is the most important resource in modern enterprise, which is the guarantee of talent, spiritual power, and intellectual support for enterprise development and the basis of enterprise prosperity [14]. Different from other resources, human resources have specific complexity and are more susceptible to the influence of internal and external environment. With the increasingly fierce competition in the market economy, the speed of technological progress and knowledge update, and the enhancement of employees’ self-awareness, enterprises will encounter many unprecedented challenges and face various risks in the process of human resource management [15, 16]. Improper handling of these risks will affect the normal operation of the enterprise and even bring a fatal blow to the enterprise [17]. In order to prevent the occurrence of various risk events and avoid risk losses, enterprises need to strengthen daily risk monitoring and management and prevent and solve various risks of HR management as early as possible [18]. However, due to the inherent special properties of human resources, the risks of human resource management are more hidden than other business risks and are often easy to be ignored [19]. Therefore, it is extremely important to build a practical HRM risk early warning model to predict risks early. At the same time, on the premise of effectively predicting HRM risks, it is also of great practical significance to establish an effective risk control system for the stable and healthy development of enterprises [20].

2. Related Work

In today’s knowledge-based economy, knowledge resources and human resources have become the core resources for enterprises to win competition through the expression of science and technology. More and more enterprises are becoming aware of the importance of rational and efficient human resource management. The special nature of human resources requires enterprises to implement standardized and appropriate management when conducting management, because improper employment not only causes waste of resources but also may induce the loss of organizational talent, resulting in irreparable losses and negative impacts. Human resource management risk refers to the enterprise in the job structure design, job analysis and positioning, recruitment and screening, staff training, performance management and salary management, career development design, and other human resource management activities, due to the neglect of the special characteristics of human resources and systemic and improper handling of the management activities and the linkage of the management activities, resulting in negative impact on the overall interests of the enterprise. When summarizing the seven types of business risks, risks in human resource management such as poor human resource management decisions, employee dishonest behavior, and employee conflicts within the company are also mentioned. The results of empirical studies by foreign scholars, such as Vers Bitsch, suggest that HRM risks can be classified as ethical risks, recruitment and allocation risks, training and development risks, performance appraisal and compensation management risks, turnover risks, and accidental risks.

Regarding the practical application of HRM risk research, it can be found that most studies are conducted from the micro perspective of enterprise HRM activities and delve into several specific details, such as senior manager departure risk identification, core employee turnover risk prevention, enterprise training risk management implementation, and HR outsourcing risk prevention countermeasures. Although these studies are rather fragmented and not systematic enough, they basically cover all aspects of HRM and have a strong applicability. However, most of these studies are focused on the exploration of HRM risks in a specific industry such as banking industry or high-tech industry, and there are very few studies on HRM risks in enterprises with general significance, and the quantitative studies on HRM risks are seriously insufficient, and there is a lack of corresponding early warning models.

3. Support Vector Machine Theory

3.1. Support Vector Machines

Support vector machine is a new machine learning method developed on the basis of statistical learning. The basic idea is to find the optimal classification hyperplane to completely separate the two classes of samples in the original space in the linearly divisible case and to solve a problem that is nonlinear in the low-dimensional space into a linearly divisible problem in the high-dimensional space by using a kernel approach in the nonlinear case. Let the expression of the optimal classification hyperplane be $$\begin{array}{}\text{(1)}& w\phi x+b=0,\end{array}$$where $w$ is the weight and $b$ is the bias.

Thus, the two classification problems in the original sample space can be expressed as $$\begin{array}{}\text{(2)}& y_{i}w\phi x+b\ge 1,\end{array}$$where $y_i\in 1,1$ is the output category.

Given that some samples may be misclassified, introducing the slack variable ${\epsilon }_i$ and the penalty factor C, the constraint on the hyperplane becomes $$\begin{array}{}\text{(3)}& \begin{array}{c}\text{min}\hspace{1em}\frac{1}{2}{w}^2+C\sum _{i=1}^l{\epsilon }_i\hfill \\ \text{s}.\text{t}.\hspace{1em}{y}_{i}w\phi x+b\ge 1-{\epsilon }_i.\hfill \end{array}\end{array}$$

Introducing the Lagrange multiplier ${\alpha }_i$ converts the original problem into a pairwise problem, i.e., $$\begin{array}{}\text{(4)}& \begin{array}{c}\text{max}\hspace{1em}\sum _{i=1}^l{\alpha }_i-\frac{1}{2}\sum _{i=1}^l\sum _{j=1}^l{\alpha }_i{\alpha }_j{y}_i{y}_j\phi x_i\phi y_i\hfill \\ \text{s}.\text{t}.\hspace{1em}0\le {\alpha }_i\le C\hspace{1em}\sum _{i=1}^l{y}_i{\alpha }_i=0.\hfill \end{array}\end{array}$$

According to the Kuhn-Tucker condition, ${\alpha }_i$ must satisfy $$\begin{array}{}\text{(5)}& {\alpha }_i{y}_{i}w\phi x_i+b-1+{\epsilon }_i=0.\end{array}$$

Therefore, the nonzero sample corresponding to ${\alpha }_i$ is the support vector. Solving the above problem, the optimal classification function is obtained $$\begin{array}{}\text{(6)}& fx=\mathrm{sgn}\sum _{i=1}^m{y}_i{\alpha }_{i}Kx,x_i+b,\end{array}$$where $Kx,x_i={\phi }^{\text{T}}x\phi x_i$ is the kernel function and $m$ is the number of support vectors.

3.2. Kernel Functions and Kernel Function Selection in SVM

There are many kinds of kernel functions, such as linear kernel functions, polynomial kernel functions, and radial basis kernel functions. There is still no unified decision on the selection of kernel functions, which are generally preselected based on experts’ a priori knowledge. The kernel function chosen in this paper is radial basis kernel, whose expression is $$\begin{array}{}\text{(7)}& Kx,x_i=\exp -y{u-v}^2,\end{array}$$where $u,v$ denote the two samples of the data set and $g$ is used to set the $y$ parameter setting in the kernel function, i.e., yin Eq.

4. Decision Tree Support Vector Machine

The original support vector machine is only applicable to binary classification problems. Support vector machines must be modified when used for multiclassification problems. Common approaches include one-to-one and one-to-many models. Decision tree support vector machine is also a common multicategorization model, which combines the features of support vector machine binary classification and regroups each category of multicategory classification according to decision tree multisubclass classification, which can identify all categories more accurately. For the $k$ classification problem, the method first needs to build $kk-1/2$ classification decision surfaces during training, but many categories are excluded from the prediction at each node, so the total number of classification surfaces used is $k-1$, and the classification efficiency is greatly improved. Figure 1 shows the schematic diagram of the algorithm to solve a four classification problem.

[figure(s) omitted; refer to PDF]

5. Implementation Processes

The following are the implementation processes:

(1) Feature extraction: the dissolved characteristic gases in oil used for the enterprise are mainly: ${\text{H}}_2,\text{C}{\text{H}}_4,{\text{C}}_2{\text{H}}_6,{\text{C}}_2{\text{H}}_4,\text{and}{\text{C}}_2{\text{H}}_2$ 5 gases, which carry 5 kinds of fault information of high temperature overheating, high energy discharge, low energy discharge, partial discharge, and low and medium temperature overheating of power transformer to some extent, so it can be used as characteristic gases [21, 22].

(2) Data normalization: in order to reduce the influence caused by the difference in values between various gases, the original DGA data (including the training set and the test set) are normalized by the following formula:

$$\begin{array}{}\text{(8)}& x_i^{\prime }=\frac{x_i-x_{\min }}{x_{\max }-x_{\min }},\end{array}$$where $x_i$ is the actual gas content, $x_{\max }$ is the maximum value of 5 gases in the same sample, $x_{\min }$ is the minimum value of 5 gases in the same sample, and $x_i^{\prime }$ is the normalized data.

(3) Third is the determination of the optimal parameters.

(4) Fourth is the substitution of the obtained optimal parameters into the support vector machine power enterprise model to obtain the corresponding classification model.

(5) Using the trained classification models, the samples to be tested are tested.

6. Experimental Results

In this paper, the neural network toolbox of MATLAB is chosen to build a decision tree support vector machine. Run MATLAB7.0 software, start the neural network toolbox (nntool), import data, set specific parameters of the early warning model and training functions and other functions, create a neural network and then train it, in which the main training parameters are set as follows (default values are used for other parameters not mentioned) [23, 24].

6.1. Training Function: traingdx Function

The function has additional momentum term, can be adaptive gradient decreasing, suitable for the problem of limited number of training samples, convergence is fast, high training accuracy, and can effectively avoid the problem of local miniaturization.

6.2. Weight Adjustment Function: clearngdm Function

The function adjusts the weights and thresholds by using the method of momentum gradient descent and converges most quickly, which shortens the learning time and improves the accuracy.

6.3. Performance Function: mse Function

This function is used to calculate the mean square error between the actual output value of the network and the desired output value.

6.4. Human Resource Management Risk Causation Analysis

Risk causation analysis, also known as risk identification, is to find out the reasons behind the hidden risks and the degree of influence of each reason on the risks. Risk causation analysis is the premise of risk early warning, and only when the potential causes or root causes of various risks are identified can risk identification and early warning be carried out before risks occur and cause losses, and control measures can be taken in a targeted manner. Because of the special characteristics of human resources, as shown in Figure 2, and the openness and systemic nature of human resource management activities, human resource management is different from other business management activities such as financial management and is also exposed to more risks, and most of these risks are not easily detected, so it is especially important to conduct early warning of human resource management risks.

[figure(s) omitted; refer to PDF]

In this paper, the normalized index data (21X8 matrix) of the first 8 cycles are used as the training input data of the decision tree support vector machine, and the corresponding expected output value is the actual expert evaluation score, while the data of the last 1 cycle is used as the testing data of the network to check the training of the network. Figure 3 below shows the training process of the network, from which it can be seen that the network converges to a stable level when it reaches the 39th step, indicating that the performance is up to the standard, so the training is finished.

[figure(s) omitted; refer to PDF]

Performance is 0.000143614; Goal is 0.001.

6.5. Detection of HRM Risk Warning Model Based on Decision Tree Support Vector Machine

After training the network model, this paper normalizes the index data of the last cycle as the detection data set of the network, after inputting the data, while its corresponding expert rating value is used as the expected output value, and the final detection results are shown in Table 1.

Table 1

Comparison of training and detection output of decision tree support vector machine.

Cycle time	T1	T2	T3	T4	T5	T6	T7	T8	T9
Expected output	0.39	0.41	0.52	0.58	0.75	0.69	0.14	0.25	0.26
Actual output	0.41145	0.4199	0.51988	0.6	0.75	0.69	0.12598	0.35	0.29
Error	-0.00959	-0.0169	-0.00997	0.02	0.011	0.012	-0.00789	-0.01	0.0029
Analysis results	Basic safety	Basic safety	Basic safety	Security	Security	Security	Higher risk	Risk	Risk
Expected results	Basic safety	Basic safety	Basic safety	Security	Security	Security	Higher risk	Risk	Risk

From comparing the actual output value and the expected output value, the test results show that the risk output value and the expert evaluation value for this quarter match very well, as shown in Figure 4. Therefore, the example of this company proves that the HRM risk warning model based on decision tree support vector machine can be used for HRM risk warning with practicality.

[figure(s) omitted; refer to PDF]

The special characteristics of human resources themselves determine the existence of human resource management risks. Unlike other objective enterprise resources such as capital and raw materials, human resources are the most active and the only factor with subjective dynamism among the factors of productivity. Human resources are directed by their brains and higher nervous systems, and they have a dynamism that material elements do not have. This dynamism is subjective and can be influenced by many factors such as cultural environment, personality traits, and knowledge reserves, making it impossible for two people to be exactly the same. As shown in Figure 5, human resources do not carry out orders in strict accordance with regulations and instructions like machines and equipment and other resources, which may lead to inconsistencies between the results of behavior and expected results due to various reasons, and even programmed work hides inconsistencies between actual and expected goals due to the involvement of people, and the effect of such inconsistent behavior is, to a certain extent, the human resources that the organization has to face.; Therefore, the specificity of human resource itself determines that human resource management will encounter inevitable management risks.

[figure(s) omitted; refer to PDF]

7. Conclusion

The risks of HRM are latent in each link, and only by paying full attention to the coordination and connection of each component or level in the actual operation can we improve its orderliness and overall operation and avoid management risks as much as possible. In this paper, we will reveal the existing risks specifically for each link of HRM and establish a detailed risk causation index system to lay the foundation for the construction of risk warning model.

References

[1] M. S. Tehrany, B. Pradhan, M. N. Jebur, "Flood susceptibility analysis and its verification using a novel ensemble support vector machine and frequency ratio method," Stochastic Environmental Research and Risk Assessment, vol. 29 no. 4, pp. 1149-1165, DOI: 10.1007/s00477-015-1021-9, 2015.

[2] R. C. Deo, O. Kisi, V. P. Singh, "Drought forecasting in eastern Australia using multivariate adaptive regression spline, least square support vector machine and M5Tree model," Atmospheric Research, vol. 184, pp. 149-175, DOI: 10.1016/j.atmosres.2016.10.004, 2017.

[3] V. H. Nhu, A. Shirzadi, H. Shahabi, S. K. Singh, N. al-Ansari, J. J. Clague, A. Jaafari, W. Chen, S. Miraki, J. Dou, C. Luu, K. Górski, B. Thai Pham, H. D. Nguyen, B. B. Ahmad, "Shallow landslide susceptibility mapping: a comparison between logistic model tree, logistic regression, naïve Bayes tree, artificial neural network, and support vector machine algorithms," International Journal of Environmental Research and Public Health, vol. 17 no. 8,DOI: 10.3390/ijerph17082749, 2020.

[4] D. Tien Bui, B. T. Pham, Q. P. Nguyen, N. D. Hoang, "Spatial prediction of rainfall-induced shallow landslides using hybrid integration approach of least-squares support vector machines and differential evolution optimization: a case study in Central Vietnam," International Journal of Digital Earth, vol. 9 no. 11, pp. 1077-1097, DOI: 10.1080/17538947.2016.1169561, 2016.

[5] E. Owusu, P. Boakye-Sekyerehene, J. K. Appati, J. Y. Ludu, "Computer-aided diagnostics of heart disease risk prediction using boosting support vector machine," Computational Intelligence and Neuroscience, vol. 2021,DOI: 10.1155/2021/3152618, 2021.

[6] B. T. Pham, T. V. Phong, T. Nguyen-Thoi, K. Parial, S. K. Singh, H. B. Ly, K. T. Nguyen, L. S. Ho, H. V. Le, I. Prakash, "Ensemble modeling of landslide susceptibility using random subspace learner and different decision tree classifiers," Geocarto International, vol. 37 no. 3,DOI: 10.1080/10106049.2020.1737972, 2022.

[7] F. B. Saghezchi, A. Radwan, J. Rodriguez, T. Dagiuklas, "Coalition formation game toward green mobile terminals in heterogeneous wireless networks," IEEE Wireless Communications, vol. 20 no. 5, pp. 85-91, DOI: 10.1007/s11440-022-01455-2, 2013.

[8] B. Ji, Y. Li, D. Cao, C. Li, S. Mumtaz, D. Wang, "Secrecy performance analysis of UAV assisted relay transmission for cognitive network with energy harvesting," IEEE Transactions on Vehicular Technology, vol. 69 no. 7, pp. 7404-7415, DOI: 10.1109/TVT.2020.2989297, 2020.

[9] X. Lin, J. Wu, S. Mumtaz, S. Garg, J. Li, M. Guizani, "Blockchain-based on-demand computing resource trading in IoV-assisted smart city," IEEE Transactions on Emerging Topics in Computing, vol. 9 no. 3, pp. 1373-1385, DOI: 10.1109/TETC.2020.2971831, 2021.

[10] D. Tien Bui, H. Shahabi, A. Shirzadi, K. Chapi, M. Alizadeh, W. Chen, A. Mohammadi, B. Ahmad, M. Panahi, H. Hong, Y. Tian, "Landslide detection and susceptibility mapping by AIRSAR data using support vector machine and index of entropy models in Cameron highlands, Malaysia," Remote Sensing, vol. 10 no. 10,DOI: 10.3390/rs10101527, 2018.

[11] S. Lee, S. M. Hong, H. S. Jung, "A support vector machine for landslide susceptibility mapping in Gangwon Province, Korea," Sustainability, vol. 9 no. 1,DOI: 10.3390/su9010048, 2017.

[12] Y. Huang, L. Zhao, "Review on landslide susceptibility mapping using support vector machines," Catena, vol. 165, pp. 520-529, DOI: 10.1016/j.catena.2018.03.003, 2018.

[13] M. Sahana, S. Rehman, H. Sajjad, H. Hong, "Exploring effectiveness of frequency ratio and support vector machine models in storm surge flood susceptibility assessment: a study of Sundarban Biosphere Reserve, India," Catena, vol. 189, article 104450,DOI: 10.1016/j.catena.2019.104450, 2020.

[14] G. Liu, W. Jia, M. Wang, A. A. Heidari, H. Chen, Y. Luo, C. Li, "Predicting cervical hyperextension injury: a covariance guided sine cosine support vector machine," IEEE Access, vol. 8, pp. 46895-46908, DOI: 10.1109/ACCESS.2020.2978102, 2020.

[15] P. An, Z. Wang, C. Zhang, "Ensemble unsupervised autoencoders and Gaussian mixture model for cyberattack detection," Information Processing & Management, vol. 59 no. 2, article 102844,DOI: 10.1016/j.ipm.2021.102844, 2022.

[16] M. S. Tehrany, B. Pradhan, M. N. Jebur, "Spatial prediction of flood susceptible areas using rule based decision tree (DT) and a novel ensemble bivariate and multivariate statistical models in GIS," Journal of Hydrology, vol. 504, pp. 69-79, DOI: 10.1016/j.jhydrol.2013.09.034, 2013.

[17] D. T. Bui, P. T. T. Ngo, T. D. Pham, A. Jaafari, N. Q. Minh, P. V. Hoa, P. Samui, "A novel hybrid approach based on a swarm intelligence optimized extreme learning machine for flash flood susceptibility mapping," Catena, vol. 179, pp. 184-196, DOI: 10.1016/j.catena.2019.04.009, 2019.

[18] R. Costache, A. Arabameri, I. Elkhrachy, O. Ghorbanzadeh, Q. B. Pham, "Detection of areas prone to flood risk using state-of-the-art machine learning models," Geomatics, Natural Hazards and Risk, vol. 12 no. 1, pp. 1488-1507, DOI: 10.1080/19475705.2021.1920480, 2021.

[19] Z. W. Zhang, D. Wu, C. J. Zhang, "Study of cellular traffic prediction based on multi-channel sparse LSTM," Computer Science, vol. 48 no. 6, pp. 296-300, DOI: 10.11896/jsjkx.210400134, 2021.

[20] Z. Guo, Y. Shi, F. Huang, X. Fan, J. Huang, "Landslide susceptibility zonation method based on C5.0 decision tree and K-means cluster algorithms to improve the efficiency of risk management," Geoscience Frontiers, vol. 12 no. 6, article 101249,DOI: 10.1016/j.gsf.2021.101249, 2021.

[21] G. F. Lin, M. J. Chang, Y. C. Huang, J. Y. Ho, "Assessment of susceptibility to rainfall-induced landslides using improved self-organizing linear output map, support vector machine, and logistic regression," Engineering Geology, vol. 224, pp. 62-74, DOI: 10.1016/j.enggeo.2017.05.009, 2017.

[22] B. T. Pham, D. Tien Bui, I. Prakash, L. H. Nguyen, M. B. Dholakia, "A comparative study of sequential minimal optimization-based support vector machines, vote feature intervals, and logistic regression in landslide susceptibility assessment using GIS," Environmental Earth Sciences, vol. 76 no. 10,DOI: 10.1007/s12665-017-6689-3, 2017.

[23] A. K. M. Masum, L. S. Beh, M. A. K. Azad, K. Hoque, "Intelligent human resource information system (i-HRIS): a holistic decision support framework for HR excellence," International Arab Journal of Information Technology, vol. 15 no. 1, pp. 121-130, 2018.

[24] S. A. Taqvi, L. D. Tufa, H. Zabiri, A. S. Maulud, F. Uddin, "Multiple fault diagnosis in distillation column using multikernel support vector machine," Industrial & Engineering Chemistry Research, vol. 57 no. 43, pp. 14689-14706, DOI: 10.1021/acs.iecr.8b03360, 2018.