Proposing a framework for specifying the appropriateness of pursuing a career regarding Employees’ healthiness and safeguard: A Case study of food manufacturing industry

Abstract


Introduction And Background
One of the most valuable assets of an organization is human resources, and the organization's growth and development highly depend on planning for employees (Desarno et al., 2021). Human Resources Management area revolves around recruiting, hiring, deploying, and managing an organization's employees (Rihan, 1998).
One of the most signi cant issues that HR managers should take into consideration is the Health and safety of employees in the workplace (Perera, 2019;Mosadeghrad, 2013;Anger et al., 2015). There is an undeniably direct correlation between job performance and employees' health conditions (Lukas et al., 2007), especially in manufacturing rms (Perera, 2019). Taking employees' Health into consideration in the workplace by managers assists prevent possible injuries in the workplace and puts workforces in their proper positions (Linnander et al., 2017;Williams and Skinner, 2003); moreover, the job insecurity rate will be reduced if applicable employees were recruited (Jiang et al., 2021). Johnson et al. (2017) stated that physical potential for employees is one of the substantial factors for managers to employ workforces. Shaffer and Gilbert (2018) mentioned that Adherencing to health standards in the workplace leads to protecting employees from hazardous and immune families from adverse effects of harming in workspaces. Guerin et al. (2020) referred to Many work environments with special conditions, especially in the production plant environment where the risk to employees is higher; hence, it is essential to concentrate on employees' Health in such environments. Identifying diseases and physical conditions level of employees is costly for the organization, and it is laborious for the organization to repeat it regularly (Ross et al., 2016); hence managers look forward to predicting employees' health conditions by approaches that are time-saving and less costly (Wu et al., 2019) (Smiti, 2020).
Machine Learning is the state-of-the-art method that can be utilized to identify the pattern in the data and predict the outcome according to data (Motwani et al., 2017). Different types of ML algorithms can be developed to identify diseases and unhealthy employees to intervene adverse effects on workforces and companies (Liu et al., 2021 Apart from ML applications in disease diagnoses, it is employed to predict the employees' workplace health. ML algorithms are perfectly functional in the eld of health management in the organization, and attempted to predict the noise-induced hearing impairment of laborers exposed to various complex occupational noises with four classi cation models, which resulted that the best model for this prediction is SVM, which could gain 80.1% accuracy.
In this paper, medical factors were utilized for Human Resource Management purposes. According to the literature, it is recognizable that machine learning methods have been widely employed in disease diagnosis. However, These algorithms are less applied in predicting the Health of human resources employees in organizations, especially the manufacturing industry, where employees experience complicated conditions. After concerning medical features, exploiting the underlying ML algorithms to predict the Suitability to carry on working in factory workplace conditions for employees is an actionable decision for the HRM department.

Problem Description
Staff health assessment and determining whether the position they worked at is appropriate or may harm them is a leading factor in HRM. Managers must safeguard the workers' health conditions in the manufacturing industry to prevent side effects the workplace may have for the workers, furthermore; It is signi cant to recognize that each individual can work in a speci c position in the factory due to their health level. Apart from this, during recruitment, a novel approach can be utilized to determine whether a person is perfectly tted for that position or the workplace may harm one's Health. Managers look forward to exploiting the e cient, time-saving method and have reasonable accuracy in predicting appropriate individuals for the manufacturing industry according to their Healthiness level. ML algorithms can be perfectly applied to recognize the effectiveness of medical components on a career, consequently paving the way for human resources managers to select the most productive individuals. The whole process of the research problem and solving is displayed in Fig 1.

Case Study
The case utilized in this Research is a food manufacturing factory. The employees who worked at the production line participated in this paper. Blood tests and general practitioners' examinations were the basis for forming factors involved in this study. This center's employees' tasks revolve mainly around packaging, labeling, or storing food for human consumption and providing food for sale or distribution to other business entities such as food processing plants or food establishments. This factory and the many food chain industries have poor conditions in terms of considering laborers' Health; however, hazards in this environment, especially facing ear-splitting facilities, should be considered so as not to affect employees' Health.

Factors
Labors' general Health has multiple factors which can be evaluated to recognize the physical Health of workers. For implementing the model, it is necessary to take the features into account which is related to the professional environment and can affect employees' task implementation (Chaudoir et al., 2013;Milat et al., 2015). Collection and preparation of information related to blood and urine can be utilized as a factor in assessing individuals' health conditions (Cornelis et al., 1996). Blood composition is plasma, red blood cells, white blood cells, and platelets, all of which can be normal, high, or low in the medical examination measurement; thus, according to the amount of these factors in the blood, health level can be assessed (Sun et al., 2019). Urine is another factor constituting Electrolytes, nitrogenous chemicals, vitamins, hormones, organic acids, which should be considered for healthiness (Karak and Bhattacharyya, 2011). Many tasks are performed in the close viewing distance in the workplace condition; thus, to carry out responsibility with high potential, employees' eye vision within the factory is critical to be examined (Anshel, 2006). Hearing conditions and their protection are substantially signi cant in the workplace, especially in laboratory conditions; hence, employees' hearing levels have to be determined (Nélisse et al., 2012). Audiometry test consists of ve levels: Normal (less than 25 DB HL), Mild (25-50 DBHL), Moderate (41-65 DB HL), Severe (66-90 DB HL), Profound ( more than 90 DBHL) (Davies, 2016) and Notch (hearing loss at 3 to 6 kHz) (Nélisse et al., 2012). The Standard threshold for hearing has to be considered in the workplaces, and appropriate employees have to be set for positions according to their hearing level (Daniell et al., 2003). Age and gender are two other factors that affect the work environment and employee performance (Wilks and Neto, 2013). In this paper, all of the factors mentioned above have been utilized to present a functional healthcare system for employees to determine whether pursuing their career is appropriate or may harm their physical condition. The criteria for measuring the factors are summarized in Table 1.

Machine Learning
The main goal of this study was to propose a system to predict whether a job position is appropriate for a speci c worker or may harm physical health. For this purpose, after processing raw data (data transformation and normalization) to be suitable for utilizing ML algorithms, four classi cation models, namely, Logistic Regression, Support Vector Machine, Decision Tree, K-Nearest-Neighbors, were tted on the dataset. Among these algorithms, LR is proposed for this study since it has the highest accuracy. LR is a functional approach in classi cation models, widely employed in assessing data in sociology, biostatistics, clinical medicine, quantitative psychology, econometrics, marketing, and many other areas (Dreiseitl and Ohno-Machado, 2002).
For the implementation of the LR Algorithm, the best parameter for the b 0 and b 1 is to be picked out. The equation for LR is as follows (Hosmer et al., 2000): Here y is the predicted output, and b 0 (the bias or intercept term) and b 1 (the coe cient for the single input value x) are to be chosen by minimizing the LR cost function. The equation for LR cost function is as follows:

Cross-validation
Many algorithms can be employed to assess the ML models, namely, Train/test split, k-fold Cross-Validation (CV), Leave-one-out Cross-Validation (LOOCV), Leave-one-group-out Cross-Validation (LOGOCV), Nested Cross-Validation (NCV), Wilcoxon signed-rank test, Mc Nemar's test, etc. The k-fold cross-validation is one of the most practical techniques for estimating the classi er's error (Anguita et al., 2012). The KCV process consists of splitting a dataset into k subsets; afterward, iteratively, some of them are utilized to build the model, while the others are exploited to analyze its performance.

Model Validation
The Confusion Matrix is a widely used method to recognize the correlation between the actual and predicted values (Xu et al., 2020). The Precision, Recall, F1-score, and Accuracy metrics are calculated based on the Confusion Matrix. The formulas for these metrics are as follows (Townsend, 1971 It is conceivable that the most contributing feature for pursuing a job is a Normal hearing condition; therefore, a healthy hearing condition has to be considered during recruitment and picking an individual for a speci c career position in the factory. After highlighting the factor importance, The ML algorithms were implemented.

Machine Learning Results
As Cross-Validation technique is utilized in this Research. Since the LR algorithm has the highest accuracy, the confusion Matrix for LR is demonstrated in Fig. 3.
Based on the Confusion Matrix, the Precision, Recall, F1-Score, and Accuracy can be extracted. In Table 2, the metrics mentioned above for all applied algorithms are represented. According to the results, The most practical algorithms to predict whether a job position is appropriate for an employee or may harm one's physical condition is the LR model with an accuracy of 84%.
The metrics, as mentioned earlier, are also calculated for each class, and the outcome is presented in Table 3. The numerical result reports that the proposed model is better performed at predicting the yes labels in the model.

Discussion
The results marked that the LR algorithm with relatively high accuracy is highly functional in predicting whether working at that job position is unimpeded or the current situation is not suitable concerning one's health level. In addition, forecasting assists employees in avoiding the physical damage that they are prone to in the factory environment. The magnitude of this problem is determined; for instance, the hearing loss rate for workers exposed to occupational noise in the textile factory is 24% higher than those who are not exposed to the noise (Shakhatreh et al., 2000). According to the Importance of this issue, researchers predict the possibility of hearing loss in noisy workplaces (Aliabadi et al., 2015). Hence, providing high-quality healthcare in a factory environment should be considered (Mosadeghrad, 2014;Bradley et al., 2009). Workers exposed to excessive noise in the workplace have to control the noise by utilizing hearing protection, and in some cases, this issue is highly signi cant that the employer has to implement a hearing conservation program for workers (Sriopas et al., 2016).
After making a prediction, human resources should be directed to address the problem if they have a health problem or take steps to protect them from further harm. In any case, for employees who are not in a suitable position, the possibility of pursuing a job has to be considered, including sending them to a specialist to x the problem, or utilizing protective equipment, or performing outpatient treatment by the general practitioners who have performed the examinations.

Managerial Insights
When considering a managerial perspective, the time and budget spent by the supervisors to implement a method are signi cant. In this paper, the proposed system assists the human resources manager determine the Suitability of industrial workers for their occupation, which prevents potential injuries to employees and boosts employee performance. The system provides the situation for performing these operations in less time with High accuracy; As a result of implementing this method, the organization's cost will be reduced due to diminishing employees' injuries and health risks.
The critical point in human resource management is to hire e cient employees for higher productivity. This procedure can also check employees' Health at hiring and be introduced as a contributing technique in the recruitment process. The level of health of employees is considered less during hiring; therefore, it has a practical utilization for the organization and human resource managers.

Limitations
In this paper, some limitations affected the whole process. First, the captured data was from a food manufacturing industry, and the number of persons who participated in this Research was only 231 members; the dataset was formed according to examination by a general practitioner. Second, Concerning the small dataset, the possibility of facing over tting is high; therefore, ve-fold crossvalidation was utilized to prevent over tting. Third, a broad number of features can be applied for recognizing the health level; however, eight independent factors were utilized to develop the Machine Learning Models in this Research since the available dataset was restricted to these factors.

Conclusion
In this Research, Machine Learning Algorithms, named SVM, KNN, LR, and Decision Tree, were utilized for training captured data to predict the feasibility of pursuing a career for employees concerning their healthiness. Fivefold cross-validation was employed for the model evaluation, resulted in opting LR algorithm as the most contributing method; furthermore, The confusion matrix represented the 84% accuracy for the model mentioned above. Eight independent features were applied to predict the dependent variable in the classi cation model, named pursuing a job with yes and no labels. The 2D Heat Map displayed the amount of correlation between variables; Hearing condition was the leading feature for workers to pursue their jobs in a factory workplace. According to this Research, relying on advancement in computer Science (ML algorithms) can assist managers to take physical condition into account for opting employees for speci c career positions.
For future Research, Taking advantage of other health factors to determine the suitability of the job for employees can be considered. Apart from the physical problems of the body, which can be solved mainly through applying protective facilities and referring to specialists, considering mental Health can be another vital factor, which should be taken into consideration for managers. Moreover, employing many other novel ML algorithms may lead to better performance in this area, which researchers can utilize to train on the captured data.

Declarations
Ethics approval and consent to participate: Not applicable.

Consent for publication:
Not applicable.
Availability of data and materials: The dataset during the current study is available from the corresponding author on reasonable request.

Competing interests:
The authors declare that they have no competing interests. Funding: Not applicable.
Authors' contributions: Ayria Behdinian prepared materials, data collection, data analysis, developed the Models and drafted the manuscript. Kamran Rezaie and Ali Bozorgi-Amiri commented on and aided in the drafting of the manuscript. All authors read and approved the nal manuscript.

Figure 1
The problem and solving process RepotingChecklist.pdf