Predictive Approach for Unemployment Management Using Data-Driven Techniques

Predictive Approach for Unemployment Management Using Data-Driven Techniques

¹Ugwu Edith A., ²Chinonso Joseph Okonkwo

¹Computer Science Department, Enugu State University of Science and Technology (ESUT)

²Computer Science Department, Chukwuemeka Odumegwu OjukwuUniversity, Anambra State

DOI: https://doi.org/10.51584/IJRIAS.2025.1005000128

Received: 29 May 2025; Accepted: 03 June 2025; Published: 25 June 2025

ABSTRACT

This study presents the modelling and implementation of an integrated system for unemployment prediction and employment tracking in Nigeria using data-driven techniques. The unemployment prediction system was initially developed using a linear regression model trained on five years of data from the National Youth Service Corps (NYSC) covering graduates of Higher Education Institutions (HEIs). In parallel, an Employment Tracking System (ETS) was developed using the Feed Forward Neural Network (FFNN) architecture trained using fingerprint data from the Federal Ministry of Labour, Employment, and Productivity (FMLEP) and both systems were implemented using MATLAB’s regression and neural network toolboxes and validated through rigorous testing and evaluation metrics. The FFNN algorithm was introduced and reconfigured to train on the NYSC dataset. The result of the applied FFNN demonstrated performance coefficient of determination with an R² of 0.99624, a Mean Square Error (MSE) of 0.0025165 and a Root Mean Square Error (RMSE) of 0.050165, showing a significant improvement in prediction accuracy. The ETS effectively classifies individuals as employed or unemployed based on biometric inputs. The integrated model provides a reliable platform for forecasting unemployment trends and tracking employment status, offering valuable insights for policy formulation and labour market planning in Nigeria.

Keywords: Unemployment Prediction; Employment Tracking System (ETS); Feed Forward Neural Network (FFNN); Linear Regression; NYSC

INTRODUCTION

To reduce unemployment rate in Nigeria drastically, there is need for a Digital Nervous System (DNS) which will intelligently detect the frauds in the government ministries, departments and agencies (MDA) by employee tracking system and also make correct estimate of the actual number of unemployed persons in the country. This will enable the government plan effectively to create jobs for the teeming unemployed Nigerian youths (Didiharyono and Syukri, 2020; Barnichon and Garda, 2016).

According to Ajah (2013), DNS is a modern vision of a computer network that imitates the biological nervous system in the coordination and processing of information. To this end, many studies have been carried out to tackle this problem of unemployment (Bello, 2003; Adesegun et al., 2020; Elijah et al., 2015; Etuk and Onwuachu., 2016). However, despite the result of these studies, none of them was able to effectively deal with the problem of internal fraud within theFederal Ministry of Labour, Employment and Productivity (FMLEP) or help the government to track and monitor employees of the ministries, departments and agencies (MDAs) (Olorunfemi, 2021; Nguyen et al., 2021).

To solve these problems, this study focused on providing a DNS system that will be able to track employment status and verify new employees. The approach utilized physiological traits of the employees to track their employment status. Secondly an intelligent regression model was developed using Machine Learning (ML) technique (Ebere et al., 2025; Habor et al., 2021) to correctly estimate in time series the number of unemployed graduates in the country over the next ten years. ML is an artificial intelligence technique which can learn and make accurate decisions without being explicitly programmed (Ituma and Asogwa, 2018; Kekong et al., 2021). This ML was trained with data collected from the relevant agencies of the federal government to develop a predictive model which will help the government determine the number of unemployed graduates from Higher Education Institutes (HEIs) for the next ten years and then effectively plan for them.

RESEARCH METHODOLOGY

Research methodology and systems development methodologies were used in this study. Development of a system that can effectively track the employed and unemployed HEIs graduates was achieved using fingerprint technology and feed forward neural network. Linear regression techniques were used to effectively predict unemployment status. The FMLEP at every state as well as the NYSC database were digitally linked via internet. The new system was analysed using Object Oriented Analysis and Design Methodology (OOADM). The reason was because of its ability to balance emphasis between process and data; using universal modelling diagrams to describe the system concepts and achieve quality system structure (Ajah and Ugah, 2013).

Data collection

The method used for the data collection is the secondary data analysis method, which made use of already existing dataset to solve research problem. In this study, the primary source of data collection is the Federal Ministry of Labour, Employment and Productivity (FMLEP), Southeast Zone, Enugu State, Nigeria. The data provided were biological and biometric information of employed persons working with the Federal Government of Nigerian agencies under the FMLEP such as National Directorate of Employment (NDE), National Productivity Centre (NPC), Michael Imoudu National Institute for Labour Studies (MINILS), Nigeria Social Insurance Trust Fund (NSITF), Industrial Arbitration Panel (IAP)from the age of 20-65yrs. These data provided the primary requirement for a biometric system for employment verification. However, from these data, 10 employees were selected to test and validate the study. The data collected were stored in a MYSQL dataset and used for the development of the employment tracking system. The instrument used for the data collection was the Damalog fingerprint scanner, which extracted the fingerprint information of the employees as shown in the samples of Figure 1.

Figure 1: Samples of fingerprints collected (Source: FMLEP)

The secondary source of data collection was the National Youth Service Corp (NYSC), Enugu State, which provided data of NYSC discharged members from 2014 to 2021. The data reflected qualified persons which graduated from 296 higher education institutes (Universities and Polytechnics). The total sample size of data collected for these graduates over the past eight years is 2.4 million people whose records were used to develop the predictive model.

The Proposed System

System analysis buttresses the processes discussed in the previous section to identify the system goal and purpose in order to design the new system which will effectively achieve them. These can be achieved using block diagram, fault tree analysis or decision tree analysis. The use of decision tree was preferred above others in this system design. Apart from its simplicity, it has the ability to define requirements and model for the employment tracking system considered. Block diagram was used for the analysis of the unemployment estimation system. It can also gather, analyze and validate information, evaluate alternatives and prioritize requirements and finally examine the information needed by the end user and enhance the system goal.

In the existing system, there is no employment tracking system currently at the Ministry of Labour, Employment and Productivity (FMLEP) or her employment agencies all over the country. This is the major reason for the problem such as employment fraud, ghost workers, very high unemployment rate, and large-scale corruption in the service. No doubt, biometric technology is current in use in some agencies, but these are applied strictly for employee management reasons and since there is no nervous system coordinating the agencies via one central database, it is very prone to manipulation and fraud. To this end, there is need for employment tracking system which can identify the position and agency one is engaged with in order not to hold two or more jobs. In the new system developed, the decision tree was used for the system analysis as shown in the employment tacking system analysis of Figure 2.

Figure 2. Decision tree of the employment tracking system

Figure 2 shows the workflow of the Employment Tracking System (ETS), indicating the various processes undertaken to achieve the DNS. The system is made up of data collected from FMLEP which contains both the biological information of the employees, like name, work status, place of work, position, date of birth, address, etc, and biometric data, such as the fingerprint. The biological information was used to develop the employment management system, while the biometric data collected with the fingerprint scanner and processed using dilation and histogram equalization techniques were trained with FFNN to generate a fingerprint verification system which was integrated with the employee management system to achieve the new employment tracking system. The other section of the system is the unemployment estimation model analysed in this case with block diagram (Figure 3).

Figure 3: Analysis of the unemployment estimation system

In Figure 3, a model for prediction of unemployed Nigerian graduates was analysed. The system collected data of Nigerian graduates of HEIs from the NYSC secretariat and then trained the data with MLNN to develop the prediction model which was used to estimate the number expected to graduate from HEIs in the next ten years. Figure 4 shows the analysis of the DNS and how the ETS and the UFS were married in the system in order to tackle the problem of graduate unemployment in Nigeria.

Figure 4: Complete system fault tree analysis.

Model of the Linear Regression Techniques for Prediction of NYSC Data

In the family of machine learning, linear regression is specialized in solving time series problem using the relationship between dependent and independent variables. The linear regression was utilized for the development of the prediction model against the other machine learning algorithms mentioned in the literature review because it is specially developed for predictions which can correctly solve time series challenges. According to Roopa and Asha (2019), it is a machine learning algorithm which is employed for time series forecasting of events and also for the determination of relationships between two variables of dependent and independent types. In modelling of the regression, linear regression model of Kavitha et al. (2016) was utilized and presented as:

\[
y_i = \beta_0 + \beta_1 X_{i1} + \beta_2 X_{i2} + \cdots + \beta_p X_{ip} + \varepsilon_i, \quad i = 1, 2, 3, \ldots, n \tag{1}
\])

\text{where } y_i \text{ is the } i^{th} \text{ response, } \beta_p \text{ is the } p^{th} \text{ coefficient of the slope, } \beta_0 \text{ is the model constant term. } X_{ip} \text{ is the } i^{th} \text{ observation on the } j^{th} \text{ predictor variable, } j = 1, \ldots, p, \varepsilon_i \text{ is the } i^{th} \text{ noise term which is the random error.} \text{Since there is only one variable which is the number of NYSC members in the past five years, the linear regression model can be presented as:}

\[
y_i = \beta_0 + \sum_{k=1}^{k} \beta_k f_k (X_{i1}, X_{i2}, \ldots, X_{ip}) + \varepsilon_i, \quad i = 1, \ldots, n \tag{2}
\]

\text{where } f_i \text{ is the scalar value function of the independent variable } X_{ip}, \beta_k \text{ is the slope coefficient and } y \text{ is the response variable.} \text{The idea in the regression model was to achieve a best fitting line with the data. The error (E) between the actual prediction and the line of best fit was given by the equation:}

\[
E(y_i) = E\left(\sum_{k=0}^{k} \beta_k f_k (X_{i1}, X_{i2}, \ldots, X_{ip}) + \varepsilon_i \right) \tag{3}
\]

\text{The mean error output in Equation 3 can be represented as Equation 4 and the variance which is Equation 5 as the error square (Kavitha et al., 2016):}

\[
E(y_i) = \sum_{k=1}^{k} \beta_k f_k (X_{i1}, X_{i2}, \ldots, X_{ip}) + E(\varepsilon_i) \tag{4}
\]

\[
V(y_i) = V\left(\sum_{k=0}^{k} \beta_k X_k (X_{i1}, X_{i2}, \ldots, X_{ip}) + \varepsilon_i \right) = V(y_i) = \sigma^2 \tag{5}
\]

The linear regression model which depicts the relationship between the dependent (y) and independent variable (x) using the data model of Equation 1. The line of best fit in Equation 6 was used to show the mean variables which was predicted using Equation 2. The difference between the actual data in Equation 2 and the predicted data in Equation 2 presents the error model in Equation 4 which was used to compute the variance in Equation 5 as a means to evaluate the regression performance. The slope shows the changes between the dependent and independent variables during the prediction performance. The line of best fit was presented using the fitted coefficient model as:

\[
\hat{y_i} = \sum_{k=1}^{k} \beta_k f_k (X_{i1}, X_{i2}, \ldots, X_{ip}) + \varepsilon_i, \quad i = 1, \ldots, n \tag{6}
\]

\text{where } \hat{y_i} \text{ is the response estimated and } \beta_k. \text{ The estimated coefficient seeks to minimize the error difference between the predicted vector } \hat{y_i} \text{ and the true vector } y, \text{ which is } \hat{y_i} – y \text{ based on the least mean-square method (Neter et al., 1996).}

The activity model of the linear regression technique

Activity model is a graphical representation of a workflow of stepwise activities and actions with support for choice, iteration and concurrency. The activity model gives the description of the process flow among multiple objects of a class during the activity processing. They were used (Figures 5 and 6) in association with the UML modelling methods to achieve the major objective of molding templates for workflow behind the system being developed.

The model was adopted for presentation of the linear regression based on its compatibility with the adopted methodology for the study. The activity model of the linear regression is as shown in Figure 5. The figure shows how the dependent and independent variables are related and then used to compute line of best fit based on minimal least square error achieved from the data computation. The idea is to achieve the least error and then generate prediction model which was later used in the modelling of unemployment prediction system.

Figure 5: Activity model of the linear regression

Modelling of the unemployment prediction system

The model of the unemployment prediction system was developed using the regression technique. The model utilized the data model of higher education graduates which have completed the national service and then fed into the regression model for training and then to generate the prediction model for the time series estimation of unemployment in Nigeria. The linear regression model of Figure 7 was loaded with the NYSC data collected over the period of 5 years and then trained with the linear regression to generate the unemployment prediction system. During the training process, key performance evaluation parameters such as regression, mean square error, average mean square error and 5-fold cross validation approach were used to analyze the performance of the prediction model and when the desired results were achieved for respective parameters. The unemployment prediction model was then generated as shown in the activity diagram of Figure 6.

Figure 6: Activity diagram of unemployment prediction model

The unemployment prediction model generated was used to develop an unemployment prediction system using regression application software in MATLAB. The activity model of the system when used to predict number of fresh graduates of HEIs for the next two years is as shown in Figure 7.

Figure 7: The model of the unemployment prediction system.

The data of HEIs graduates who completed their national service of the NYSC for the last two years were loaded into the unemployment prediction model which computed and predicted the number of NYSC discharged members who will be joining the labour market in the next two years.

Model of the Feed Forward Neural Network (FFNN) for the ETS

The section presents the model of the FFNN utilized for the development of the ETS. The FFNN utilized for the study was adopted from Eneh et al. (2022) and reconfigured for the training of the employee data fingerprint data. The FFNN is a simple form of neural network which can solve pattern recognition problem using image data. FFNNfrom relevant literatures where applied has achieved good classification accuracy. The FFNN in this work was developed from a basic neuron which have weight, bias and activation function as shown in Figure 8.

Figure 8: Architectural model of a neuron

Figure 8 shows the model of the FFNN utilized for the training of the fingerprint data to generate the ETS model. The model shows the interconnection of neurons which have weights and bias function summed up and activated to ensure compatible output when data were fed. The mathematical model of the FFNN is presented as:

\( y_k = \theta_k \left(w_k p + \sum_{i=1}^{v_k} x_i w_i \right) \tag{7} \)

where \( w_i \) is the weight of the neuron, \( p \) is the number of inputs, \( w_k \) is the bias, \( v_k \) is the net input function, \( x_i \) is the data, \( \theta_k \) is the activation function, and \( y_k \) is the activated output.

The number of hidden layers of the neural network was determined based on the model in Equation 8 (Mzachary, 2019):

\( v_k = \frac{N_s}{\alpha (N_i + N_o)} \tag{8} \)

where \( N_i \) is the number of input neurons; \( N_o \) is the output neurons, \( N_s \) is the samples of the fingerprint data in the training set; \( \alpha \) is the scaling factor.

The model of the activation function is presented as:

\[
f(g) =
\begin{cases}
0 & \text{for } x < 0 \\
x & \text{for } x \geq 0
\end{cases}
=
\begin{cases}
0 & \text{for } x < 0 \\
1 & \text{for } x \geq 0
\end{cases} \tag{9}
\]

where \( x \) is the data input. The ReLU is a nonlinear function which ensures compatibility on the output of the data from neurons and within a range of 0 and 1.

The mathematical model of the FFNN with the ReLU in Equation 7 is given as:

\( y = g \left(w_o + X^T W \right) \tag{10} \)

where g is the ReLU, X is the sum of net inputs, T is the total number of input. To train the FFNN model, a training algorithm was used which allows the neuron to learn with minimum loss function.

Modelling of the Employment Tracking System using Feed Forward Neural Network 

The ETS was developed using the model of the FFNN and the data model of the employee collected from the federal ministry of labour, employment and productivity. The FFNN model in Equation 10 was trained with the data collected using the back-propagation algorithm which ensured that the optimization problem and the error during the training process were reduced to the lowest minimum. The activity model of the FFNN is as presented inFigure 9, while that of ETS developed with it is as presented in Figure 10.

Figure 9: Activity model of the FFNN

Figure 9: is the FFNN model which shows the workflow of the neural network interconnected and activated for training using back-propagation algorithm. The training adjusts the weights of the neurons with respect to the bias to solve the optimization problem and achieve least error and receiver operator values which is approximately 1, before genera tin the classification model. The FFNN was used to train staff fingerprint data collected from the data model of all the staff to generate a model of ETS. The activity diagram of the ETS is as presented in the Figure 10.

Figure 10: Activity model of the ETS

The data was loaded into the FFNN and was split into training, test and validation sets. The training set was used to learn the neurons of the fingerprint patterns, the test set was used to test the training performance, and then the validation set was used to validate the results. The output of the training generated the ETS model which was used to develop the ETS for the detection of employed persons in Nigeria. The model of the ETS is as shown in Figure 11.

Figure 11: Activity diagram of the ETS

Figure 11 shows the application of the employment tacking model in the development of the ETS which was used to identify workers with the Federal Ministry of Labour, Employment and Productivity (FMLEP). The system was developed by loading test data into the ETS classification model developed with neural network and then classify already employed persons and detect the unemployed.

System Implementation

The system was implemented with Simulink. The unemployment prediction model was generated using regression application software in MATLAB while the ETS was generated using the neural network application software. The regression application software was loaded with the data and then trained using the linear regression model to generate the result of the unemployment prediction model which was exported and used for time series prediction of the NYSC corps members from the Nigerian approved HEIs.

System Result of the Prediction Model

The performance of the unemployment prediction model was evaluated using mean square error, absolute mean square error, regression and determinant coefficient. Figure 12 shows the performance result of the linear regression model which was used to train the NYSC data and determine the line of best fit which shows the perfect point of prediction.

Figure 12: Linear regression training result with the line of best fit

The result shows the relationship between the predicted values and true values using the line of best fit as modelled in equation 6. From the result shown, it was observed that the coefficient determinant  which indicated the probability of correct regression values is 0.88. The implication of the result is that the prediction model is good as it is approximately 1 which is the ideal   for a prediction model.This further shows that 88% of the predicted points fall within the line of best fit which is very good (Figure 13)

Figure 13. Result of the prediction model

As can be seen from Figure 13, the result of the prediction model generated from the training of the NYSC data. The model was generated using the relationship between the actual values in blue dots as the dependent variable (x) and the predicted values on yellow dot as the independent variables (y), based on the prediction model in Equation 2. The error which occurred as the difference between x and y as in equation4 is presented in Figure 14.

Figure 14: Prediction model with error

Figure 14 shows the prediction model with error which is the straight line that connects the actual value with the predicted values. The idea is to achieve a mean square error which is minimal or approximately zero. From the result shown, it was observed that the mean square error which occurred after the training and testing of the prediction model is 9.5777e+05. The implication of the result is that the error is minimal (tolerable) as it is approximately zero, thus showing good training performance. However, the RMSE and MAE do not agree with the overall result achieved in Figure 14 and here in Figure 15. From the two figures mentioned, RMSE is 978.06 which is very high and implied high level of overshot during the training of the linear regression model while the MAE is 490.48 which is very high as the absolute difference between the predicted and actual values. Hence, the linear regression model developed for the prediction of unemployment, despite the success is not reliable.

To solve this problem, since a FFNN model was already developed in the study for ETS, it was utilized and reconfigured to be compatible with the NYSC data collected using 296 input neurons in the input layers, and 301 hidden layers before the data were uploaded for training as shown in the neural network time series application software of Figure 15.

Figure 15: Result of the FFNN training for prediction model

Figure 15 shows the result of the FFNN when loaded with the NYSC data collected from 296 HEIs approved for NYSC. The data was divided by the neural network toolbox automatically into training, test and validation sets. The data in each of the sets were further divided into test and target values which were used to train the neurons. Testing the ability to reach the target values using back-propagation algorithm and the parameters. To evaluate the performance of the prediction model generated after training, the regression result in Figure 16 was used.

Figure 16: Regression result of the FFNN for unemployment prediction

The result shows the average overall regression values achieved from the test, training and validation sets as 0.99624. The implication of this is that 99.62% of the predicted data are within the line of best fit which is very good and indicates correct regression performance (Figure 17).

Figure 17: Result of the MSE performance

As can be seen from Figure 17, the MSE performance of the prediction model developed. The result shows that the average MSE performance of the FFNN training is 0.0025165Mu. The implication of this is that the difference between the predicted values and the actual values is approximately zero, which is very good and shows reliable unemployment prediction model. The RMSE was determined using the square root of the MSE and gave the result as 0.050165Mu which is also good, as it approximates zero.

CONCLUSION

In this research, a digital nervous system was successfully developed using machine learning technique to attempt a solution for the problem of unemployment in Nigeria. The system was developed using a model of unemployment prediction and Employment Tracking System (ETS). The idea was to estimate the rate of unemployment and then solve the problem using biometric technology. The study developed a model which could predict unemployment rate in time series using data of past discharged NYSC members and linear regression technique. However, from the evaluation of the model developed, the system lacks reliability due to high error rate observed between actual and predicted values. To solve this problem, the feed forward neural network was also used to develop a prediction model. This system developed was implemented with simulation and then tested. The result showed that the feed forward neural network-based model achieved 20.4% improvement when compared to the linear regression model. The number of unemployed persons predicted can therefore be used for effective planning in order to provide employment. The employment fraud perpetrated in recruitment process was addressed using ETS. The system used fingerprint to detect full time staff of the FMLEP and then prevented them from multiple job holdings, thereby creating opportunity for employment.

REFERENCES

1. Adesegun, O., Mathew, O., & Omotola, S. (2020). Modelling unemployment rates in Nigeria using time series approach. Asian Journal of Mathematical Sciences. https://ssrn.com/abstract=3813692
2. Ajah, A. I., &Ugah, J. O. (2013). Comparative analysis of software development methodologies. International Journal of Advanced Research in Computer Science and Software Engineering, 3(6), 655.
3. Ajah, I. A. (2013). Digital nervous system based credit worthiness system for Nigeria banks. Zenith International Journal of Multidisciplinary Research, 3(12), 1–22.
4. Barnichon, R., & Garda, P. (2016). Forecasting unemployment across countries: The ins and outs. European Economic Review, 84, 165–183. https://doi.org/10.1016/j.euroecorev.2015.10.006
5. Bello, T. (2003, September). Attacking unemployment hurdles in the fragile economies of the Sub-Saharan Africa: The experience of Nigeria. Paper presented at the Economics for the Future Conference, Cambridge, United Kingdom.
6. Didiharyono, D., &Syukri, M. (2020). Forecasting with ARIMA model in anticipating open unemployment rates in South Sulawesi. International Journal of Scientific & Technology Research, 9(3), ISSN 2277-8616.
7. Ebere Uzoka Chidi, E Anoliefo, C Udanor, AT Chijindu, LO Nwobodo (2025)” A Blind navigation guide model for obstacle avoidance using distance vision estimation based YOLO-V8n; Journal of the Nigerian Society of Physical Sciences, 2292-229; https://doi.org/10.46481/jnsps.2025.2292
8. Elijah, J., Mishra, A., Gana, U., Udo, M., &Aibinu, A. (2017). Staff monitoring system using biometric. International Journal of Engineering and Computer Science, 6(5), 21448–21458. https://doi.org/10.18535/ijecs/v6i5.43
9. Eneh, P., Eneh, I., Egoigwe, S., & Ebere, C. (2022). Deep artificial neural network based obstacle detection and avoidance for a non-holonomic mobile robot. International Research Journal of Applied Sciences, Engineering and Technology, 5(1), ISSN 1573-1405.
10. Etuk, A., &Onwuachu, C. (2016). An Android based employee tracking system. International Journal of Computer Applications, 153(3).
11. Harbor M.C, Eneh I.I., Ebere U.C. (2021). Nonlinear dynamic control of autonomous vehicle under slip using improved back-propagation algorithm. International Journal of Research and Innovation in Applied Science (IJRIAS); Vol. 6; Issue 9; https://rsisinternational.org/journals/ijrias/DigitalLibrary/volume-6-issue-9/62-68.pdf
12. Ituma, C., & Asogwa, T. C. (2018). The application of machine learning for digital recognition of identical twins to support global crime investigation. International Journal of Computer Science and Engineering, 4(12), 427–433.
13. Kavitha, S., Varuna, S., & Ramya, R. (2016). A comparative analysis on linear regression and support vector regression. In Online International Conference on Green Engineering and Technologies (IC-GET) (pp. 1–5).
14. Kekong P.E, Ajah I.A.,   Ebere U.C. (2019). Real-time drowsy driver monitoring and detection system using deep learning based behavioural approach. International Journal of Computer Sciences and Engineering 9 (1), 11-21; http://www.ijcseonline.isroset.org/pub_paper/2-IJCSE-08441-18.pdf
15. Mzachary, M. (2019). An attempt to find the right hidden layer size for your tabular learner. fast.ai Forums. https://forums.fast.ai/t/an-attempt-to-find-the-right-hidden-layer-size-for-your-tabular-learner/45714
16. Neter, J., Kutner, M. H., Nachtsheim, C. J., & Wasserman, W. (1996). Applied linear statistical models. IRWIN, The McGraw-Hill Companies, Inc.
17. Nguyen, P.-H., Tsai, J.-F., Kayral, I. E., & Lin, M.-H. (2021). Unemployment rates forecasting with grey-based models in the post-COVID-19 period: A case study from Vietnam. Sustainability, 13, 7879. https://doi.org/10.3390/su13147879
18. Olorunfemi, G. C. (2021). References addressing the state of youth unemployment in Nigeria. International Journal of Innovative Psychology & Social Development, 9(4), 102–113. ISSN: 2467-8546.
19. Roopa, H., & Asha, T. (2019). A linear model based on principal component analysis for disease prediction. IEEE Access, 7, 105314–105318.