J.

Electrical Systems 20-7s (2024): 744-754

1
Jerin Johnkutty Exploratory Data Analysis using Machine
2
Dr.Serajul learning – Behaviour Based Safety
Haque
3
Jerry Davis T

Abstract: - This study takes a gander at the usage of machine learning techniques for exploratory data examination (EDA) in the field of
behavior-based safety (BBS), with a particular focus on the examination of safety dimension datasets got from worker reviews drove in
industrial settings. We utilize a methodology that integrates an extent of visualization methods, statistical examinations, and parameter
evaluations to uncover complex encounters into safety perceptions and ways of behaving. We do this by utilizing the Python programming
language and its strong data investigation libraries, including MATplotlib, Seaborn, and Pandas. Our survey means to assist proof based
decision-making strategies, proactively distinguish potential for developing safety protocols, and develop a organizational culture that is
safety-centric driven by eagerly examining worker feedback and behavioral patterns. This research features the significant role of EDA and
machine learning in deciphering complex datasets, advancing substantial improvements in occupational safety, and putting a high priority
on workers' well-being in dynamic workplaces through synergistic cooperation between data analytics and domain expertise.

Keywords: Exploratory Data Analysis (EDA), Behaviour Based Safety (BBS), Machine Learning Techniques, Worker
Reviews, Occupational Safety, Data Analytics

Introduction

Rapid technological advancement and ideas like lean production in today's culture create brand-new, complex
hazards [1]. Consequently, businesses that want to operate successfully in a more sustainable way must take
proactive measures to reduce their negative effects on the economy, environment, and society[2, 3]. One factor
that can affect an organization's safety performance is the impact on society, the environment, and the economy
[4]. The definition of safety performance is the quality of safety-related work. Improving an organization's
safety performance can raise its resilience or robustness, lowering the chance of accidents. Poor safety
performance, on the other side, might raise the organization's vulnerability and hence the likelihood of accidents
[5]. Poor design, gaps in oversight, and unworkable processes are examples of latent circumstances that are
hypothesized to cause accidents in organizations. Employee attitudes, beliefs, perceptions, and values (safety
culture), the environment that impacts employees (working environment), and routines and procedures are all
examples of latent conditions (safety activities) [6, 7]. The work defines safety performance as the complete
performance in a safety culture, working environment, and safety activities,Workplace safety culture and
environment [8]. Statistics on occupational injuries are critical for determining how well workers are protected
from workplace hazards and dangers. Workplace safety and health are critical components of decent work [9,
10]. An occupational accident is defined as an unexpected and unplanned incident, including acts of violence,
which occurs during or in connection with work and causes personal harm, sickness, or death to one or more
workers [11]. A case of occupational injury is one worker who sustains an occupational injury as a result of a
single occupational mishap [12, 13]. An occupational injury can be deadly (as a result of an occupational
accident and death occurs within one year of the event) or non-fatal, resulting in missed work time.
Behaviour Based Safety (BBS) is a strategy that uses safety observations to inform management and employees
about the overall safety of the workplace [15]. BBS is designed to draw workers' attention to their own and their

1 Research Scholar, B.S. Abdur Rahman Crescent Institute of Science and Technology, Chennai
2
Assistant Professor, Dept. of Mechanical Engineering, B.S. Abdur Rahman Crescent Institute of Science and Technology, Chennai
3
Assistant Professor, Dept. of Mechanical Engineering, Government Engineering College Thrissur, Kerala

* [email protected] ,** [email protected]

***[email protected]

Corresponding Author - Dr.Serajul Haque

([email protected])

Copyright © JES 2024 on-line : journal.esrgroups.org

744
 J. Electrical Systems 20-7s (2024): 744-754

colleagues' regular safety behaviour. BBS program aims to increase the employee safety of the organization [16,
17]. Observers (workers trained to perform on-site safety checks)also conduct reviews of other employees with
an eye on their behaviour when implementing a BBS [18]. These observers document safe and dangerous
conduct as well as safe and unsafe workplace circumstances. The observer then informs the worker of his or her
observations and gives comments [19, 20]. Positive feedback is welcome. Discussing strategies for employees to
conduct their activities more safely helps workers and observers become more conscious of their actions. BBS
programmers are built on a continuous feedback loop in which employees and observers offer comments on
how to enhance safety to one another, and safety professionals use the data gathered during the observations to
continuously improve the BBS programme. Organizations that adopt a BBS programmer establish the proper list
of behaviours to watch depending on their organization's particular habits and hazards [21-23]. Safety
professionals often create a checklist style that is simple and quick for field observers to complete and outlines
the objective behaviours [24-25]. Each organization must have a unique strategy for installing BBS. Employees
are responsible for their own personal safety while working with BBS, according to the firm. For the BBS
deployment inside the organization to be effective, all workers must participate in the program [26-27]. Every
employee, regardless of their position in the organization's structure, must comprehend and implement the
benefits of BBS. Every employee, from the CEO to the front-line workers, must be included in the BBS
implementation process, including hourly, salaried, and union personnel. Change is unavoidable since the
organization's policies, methods, and/or systems will need to be modified in order to achieve the intended
behavioural changes. Any modification requires the participation of the whole crew [23, 29].

Literature Review

Occupational safety and health (OSH) remain critical concerns in various industries worldwide. Understanding
the dynamics of safety stressors, social support systems, and safety performance is crucial for ensuring worker
well-being and organizational productivity.Sampson, [21] explore the intricate relationship between safety
stressors, social support, and safety performance among unionized pipefitters. Their study, grounded in action
theory, highlights the nuanced impact of safety obstacles and uncertainty on safety compliance and
participation. While safety uncertainty and obstacles were negatively correlated with safety participation, safety
compliance was mainly affected by uncertainty. Manager support, especially positive job-related
communication, significantly improved safety performance, highlighting the significance of viable
communication channels in advancing safety culture. Interventions on the basis of (BBS) are a proactive method
for decreasing working environment injuries and accidents. A meta-analysis was done by [22] to evaluate the
viability of BBS interventions in various occupational settings. Their outcomes show a statistically significant
diminishing in accidents and injuries following the execution of BBS, despite methodological limitations. The
authors do, however, issue a warning against exaggerating these discoveries since some examination had subpar
procedure. Solid intervention plans that are adjusted to specific workplace prerequisites are supported, just like
the utilization of control groups in evaluations to expand their validity.

As the tasks of the construction industry are by their very nature complex, cautious planning and execution are
important to maximize efficiency and productivity. [23] use structural equation modeling (SEM) to dissect the
critical factors influencing the safety risk tolerance of construction workers. Their review features what risk
versatility is a perplexing construct that is impacted by a people perspectives, past experiences, traits of the job,
and safety techniques. Curiously, safety management is viewed as a central part influencing risk tolerance,
featuring the fundamental job that organizational safety culture plays in picking the mindsets of representatives
toward security. The pursuit for expanding construction productivity and efficiency has earned attention,
particularly in quickly arising economies like India. As per [24], labor deficiencies, unexpected disturbances,
material delays and plan revisions are a piece of the significant obstructions preventing construction efficiency.
To expand construction activities and shortening delays, their study underlines the significance of effective
project management methodology, brief material procurement, and proactive risk moderation techniques. Credit
risk the management is fundamental to safeguarding financial stability and decreasing the risk of loan in the
banking business. [28] investigate loan default patterns and credit risk in Ghanaian banks. Banks actually
struggle with high credit default rates even with an assortment of credit risk management techniques in light of
the CAMPARI model. The authors stress that to decrease credit risks and assurance financial supportability,

745
 J. Electrical Systems 20-7s (2024): 744-754

severe credit assessment practices, the formation of credit reference agencies, and further developed client
instruction programs are fundamental.

Exploratory data analysis (EDA) methods help identify patterns and generate hypotheses by providing insightful
information about intricate datasets. A thorough introduction to EDA principles and computational tools is given
by [29], who also emphasizes how complementary EDA is to confirmatory data analysis (CDA). Through the
development of a deeper understanding of data structures and underlying patterns, analytical model refinement
and the formulation of strong hypotheses are made possible by EDA.

RESEARCH GAP:

Author Proposed Methodology Results Research Gap

Action theory Identified negative relationship Lack of exploration into

Sampson et al. framework, survey between safety stressors and specific safety stressors'
method performance impact on performance

Need for methodologically

Meta-analysis of BBS Statistically significant reduction
Tuncel et al. robust studies in BBS
interventions in accidents/injuries
interventions

Behavioral safety Lack of detailed examination

Kabil & Improved safety behavior
partnership, survey on specific safety behavior
Sundararaju outcomes observed
method interventions

Structural equation Limited understanding of

External factors had larger
Wang et al. modeling (SEM), internal vs. external factors
impact on risk tolerance
questionnaire surveys influencing risk

Identified critical factors Need for empirical studies to

Subramani & Survey method, factor
affecting construction evaluate effectiveness of
Rajiv analysis
productivity interventions

Survey method, analysis Banks using varied risk Inefficacy of current risk
Ntow-Gyamfi
of credit risk management tools, default rates management tools to mitigate
& Boateng
management tools remain high loan defaults

Literature review, EDA complements CDA, Limited incorporation of EDA

Behrens contrasting EDA with emphasizes need for EDA techniques in statistical
CDA integration in data analysis training

The table summarizes various research approaches and results from multiple studies. The application of action
theory by Sampson et al. clarifies the complex relationship between safety stressors and performance,
emphasizing the need for more targeted research into the effects of particular stressors. The meta-analysis by
Tuncel et al. highlights that in order to validate the efficacy of Behavior-Based Safety (BBS) interventions,
methodological rigor is necessary. The structural equation modeling of Wang et al. highlights the limited
understanding of internal versus external risk factors and emphasizes the dominance of external factors in
influencing risk tolerance. Critical factors influencing construction productivity are identified by Subramani &
Rajiv's factor analysis, which calls for empirical research to assess the efficacy of interventions. Ntow-Gyamfi
& Boateng's analysis of credit risk management tools reveals the inefficacy of current practices in mitigating

746
 J. Electrical Systems 20-7s (2024): 744-754

loan defaults, indicating a need for improved risk management strategies. Behrens' review accentuates the
complementary nature of Exploratory Data Analysis (EDA) alongside Confirmatory Data Analysis (CDA),
advocating for the integration of EDA into statistical training and research methodologies to enhance data
interpretation and hypothesis development. Overall, these findings collectively stress the imperative for
methodological robustness and targeted investigations to address existing research gaps in safety, risk
management, productivity enhancement, and data analysis across various domains.

Proposed Methodology:

1. Data Acquisition and Preprocessing:

We commence our methodology by acquiring tabular data from safety dimensions review datasets. The dataset,
represented as D, comprises m samples and n features, where m=473 and n=38. Each sample pertains to worker
reviews on various safety dimensions. Upon loading the dataset into a Pandas DataFrame, denoted as X, we
preprocess the data to ensure its readiness for analysis. The preprocessing steps include handling missing values,
encoding categorical variables if any, and standardizing numerical features, if necessary.

2. Exploratory Data Analysis (EDA):

Under this phase, we embark on exploring the fundamental characteristics and relationships within the dataset.
Our exploratory analysis encompasses:

2.1. Data Exploration:

We employ descriptive statistics to summarize the central tendencies, dispersions, and distributions of the
dataset. The summary statistics include the mean (μ), standard deviation (σ), and quartiles of each feature.

2.2. Data Visualization:

Utilizing Python's visualization libraries, including Matplotlib and Seaborn, we create various visualizations
such as histograms, scatter plots, and box plots to elucidate the distributional properties and relationships
between features.

2.3. Correlation Analysis:

To unveil the interdependencies among safety dimensions, we compute the Pearson correlation coefficient (ρ)
between pairs of features. The correlation matrix C generated facilitates the identification of strong positive or
negative correlations between safety dimensions.

Figure 1: Machine Learning Process (Data analysis & Visualization techniques)

747
 J. Electrical Systems 20-7s (2024): 744-754

Figure 1 depicts a sequential flow of activities within a machine learning process. It begins with "Data
Acquisition," signifying the initial step of obtaining tabular data from relevant sources. The procedure then
switches to "Exploratory Data Analysis," in which the obtained data is scrutinized to determine its essential
traits and connections. After data exploration, "Data Visualization" techniques are used to produce a variety of
visual representations, including scatter plots and histograms, to clarify the relationships and distributional
properties of the features. Following an adequate measure of data exploration and visualization, the process
continues to "Model Building," where predictive models are fabricated utilizing the information accumulated
from the previous stages. In the "Model Evaluation" stage that follows model construction, the models'
performance is assessed utilizing measurements like mean squared error and coefficient of determination. In the
"Results Interpretation" stage, the results are interpreted, offering critical information about the models'
predictive capacity and highlighting important variables impacting the phenomena under study.

3. Modeling:

We move to model building to forecast the Safety Behavior dimension based on other safety dimensions after
gaining insights from EDA. The following steps are part of our modeling pipeline:

3.1. Feature Selection:

A subset of safety dimensions (Xselected) that are thought to be significant in predicting safety behavior are
chosen. The components of stress recognition, safety awareness, safety commitment, teamwork, and safety
compliance are incorporated in this subset.

3.2. Data Splitting:

Utilizing a 60-40 split, we partition the dataset into training (Xtrain) and test (Xtest) sets. While the test set
surveys the model's generalization performance, the training set makes model parameter assessment simpler.

3.3. Linear Regression Modeling:

To model the relationship between Safety Behavior and the chosen safety dimensions (Xselected), we use linear
regression. The following represents the linear regression model:

𝑦̂ = 𝛽0 + 𝛽1 𝑥1 + 𝛽2 𝑥2 + ⋯ + 𝛽𝑝 𝑥𝑝

Where xi are the chosen safety dimensions, y^is the expected safety behavior, β0 is the intercept, and the
coefficients are βi.

3.4. Model Evaluation:

Measurements like the coefficient of determination (R2) and mean squared error (MSE) on the test set (Xtest)
are utilized to assess the model's performance. These measurements measure how well the model predicts
results and makes sense of variety in safety behavior.

4. Results Interpretation:

The outcome of our examination shed light on significant factors affecting working environment safety
discernments and ways of behaving and offer insightful data about the predictive force of safety aspects on
safety behavior.

We understand the intricate connections among safety aspects thorough numerical and statistical investigations,
and we construct a predictive model that can direct designated interventions intended to advance a more secure
and better workplace.

RESULTS:

Linear Regression Modeling and Performance Evaluation:

748
 J. Electrical Systems 20-7s (2024): 744-754

The developed linear regression model, which forecasts safety behavior in light of the picked security
dimensions, displays favorable outcomes. Utilizing critical measurements on the test dataset, like mean squared
error (MSE) and coefficient of determination R2 , is vital for an exhaustive evaluation of the model's
effectiveness. While the MSE estimates the precision of the model's predictions, the R2 value is a strong
indicator that clarifies the percentage of changeability in safety behavior explained by the chose aspects.
Carefully laid out performance measurements for the model alongside the coefficients identified by the linear
regression model are reflected in the organized showcase given by Table 2.

Table 2: Linear Regression Model Performance and Coefficients

Model Performance Metrics

R2 0.76

60% of the data set is taken as training data set and the balance is test data set. From the training data set, we
kept the five dimensions (Safety Commitment, Safety compliance, Safety awareness, Teamwork, and Stress
Recognition) as independent and one dimension as dependent. We trained a linear regression model to predict
one dimension (Safety Behavior) of a 40% data set. The model predicted the output with an accuracy of 76%.
(R2 value 0.769). Important information about the operation and coefficients of the linear regression model used
to forecast safety behavior based on chosen dimensions is summarized in Table 2. The model performance
metrics show that the selected dimensions account for 76% of the variance in safety behavior, with a coefficient
of determination R2 value 0.76. The coefficients of the linear regression model further delineate the impact of
each safety dimension on safety behavior. This comprehensive overview aids in understanding the predictive
capacity of the model and underscores the relative importance of individual safety dimensions in shaping
workplace safety perceptions and behaviors.

Working With the Data Sets

The data we're using is from safety dimensions review data set. We will analyze the data and consider possible
options.

1. Import the Pandas libraries in the first step from Numpy Package.

2. Read the relatively large Safety Questionnaire CSV file as a data frame df (variable name). It displays
the data sets as rows and columns. There are 473 rows and 38 columns in our CSV file. To return the top 5 rows
of the data frame, we used the .head () method.

3. To learn more about the data frame, we used df.describe(). This will return the average, mean, standard
deviation, and so on for integer and float-type values in the data frame.

4. In the Next step organised the column names of questions from each dimension into a separate list.

5. We used df.corr () to find correlations between each question and created a heat map for each
correlation. Figure 4 depicts this pairwise correlation as a heat map.

749
 J. Electrical Systems 20-7s (2024): 744-754

Figure 4. Heat Map of all features

6. Then we obtained the correlation between the questions raised in each safety dimension. The plots of
each dimension are shown in Figure 5.

(a) (b)

(c) (d)

750
 J. Electrical Systems 20-7s (2024): 744-754

(e) (f)

(a). Safety Commitment (b) Safety Compliance (c) Safety Awareness (d) Safety behaviour (e) Stress
recognition (f) Team Work

Figure 5 Heat Map of various dimensions.

7. To conclude on most influenced questions, we have counted each rating with values greater than or
equal to 4. Then assigned it as a data frame with a variable name influ. Figure 6 shows the “for loop” used to
obtain the above.

Figure: 6. Features influences

751
 J. Electrical Systems 20-7s (2024): 744-754

8. We have used the maximum function method on the data frame influ to get the question number of
which factor is most rated/influenced by the workers.

9. Also, we have plotted the data frame to visualize the influence of each questionnaire as a bar chart.

Figure: 7. Influenced questions

Conclusion

In conclusion, this research paper delved into the application of Exploratory Data Analysis (EDA) and machine
learning techniques within the domain of Behavior Based Safety (BBS) to analyze safety dimension datasets
derived from worker reviews in industrial settings. We used a thorough methodology that included statistical
analyses, parameter evaluations, and visualization techniques to glean insights into safety perceptions and
behaviors. We did this by utilizing Python programming and reliable data analysis libraries. Our study sought to
identify opportunities for improving safety protocols, provide evidence-based decision-making processes, and
promote a safety-centric culture within organizations through the careful analysis of worker feedback and
behavioral trends. The outcomes demonstrated how well EDA uncovered the underlying features of safety
dimension datasets and how machine learning models could predict safety behavior based on specific
dimensions. The interrelationships between safety dimensions and their influence on perceptions of workplace
safety were better understood through the use of descriptive statistics, visual aids, and linear regression
modeling. The results underscore the critical function of data analytics in interpreting intricate datasets,
propelling concrete progress in occupational safety, and placing the welfare of employees at the forefront of
changing work settings. This study lays the groundwork for future efforts to use data-driven methods to address
new safety issues and foster an excellence in safety culture in a variety of industrial settings.

References

[1] Yu, Kai, Qinggui Cao, Changzhen Xie, Nannan Qu, and Lujie Zhou. "Analysis of intervention strategies for coal miners'
unsafe behaviours based on analytic network process and system dynamics." Safety Science 118 (2019): 145-157.

[2] Yu, Heyao, Jay Neal, Mary Dawson, and Juan M. Madera. "Implementation of behaviour-based training can improve
food service employees’ handwashing frequencies, duration, and effectiveness." Cornell Hospitality Quarterly 59, no.
1 (2018): 70-77.

[3] Houette, Beate, and Natascha Mueller-Hirth. "Practices, preferences, and understandings of rewarding to improve safety
in high-risk industries." Journal of safety research 80 (2022): 302-310.

752
 J. Electrical Systems 20-7s (2024): 744-754

[4] Soltanmohammadlou, Nazi, Sanaz Sadeghi, Carol KH Hon, and Fariba Mokhtarpour-Khanghah. "Real-time locating
systems and safety in construction sites: A literature review." Safety Science 117 (2019): 229-242.

[5] Fu, Hanliang, Gunasekaran Manogaran, Kuang Wu, Ming Cao, Song Jiang, and Aimin Yang. "Intelligent decision-
making of online shopping behaviour based on the internet of things." International Journal of Information
Management 50 (2020): 515-525.

[6] Kalteh, Haji Omid, Seyyed Bagher Mortazavi, Eesa Mohammadi, and Mahmood Salesi. "The relationship between safety
culture and safety climate and safety performance: a systematic review." International journal of occupational safety
and ergonomics 27, no. 1 (2021): 206-216.

[7] Zhou, Cheng, Hanbin Luo, Weili Fang, Ran Wei, and Lieyun Ding. "Cyber-physical-system-based safety monitoring for
blind hoisting with the internet of things: A case study." Automation in Construction 97 (2019): 138-150.

[8] Goh, Yang Miang, Chalani U. Ubeynarayana, Karen Le Xin Wong, and Brian HW Guo. "Factors influencing unsafe
behaviours: A supervised learning approach." Accident Analysis & Prevention 118 (2018): 77-85.

[9] Chen, Hainan, Xiaowei Luo, Zhuang Zheng, and Jinjing Ke. "A proactive workers' safety risk evaluation framework
based on position and posture data fusion." Automation in Construction 98 (2019): 275-288.

[10] Omidi, Leila, Seyed Abolfazl Zakerian, Jebraeil Nasl Saraji, Esmaeil Hadavandi, and Mir Saeed Yekaninejad.
"Prioritization of Human Factors Variables in the Management of Major Accident Hazards in Process Industries Using
Fuzzy AHP Approach." Health Scope 7, no. 4 (2018).

[11] Zhou, Tuqiang, and Junyi Zhang. "Analysis of commercial truck drivers’ potentially dangerous driving behaviours
based on 11-month digital tachograph data and multilevel modelling approach." Accident Analysis & Prevention 132
(2019): 105256.

[12] Xia, Nini, Mark A. Griffin, Xueqing Wang, Xing Liu, and Dan Wang. "Is there agreement between worker self and
supervisor assessment of worker safety performance? An examination in the construction industry." Journal of safety
research 65 (2018): 29-37.

[13] Fang, Weili, Lieyun Ding, Peter ED Love, Hanbin Luo, Heng Li, Feniosky Pena-Mora, Botao Zhong, and Cheng Zhou.
"Computer vision applications in construction safety assurance." Automation in Construction 110 (2020): 103013.

[14] Dong, Shuang, Heng Li, and Qin Yin. "Building information modelling in combination with real-time location systems
and sensors for safety performance enhancement." Safety Science 102 (2018): 226-237.

[15] Hakanen, J.J.; Perhoniemi, R.; Toppinen-Tanner, S. Positive gain spirals at work: From job resources to work
engagement, personal initiative and work-unit innovativeness. J. Vocat. Behav. 2008, 73, 78–91.

[16] Hinze, J., Hallowell, M., Baud, K., 2013a. Construction-safety best practices and relationships to safety performance. J.
Constr. Eng. Manage. 139(10), 04013006.

[17] Li, H., Lu, M., Hsu, S.C., Gray, M., Huang, T., 2015. Proactive behaviour-based safety management for construction
safety improvement. Saf. Sci. 75(6), 107–117.

[18] Martínez-Córcoles, M., Gracia, F., Tomás, I., Peiró, J.M., 2011. Leadership and employees’ perceived behaviours in a
nuclear power plant: a structural equation model. Saf. Sci. 49, 1118–1129.

[19] Nemanich, L.A.; Keller, R.T. Transformational leadership in an acquisition: A field study of employees. Leadersh. Q.
2007, 18, 49–68.

[20] Øyvind Dahl, Olsen, E., 2013. Safety compliance on offshore platforms: a multi-sample survey on the role of perceived
leadership involvement and work climate. Saf. Sci., 54(4), 17–26.

[21] Sampson, J.M., DeArmond, S., Chen, P.Y., 2014. Role of safety stressors and social support on safety performance.
Saf. Sci. 64(3), 137–145.

[22] Tuncel, S., Lotlikar, H., Salem, S., Daraiseh, N., 2006. Effectiveness of behaviour based safety interventions to reduce
accidents and injuries in workplaces: critical appraisal and meta-analysis. Theor. Issues Ergon. Sci. 7(3), 191–209.

[23] Wang, J., Zou, P.X.W, Li, P.P., 2016. Critical factors and paths influencing construction workers’ safety risk tolerances.
Accident Anal. Prev. 93, 267–279.

753
 J. Electrical Systems 20-7s (2024): 744-754

[24] Subramani, T., and S. R. Rajiv. "Improving construction efficiency and productivity of industry using
SPSS." International Journal of Application or Innovation in Engineering & Management (IJAIEM) 5, no. 5 (2016):
239-250.

[25] Kabil, GV Arockia, and V. Sundararaju. "Behaviour Based Safety in Workplace." International Journal of Research in
Engineering, Science and Management 2, no. 12 (2019): 327-333.

[26] Perera, H., & Costa, L, “Personality Classification Of Text Through Machine Learning And Deep Learning: A Review
(2023),” International Journal for Research in Advanced Computer Science and Engineering, 9(4), 6–12.
https://doi.org/10.53555/cse.v9i4.2266.

[27] Bogumil M. Konopka, Felicja Lwow, Magdalena Owczarz, Łukasz Łaczmański, “Exploratory Data Analysis of a
Clinical Study Group: Development of a Procedure for Exploring Multidimensional Data,” PLOS ONE, [Online]
https://www.ncbi.nlm.nih.gov/pmc/articles/ PMC6107146/pdf/pone.0201950.pdf, August23, 2018, pp. 1-21.

[28] Matthew Ntow-Gyamfi and Sarah Serwaa Boateng, “Credit Risk and Loan Default among Ghanaian Banks: An
Exploratory Study,” Management Science Letters, Vol. 3, 2013, pp.753–762.

[29] John T. Behrens, “Principles and Procedures of Exploratory Data Analysis,” Psychological Methods, 1997, Vol. 2, No.
2, pp.131-160.

754