PosterPDF Available

Behavioral Business Analytics Approach for Occupational Health & Safety

  • DNB Analytics (aka Decision & Behaviour Analytics)

Abstract and Figures

This research aims to develop a behavioral based and collaborative Business Analytic approach for investigating problems of Occupational Health and Safety (OHS) at a chemical manufacturing company located in Turkey.
Content may be subject to copyright.
Behavioral Business Analytics Approach for Occupational Health & Safety
1,2 Salih Tutun, 3Abdulaziz Ahmad, 2 Sedat Irgil, 2 Ilker Yesilkaya, 4Tülay Yorulmaz, 2 Oguzhan Kirkpinar
1 Binghamton University, 2 DNB Analytics, 3University of Minnesota Crookston, 4Literatur Chemical Industry
After preprocessing and analyzing the dataset, we understand A615:Insufficient Instructions, A616: Operator
Mistakes, A622:Lack of Information, A621:Education Deficiency, A618:Manager Mistakes, A619:Lack of attention,
A624:Behavioral and Cognitive Deficiencies, A611: No use of gloves, and A613: No use of work shoes are
significant and the company need to focus on these features to prevent future accidents. After getting relations
based on Eq. (2), we build the network (as seen in Figure 2) and found outliers (blue rectangles) and popular
behaviors (red circle). In this way, we defined around 30 outliers workers that have problems, and we defined
alpha/alphas of the group for managing the system. For example, Worker 54 is one of the leaders of the group in
the company. We can focus on him, and understand why others learn from him.
DNB Analytics developed behavioral BI to
understand workers. We also would like to thank
Psikomed and Literatur Chemical Industry for
helping us.
This research aims to develop a behavioral based and
collaborative Business Analytic approach for
investigating problems of Occupational Health and
Safety (OHS) at a chemical manufacturing company
located in Turkey. The collected dataset is composed of
270 workers, in which each worker has 630 features
(including psychological questions). Then, we utilized
the Heterogonous Similarity Function (HSF) to capture
the relationships among workers by looking at the
psychological behaviors of works as well as the
Company's structure. Afterward, based on the captured
relations, a network model is developed to analyze
workers’ personalities, and it explains the relationship
between the psychological status of the workers and
the reason for occupational accidents. Finally, we
defined around 30 outliers (workers) that have
problems for OHS, and we defined alpha/alphas leaders
of the group for managing the OHS system. We
evaluate the behavioristic patterns of these alphas and
outliers and then improve those behaviors for the
purpose of improving the OHS of workers
Creating systematic and comprehensive strategies to
prevent occupational hazards and risks is challenging.
International Labor Organization (ILO) stated that there
are about 2 million occupational fatal accidents
occurred across the world per year [1]. The overall
annual rate of Occupational accidents, fatal or non-
fatal accidents is estimated at 270 million [2]. After
tragedy happened in 1976, in a small village located in
Italy called "SEVESO", European zone has been
declared as "Seveso Directive" (82/501/EEC). Over the
past decades, this directive was developed in three
directions. The first one focused on controlling major
hazards for better OHS. While the second directive was
based on improving the occupational processes and
standardize with OHSAS 18001. The third directive was
based on the improvement of human behavior and
reliability and standardize with ISO 45001. Yet, despite
continuous improvements, occupational accidents, and
diseases are still too frequent and their cost in terms of
human suffering and economic burden continues to be
significant [3].
Many solutions focused on occupational processes and
standardization. However, human mistakes are the
most common reason of occupational accidents [3]. In
our research, we integrate network science and
machine learning to understand workers’ behaviors
and detect accidents before happening. This research
will focus on behaviors and eliminate problems for
workers. In this way, the company will increase
motivation, health, and safety.
OHS aims to decrease Occupational Hazards and
Risks based on regulations, standards, and
assessments. This research is done by subjective
inspections and analyze based on mostly
qualitative scoring. The primary conclusion of this
research is understanding the OHS interaction
between the company population and find out the
alpha/alphas of the group and outliers and
evaluate the behavioristic pattern of this alphas
and outliers and improve those behaviors for the
sake of the OHS based on understand relation of
causality about incidents happened in the that
occupational ecosystem.
[1] International Labour Organization, Fundamental Principle of
[2] Hämäläinen, P., Takala, J., & Saarela, K. L. (2006). Global
estimates of occupational accidents. Safety science, 44(2), 137-156.
[3] Hao, B., Li, L., Li, A., & Zhu, T. (2013, July). Predicting mental
health status on social media. In International Conference on Cross-
Cultural Design (pp. 101-110). Springer, Berlin, Heidelberg.
[4] Tutun, S., Khasawneh, M. T., & Zhuang, J. (2017). New
framework that uses patterns and relations to understand terrorist
behaviors. Expert Systems with Applications, 78, 358-375.
[5] Tutun, S., Ahmed, AA., Irgil, S., Yesilkaya, I., Khasawneh, M.
(2019) Detecting Psychological Symptom Patterns Using
Regularized Multinomial Logistic Regression, the 2019 IISE Annual
Conference, Orlando, Fl, USA.
This research was conducted at the Literatur Chemical Industry company, which is located in Turkey. At the
company, workers learn success and mistakes by looking at each other. Therefore, we need to understand how
workers interact with each other, and we need to find popular behaviors and outliers by looking for the
companys structure and psychological answers. For better analysis, Step 1: we collected a dataset of 270
workers and 630 features (including psychological attributes) for each worker. Step 2: we defined the important
features to focus on them as a strategy by using correlations between features and accidents feature. Step 3:
The heterogonous similarity function approach (as seen in Eq.(1)) is utilized to capture relations of workers for
understanding outliers and popularity in the company. If workers are very similar and they have interactions
well, we are finding and defining there is a popularity among workers. For the outliers, if they have non-similar
behaviors in the building network, it means that they did not share their information. As seen in Eq.(1), the
dataset has n=worker, N=270, and xNd=630, and we would like to understand the relations of workers (n). To
calculate relations, we are looking at how they are similar (and interact with each other). In the dataset, for each
worker, we are comparing by looking at the features (F1, F2, F3, Fd). As seen in Eq.(2), If the feature is
categorical, we are looking that features are the same (it is 1) or not (0), and calculating bh value. If the feature is
continuous (like age), we are calculation the rate (Xh/Yh) for understanding how workers interact with each
other. Therefore, we calculate the similarity rate S(X, Y) between worker X and worker Y [4]. Step 4: we build the
network to show popular behaviors and outliers in the company.
Figure 1: Correlation of features and significance features Figure 2: Networks for understanding behaviors of workers.
Equation 2: Similarity function for capturing relations [5].
Therefore, for these 30 people, we will look at all
significance features as A615, A616, A622, A621,
A618, A619, A624, A611, and A613. The company
needs to give training for solving the lack of
information, education deficiency, manager
mistakes, and insufficient instructions. At the same
time, for solving a lack of attention and behavioral
and cognitive deficiencies, we sent outliers to
Psikomed Clinique to cure and solve their problems
[5]. After completing all the issues in the company,
outliers will be in a popular group, and we will
increase motivation and productivity. Therefore, we
will prevent future accidents in the company.
Equation 1: Structure of the dataset [5].
Methods & Materials
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.