SEMESTER

PROJECT
Cem Çağrı
Dönmez
EM7198.9.0

Eren ERDEM
924820267

DATA SCIENCE IN DECISION MAKING

PROCESSES
Dr. Cem Çağrı Dönmez
 Data Science in Decision-Making Processes

Data Science in Decision-Making Processes..............................................................................3

Introduction...........................................................................................................................4
1.What is a Decision?.........................................................................................................4
1.1 Framing........................................................................................................................5
1.2 Gathering.....................................................................................................................5
1.3 Deciding.......................................................................................................................6
1.4 Learning.......................................................................................................................6
2. What is Data?.....................................................................................................................6
3. What is Data Science?........................................................................................................7
3.1 Evolution of Data Science............................................................................................8
4. Machine Learning Based Analytics.....................................................................................9
4.1 Cohort Analysis............................................................................................................9
4.2 Regression Analysis....................................................................................................10
4.3 Classification Analysis................................................................................................10
4.4 Cluster Analysis..........................................................................................................10
4.5 Association Rule Analysis...........................................................................................11
4.6 Log Analysis................................................................................................................11
4.7 Neural Networks and Deep Learning Analysis...........................................................11
5. Data Terms and Definitions..............................................................................................12
6. Artificial Intelligence........................................................................................................13
6.1 Using Artificial Intelligence for Decision Making........................................................13
6.2 Human vs AI Based Decision Making.........................................................................14
6.3 AI from human...........................................................................................................15
7. Decision Making Response..............................................................................................16
7.1 Repeatability..............................................................................................................16
8. Hybrid Decision-Making Models......................................................................................16
8.1 Collective Human Artificial Intelligence Decision Making Structures.........................17
8.2 The Data Science Life Cycle........................................................................................19
9. Big Data & Decision-making.............................................................................................19
9.1 Dissemination............................................................................................................20
9.2 Five Vs of Big Data......................................................................................................21
9.3 Big Data Analytics.......................................................................................................22
9.4 Text Analysis..............................................................................................................23
9.5 Audio and Video Analysis...........................................................................................23
9.6 Internet analysis.........................................................................................................23
9.7 Network Analytics......................................................................................................23
10. Role of Big Data in Decision Making...............................................................................23
10.1 Benefits of Decision Making with Data Analytics.....................................................24

1
 Data Science in Decision-Making Processes

10.2 Opponent Intelligence result...................................................................................25

10.3 Time & Cost Saving..................................................................................................25
10.4 Optimization and Simulations of Supply Chains.......................................................25
10.5 Process of Decision making with Big Data Analytics................................................26
Conclusion...........................................................................................................................28

2
 Data Science in Decision-Making Processes

Data Science in Decision-Making Processes

Submitted 30/01/2021 by Eren ERDEM

Abstract:

Information is an important success factor that affects the performance of decision makers, especially
the quality of their decisions. These days, there is a huge amount of data available for organizations to analyze.
Data is considered the raw material of this century and abundance is assumed with today's 15 billion internet-
connected devices. Accordingly, by using big data analytics, solution manufacturers can gain valuable insights
from rapidly changing data of high volume, velocity, diversity, accuracy and value. This article aims to explain
how big data analytics can be integrated into the decision-making process. Accordingly, Big Data Analytics and
decisions framework was developed to match big data tools and analytics with different decision-making
stages, using design science methodology.

I have included many definitions and meanings. My aim was to examine and understand the types and
methods of decision making with today's conditions. In the new century, companies benefit from data in many
ways. The first of these is using data analysis and visualization for management and decision making, using
artificial intelligence for production and planning, using machine learning for repeatable and routine works. As
a matter of fact, all of these are to make our business life easier and to reach the right decisions with more
analytical methods. We do not need to go far, I use data analytics, data, outputs and visuals when making
operational decisions in my own company, and we call this decision-making using data science. Or, determining
the next song from our routine listening habits using a machine learning method while listening to Spotify is an
example of data science and machine learning. Finally, robot vacuums, which have just entered our lives, are an
excellent example of artificial intelligence, of course compared to today.

3
 Data Science in Decision-Making Processes

Introduction
The most important element in the decision-making process is to have the right
information at the right time. The main problem in recent years is not access to information,
but finding the key parts of the right decision in this flood of information. Information
technologies and people who know how to use them can help. The size and speed of data
generation is so high that traditional database and information network technologies cannot
properly manage and process data. Today's world cannot function without information and
communication technologies because the total amount of data in the world is expected to
increase over the years.

1.What is a Decision?
We define the word “decision” to mean any choice between options by any entity. In
this terminology, adding a tag to a user's photo such as cat and not cat is a decision driven by
a computer system, while deciding whether to start that system is made by the human leader.
From my perspective, the “decision maker” is the person responsible for decision architecture
and context framing, not that stakeholder or investor trying to veto the machinations of the
project team. In other words, it is the creator of meticulously articulated goals as opposed to
its destroyers. But what is decision-making? Decision making is a word used differently by
different disciplines, so it may refer to; taking action when alternative options are available (in
this sense, it is possible to speak of decision making by a computer or a lizard).
Fulfilling the function of a (human) decision maker, part of which is to take
responsibility for decisions. Even if a computer system could implement a decision, it would
not be called a decision maker because it is not responsible for its outputs, that responsibility
rests entirely on the shoulders of the people who created it. Not all outputs/recommendations
are decisions. In decision analysis terminology, a decision is made only when an irreversible
resource allocation has occurred. As long as you can change your mind, no decision has been
made yet. Therefore, there is a difference between making calculations and making decisions.
You can make as many calculations as you want, but you only have to decide once.
The disciplines that make up the qualitative side are traditionally called decision
sciences. Decision sciences deal with questions such as:
“How should you set decision criteria and design your metrics?” (All)
"Is the metric you selected incentive compatible?" (Economy)
“With what quality should you make this decision and how much should you pay for
excellent information?” (Decision analysis)
"How do emotions, heuristics, and biases play a role in decision making?"
(Psychology)
“How do biological factors such as cortisol levels affect decision making?”
(Neuroeconomics)
“How does changing the presentation of information affect selection behavior?”
(Behavioral Economics)
“How do you optimize your results when making decisions in a group context?”
(Experimental Game Theory)
“How do you balance multiple constraints and multi-stage goals when designing the
decision context?” (Design)
“Who will experience the consequences of the decision and how will various groups
perceive this experience?” (UX Research)
“Is the decision objective ethics?” (Philosophy)

4
 Data Science in Decision-Making Processes

This is just a small example of course. Many more questions can be asked. This is also
far from a complete list of related disciplines. Think of the decision science side as dealing
with decision setup and information processing in the more fuzzy form of storage (the human
brain) rather than the neatly typed kind of semi-permanent storage (on paper or electronically)
we call data.

1.1 Framing
It is necessary to distinguish between decision frames and thinking frames. Decision
frames describe actions, possibilities, and outcomes as perceived by the decision maker.
Thinking frameworks are related to deeper cognitive structures such as knowledge bases,
scenarios, schemas, cognitive maps, and inference mechanisms that shape the decision
framework. Key aspects of the decision framework are boundaries (eg, region, time and
market coverage), benchmarks (eg required rates of return, performance benchmarks, relevant
competitors) and metrics (eg return on investment, market share, and metrics). product
quality). Many companies use their own or close competitors' past performance as a relevant
reference point to evaluate their success. This kind of nearsighted framing plagued the UK
industry as well as Detroit automakers in the 1970s. A more nuanced framing issue in new
technology decisions concerns the 'do not invest' option, which often assumes a continuation
of current trends as a reference point. But this static view ignores the actions of rivals that
would likely erode the status quo. As well as shifting the metaphor (for example, towards
biological evolution), game theory can help challenge such myopic frameworks. The failure
to embrace the portfolio perspective is another notable framing bias of the behavioral decision
literature. Each decision is handled separately from the others.

1.2 Gathering
The primary biases at this stage are a tendency towards Overconfidence, reliance on
flawed heuristics in prediction, and a preference for confirmation over disapproval of
evidence. Overconfidence or arrogance reflects poor secondary knowledge, that is, not
knowing that we don't know. This can be improved, in part, through repeated feedback or
attempts to challenge core premises through reasoning, fault trees, or scenario building. The
overconfidence bias complicates decisions, especially when there is little data and where
judgment must necessarily play an important role. The key is knowing when not to trust their
intuition and how to bring basic assumptions to the surface, especially in small groups.
5
 Data Science in Decision-Making Processes

1.3 Deciding
In particular, the factors that complicate strategic choices, people's natural aversion to
uncertainty, rational choice models, uncertainty should not be important. Uncertainty, and
even uncertainty, is ignored by integrating it over a putative subjective probability distribution
defined over the target probability. Behaviorally, however, people tend to prefer a known
probability over an unknown probability with equal mathematical expectation. Therefore,
projects with a high level of uncertainty stemming from technological or market uncertainties
are likely to be systematically undervalued in people's informal screening of projects. By
definition, it is difficult to predict objectively.

1.4 Learning
Various barriers prevent learning from experience. They range from rationalization
and ego defenses to incomplete or confusing feedback. Because organizations can make only
a small number of truly strategic decisions within any generation of management, they rarely
face the problem of a lack of independence in feedback and possibly results. This suggests
that outcome feedback will be noisy and limited, and the emphasis should shift to process
feedback. This includes premises, datasets, selection procedures, incentive alignments,
implementation, etc. requires an examination of how the decision was reached. While autopsy
is a great way to learn from mistakes, the ultimate goal is to translate lessons learned the hard
way before death. For example, learning about new technologies may require a new
organizational unit separate from mainstream business or technology. IBM took this path
when developing its computer, for example, as General Motors (GM) did for the Saturn
project. Such a distinction is a way of resolving the inherent conflict between performance
and learning cultures in organizations (Senge, 1990). To optimize performance over the next
few periods, the firm should use what it knows best. To maximize its long-term survival, the
firm must expand its capabilities through exploration. Long-term success may require short-
term sacrifices.

2. What is Data?
Without going too far into vague etymological details, the word 'data' is the plural
form of the Latin word 'data' meaning 'given'. It is also derived from the Latin verb 'courage'
meaning 'to give'. In essence, data is a multitude of observations; data is given. Data existed
before computers. For example, when scientists observed the world, they first collected data
with thoughtful approaches and then drew conclusions from them after extensive data
analysis. From the very beginning, data was critical for scientists to make objective
observations, and the purpose of data collection was to draw unbiased information, hence
conclusions. Meanwhile, details on how data were collected, analyzed, sampled and
calculated are presented in the final published research paper. In other words, the research
paper itself detailed data that was meaningful and understandable to anyone reading the
article. Here, data about data is what we call meta-data today, and the ability of scientists to
read, understand, create and communicate data is data literacy.
By definition, the data itself must reflect a fact, or a set of facts, that is its most
important real-world property. Data can be collected in different ways and stored in various
formats, and the most appropriate way depends on usability and efficiency. A fact can have
many properties. While collecting all the details takes much more resources and time, a subset

6
 Data Science in Decision-Making Processes

may not be enough in the end. Efficiency also depends on data format and structure. Putting
data in a numeric form requires the least amount of storage and enables fast data processing.
We live in an age of data science and advanced analytics. Considering that almost
everything in our daily life is digital, almost our every move is recorded as data. Thus, the
current electronic world; It consists of structured, semi-structured or unstructured financial
data, health data, species data, internet data, and cybersecurity data. Data science is typically
"the set of methods used to understand and analyze statistics, data analysis, and related real
situations. Data science is the science that studies systems such as decision-making, thinking,
and models in situations where data can be given and manipulated. The popularity of data
science is increasing day by day." You can see the trend changes in google data as shown in
Figure 1. As we understand, the trend towards topics such as "Big data", "Machine learning"
has increased in recent years. This statistical information is sufficient to motivate us,
especially in real-world applications where data-driven smart decision making can be applied.
Generally, data science is the science that uses advanced analytical methods to make
useful decisions for our business. By advanced analytical methods, it is a pattern used to
determine what may occur in the future. While basic analytics generally provides a
description of the data, advanced analytics are steps taken to provide deeper solutions and
help analyze granular data. Popular among the various types of analytics, "Descriptive
analytics" answers what it is, "Diagnostic analytics" explains why the event occurred,
"Predictive analytics" what might happen in the future, Prescriptive analytics" which
prescribes what action should be taken, discussed briefly in “Advanced analytics methods and
smart computing”.

Figure 1

3. What is Data Science?

Data science has received widespread attention in academic and industrial circles. new
data. Science research institutes and organizations such as Columbia University Institute of
Data Science and Engineering and New York University Data Science Center. University of
California at Berkeley, Columbia University and other universities have launched data science
courses and degree programs. Cleveland and Smith suggested that data science should be seen
as an independent discipline. Facebook, Google, EMC, IBM and other companies created jobs
positions for data scientists. According to the Harvard Business Review, the data scientist "
The sexiest job of the 21st century.” Currently, there are several viewpoints on the definition.

7
 Data Science in Decision-Making Processes

1
However, there is no consensus definition. As we believe, new science, data science research
goals are different from others, more established disciplines. Also, scientific topics that data
science addresses include: not studied by the natural or social sciences.
The basic ideas underlying the definitions given above are that data science is used to
gain knowledge. Obtaining information from data in some related fields and providing
support to current scientific research and management decision-making scheme. However, all
the studies described above are still it is not enough to establish data science as a new, unique
science. This is because, the objects of his work are things in the natural world, and his
research topics are also covered in current scientific fields. With the development of digital
equipment, things in the natural world are increasingly stored in the form of data in
cyberspace. Data is entered, generated and created in cyberspace in various ways and became
more and more diverse, complex and inhuman. control. More and more data is unknown or
not fully understood by humans. data in cyberspace already shows features of an independent
world like the natural world, so all data cyberspace is referred to here as data nature. It should
be noted that there are two types of data in cyberspace. The first of these is the data represents
things in the natural world, here called real data. An example is personal information, Data
representing personal characteristics. The second is non-representative data things in the
natural world, here called virtual data. Virtual data, such examples data has no reference in
the natural world. An example is computer viruses, which are neither data representation of
viruses in the natural world or real viruses; instead, they exist only in cyberspace

3.1 Evolution of Data Science

The term "data science" - and the application itself - has evolved over the years. In
recent years, its popularity has grown significantly due to innovations in data collection,
technology, and serial data generation around the world. Gone are the days when those
working with data had to rely on expensive programs and mainframes. The proliferation of
programming languages like Python and procedures for collecting, analyzing and interpreting
data paved the way for data science to become the popular field it is today.
Data science started in statistics. Part of the evolution of data science was the
incorporation of concepts such as machine learning, artificial intelligence, and the internet of
things. With the incoming flood of new information and businesses looking for new ways to
increase profits and make better decisions, data science has begun to expand into other fields
including medicine, engineering and more. Statistics and the use of statistical models are
deeply rooted in Data Science. Data Science started with statistics and has evolved to include
concepts/applications such as Artificial Intelligence, Machine Learning and Internet of
Things, to name a few. As more and more data becomes available, businesses collect and
store this data in ever greater quantities, first through recorded shopping behaviors and trends.
With the growth of the Internet, the Internet of Things, and the exponential growth of the
volumes of data available to businesses, there has been a flood of new information or Big
Data. As the doors were opened by businesses looking to increase profits and enable better
decision making, the use of Big Data began to be applied to other fields such as medicine,
engineering and social sciences. We can say that data science was born from the idea of
combining applied statistics with computer science. The resulting workspace will harness the
extraordinary power of modern computing. Scientists realized that they could not only collect
data and solve statistical problems, but also use that data to solve real-world problems and
make reliable fact-based predictions.
In 1962 American mathematician John W. Tukey first expressed his dream of data
science. In his now-famous article "The Future of Data Analytics," he predicted that a new
1
(Sarker, 2021)

8
 Data Science in Decision-Making Processes

field would inevitably emerge nearly two decades before the first personal computers. While
Tukey was ahead of his time, he was not alone in his early appreciation of what would
become known as "data science." Another early figure was Peter Naur, a Danish computer
engineer with his book Concise Survey of Computer Methods, which offered one of the first
definitions of data science: "The science of dealing with data, once created, is transferred to
other fields and sciences as the relation of data to what they represent." The theories and
predictions of "preliminary" data scientists like Tukey and Naur became more tangible in
1977 with the founding of the International Association for Statistical Computing (IASC),
whose purpose was to connect traditional statistical methodology, modern computer
technology, and knowledge.
According to the Global Data Protection Index 2020 Snapshot survey published by
Dell Technologies, the amount of data companies manage globally has increased by 40
percent over the past year. Compared to 2016, this is an 830 percent jump. The coronavirus
pandemic has caused this rate to accelerate further since the first months of 2020. The
transition of most companies to remote work, the transformation of the data mix created into a
richer dataset that includes video communication and a tangible increase in workforce. the
consumption of downloaded and streamed video all contributed to this. The total amount of
global data is constantly growing. IDC's Global Data Sphere predicts that more than 79
zettabytes (79 trillion gigabytes) of data will be created, captured, copied and consumed
globally in 2021, with more data created over the next three years than in the past. The
statistics and forecasts presented clearly show that we are currently experiencing a digital
transformation with significant social and economic consequences. Therefore, advanced tools
are essential for analyzing large datasets. A branch of science called data science (DS) has
also emerged. It uses scientific methods, processes, algorithms and systems to extract
knowledge and insights from structured and unstructured data. By shaping and filtering the
collected data, it enables people to make more informed choices than in the past. DS is an
interdisciplinary field focused on extracting information from often large datasets and
applying knowledge and actionable insights from data to decision-making processes in a wide
variety of fields.

4. Machine Learning Based Analytics

In this section, we briefly discuss various advanced analytics methods based on
machine learning modeling, can make the computing process smart through smart decision
making in a business process. The figure shows a general structure of a machine learning-
based prediction
modeling, taking into account both the training and testing phase. Below we discuss a wide
variety of methods such as: regression and classification analysis, association rule analysis,
time series analysis, behavior analysis, log analysis, etc. within the scope of our study.

4.1 Cohort Analysis

Behavioral analytics is a new trend that is typically emerging. New information about
e-commerce sites, online games, mobile devices and smartphone applications, IoT user
behavior and many more. Behavioral analysis aims to understand how. and why consumers or
users are behaving, right predictions of how they will behave in the future. For example, it
allows advertisers to make the best bids. The right customer segments at the right time.
Behavior analytics, including traffic data such as navigation paths, clicks, social media
interactions, purchasing decisions and marketing responsive, use large amounts of raw user

9
 Data Science in Decision-Making Processes

events information collected during sessions that people use apps, games or websites. In the
real-world scenario, behavioral analytics is usually used in e-commerce, social media, call
centers, billing systems, IoT systems, political campaigns and other applications, to find
optimization opportunities to achieve specific goals results. Cohort analysis is a branch of
behavioral analysis. Analytics, which involves examining groups of people over time to see
how their behavior has changed. For example, it takes data from a specific dataset (for
example, an e-commerce website, web application or online game) and divides it into related
sections. groups for analysis. Various machine learning techniques such as behavioral data
clustering, behavioral decision tree classification, behavioral association rules, etc. can be
used in areas according to the purpose.

4.2 Regression Analysis

Regression analysis is typically done for either purposes: to estimate the value of the
dependent variable status of individuals with some relevant information can be used for
explanatory variables or to predict the effect of some explanatory variables on the dependent-
variable, that is, find the causal effect relationship are among the variables. Linear regression
cannot be used to: fits nonlinear data and can cause underfitting. In in this case, polynomial
regression performs better, but increases model complexity. regularization techniques Like
Ridge, Lasso, Elastic-Net can be used to optimize the linear regression model. Outside,
support vector regression, decision tree regression, random forest regression techniques can
be used for building effective regression models depending on the problem type, for example
non-linear tasks. financial forecast or forecast, cost estimation, trend analysis, marketing, time
series prediction, drug response modeling, etc. are some examples where regression models
can be used to solve real world problems in data science and analytics.
4.3 Classification Analysis
The classification is one of the most widely used and best known. data science processes. This
is a form of supervised machine learning approach that also refers to a predictive modeling
the problem where a class label is estimated for a given instance. Spam identification such as
"spam" and "not spam" In e-mail service providers, an example of classification might be
problem. There are various forms of classification analysis. such as the binary classification
available in the field expresses an estimate of one of the two classes; multiclass classification
involving more than one's guess more than two classes; multi-label classification is a
generalization multi-class classification in which the classes of the problem are arranged
hierarchically. A few popular classification techniques, such as k-closest neighbors, support
vector machines, navies Bayesian, adaptive reinforcement, extreme gradient boosting, logistic
regression, decision trees, and random forests are available to solve classification. problems.
tree-based classification technique, e.g. random forest that considers multiple decision trees,
outperforms others for solving real-world problems due to its ability to generate logic in most
cases.
4.4 Cluster Analysis
Clustering is a form of unsupervised machine learning technique and well known in
many data science applications statistical data analysis fields. Generally, clustering techniques
look for structures within a dataset and, if classification not defined before, classify as
homogeneous case groups. This is the data points same and different within a set data points
in another set. In general, the purpose of the cluster analysis is to group various data points
into groups (or clusters) internally homogeneous and heterogeneous from outside. To get an
idea of how data is distributed on a given dataset or as a pre-processing step other algorithm,

10
 Data Science in Decision-Making Processes

clustering are often used. data clustering, for example, it helps customer shopping behavior,
sales. campaigns and consumer retention for retail businesses, anomaly detection, etc.
Recently, some articles proposed a hierarchical clustering method, based on the bottom-up
agglomerative technique to capture similar behavioral characteristics of the user over time.
Key the benefit of agglomerative hierarchical clustering is tree structure hierarchy created by
agglomerative clustering is more informative than unstructured flat sets, which can assist in
better decision making in the relevant application fields in data science.

4.5 Association Rule Analysis

Association rule learning is known as rule-based machine learning system, an
unsupervised learning method typically It is used to establish relationships between variables.
It a descriptive technique often used to analyze large datasets to discover interesting
relationships or patterns. This is the main strength of the association learning technique is that
it is comprehensive for producing all the associations that meet user-specified restrictions,
including minimal support, and trust value. Association rules allow a data scientist to identify
trends, relationships and co-occurrences between datasets in large data collections. For
example, in a supermarket, if associations get information about purchasing behavior
consumers for different items that help change marketing and sales plan. In health, for better
diagnosis patients, doctors can use the association guidelines. Doctors can assess the
conditional probability of a particular disease by comparing symptom associations in previous
data machine learning-based cases using association rules data analysis. Similarly, association
rules are useful to consumers. behavior analysis and forecasting, customer market analysis,
bioinformatics, blog mining, advice systems etc. Several types of association rules have been
proposed. frequent pattern based, logic based, tree based, fuzzy rules, belief rule etc. area like.
A priori is the most widely used. Algorithm to discover association rules from a given data.
Dataset among association rule learning techniques. Recent developments can yield important
results in terms of production. non-redundant rules that can be used for smart decision made
according to human preferences within the field data science applications.

4.6 Log Analysis

Logs are widely used daily in system administration it is often the only data available
to record the detailed system runtime activities or behaviors in production. Record analysis
can therefore be considered a method of analysis, capable of interpreting and understanding
computer-generated logs or messages, also known as logs. Its device log, server log, system
log, network log, event log, audit trail, audit record, etc. This type of creation process records
are called data records. Various techniques such as classification and labeling, correlation
analysis, pattern recognition methods, anomaly detection methods can be used for machine
learning modeling, effective log analysis. Log analysis can help in accordance with safety
policies and industry regulations, as well as providing a better user experience by promoting
Troubleshooting and identifying technical problems areas where productivity can be
increased. For example, web servers use log files to record data about website visitors.
Windows event log analysis can help a researcher draw a timeline based on the log
information and discovered artifacts. In general, advanced analytical methods can play
machine learning, taking into account modeling an important role in extracting insightful
patterns from these diaries Data that can be used to automatically and intelligently generate
applications and therefore can be considered an important study in the field of data science.
2

2
(Filip Björkman, Sebastian Franco, 2017)

11
 Data Science in Decision-Making Processes

4.7 Neural Networks and Deep Learning Analysis

Deep learning is a form of machine learning that uses machine learning. Neural
networks to build a computational architecture learning from data by combining multiple
processing layers, such as input, hidden and output layers. This main benefit of deep learning
over traditional machine learning methods is that it performs better in various fields.
situations, especially when learning from large datasets. The most common deep learning
algorithms are: multi-layer perceptron (MLP), convolutional neural network (CNN or
ConvNet), repetitive long short-term memory plexus. Figure 2 shows a structure modeling a
multiple neural network. processing layers Back propagation technique, used to set weight
values internally while building model. Convolutional neural networks (CNNs) improve the
design of traditional artificial neural networks (ANN), including convolutional layers, pooling
layers and fully connected layers. widely used various fields, including natural language
processing speech recognition, image processing and other auto-related data because it makes
use of two-dimensional data. (2D) structure of input data. AlexNet, Xception, Inception,
Visual Geometry Group (VGG), ResNet etc. Other advanced deep learning models It is also
used in the field based on CNN. In addition to CNN, recurrent neural network (RNN)
architecture is another popular method used in deep learning. Long short-term memory
(LSTM) is a popular genre, widely used recurrent neural network architecture deep learning
domain. Unlike traditional feedforward neural networks have LSTM feedback links. Thus,
LSTM networks are well suited for sequential analysis and learning. Data such as classifying,
sorting and predicting data based on time series data. Therefore, the data is LSTM, time,
sentence etc. in sequential format like can be used and is widely used in time series fields.
analysis, natural language processing, speech recognition, etc.

Figure 2 A structure of an artificial neural network modeling with multiple processing layers

12
 Data Science in Decision-Making Processes

5. Data Terms and Definitions

There are a number of key terms in data science. Analytics, data mining, data
analytics, big data, data science, advanced analytics, machine learning, and deep learning are
interrelated terms that are just as confusing. I will try to explain them to you as much as
possible.
The term data analysis refers to the processing of data. In other words, data analytics
talks about meaningful data, knowledge extraction and knowledge discovery with few other
terms like knowledge mining. Data mining is another popular term and can be expressed with
words such as information extraction, data extraction. In summary, data mining is processing
big data and extracting interesting results and meanings from them. Big data, on the other
hand, is the whole of unprocessed data that we cannot yet make sense of or lead us to a
decision by combining complex and irrelevant data. Big data can be generated by social
networks, mobile phones or other smart devices. For example, the smart watches we carry in
recent years or the smartphones that have been with us for years transmit our health, sleep
patterns, daily habits and routines to the big data network. But they don't make much sense
before they're processed.

6. Artificial Intelligence
The rise of Artificial Intelligence in recent years has raised many controversial issues.
notes. For example, IBM CEO Ginni Rometty “Technologies that increase human
intelligence. We see a world where this is a world in general. Partnership between man and
machine and that will actually make us better and allow us to do the best that the human
condition can do. Stephen Hawking, on the other hand, “Development of full AI could spell
the end of humanity. These very divergent views of leading experts warrant further research
into how humans became human, how assets can coexist with artificial intelligence and how
the negative impact of technology can be minimized. There is no widely accepted definition
of AI. It is normally referred to as a skill's ability. Machine to learn from experience, adapt to
new inputs, and perform human-like tasks. The terms AI and AI systems were first introduced
in the 1950s.
With the rapid development of Big Data technologies, eg. advanced computing storage
capacity and super high data rate processing machines are being revived with the availability
and power of AI, Big Data. That's why, after years of hope and promise, AI is gaining some
meaningful traction at the highest level. Companies. AI-enabled systems reportedly taking
over organizations are expanding rapidly and artificial intelligence is transforming business.
The new wave of AI systems, an organization's the ability to use data to make predictions and
significantly reduced the cost of making predictions. According to Gartner's 2018 technology
trend survey lists AI as the #1 strategic technology. ability to use improve decision making,
reinvent business models and ecosystems, and customer experience will increase the yield of
digital initiatives by 2025. Gartner survey showed that 59% of organizations are still
collecting information to build their AI strategies, the rest have made progress in testing or
adopting AI.
Answers. Organizations that are busy using next-generation artificial intelligence systems,
"artificial intelligence It faces the usual barriers to the advancement of any unproven and
unfamiliar technology,” says Whit. Andrews, Gartner VP and distinguished analyst.
"However,
3
(Yanqing Duan, John S. Edwards, Yogesh K Dwivedi, 2019)

13
 Data Science in Decision-Making Processes

early AI projects offer valuable lessons and perspectives for enterprise architecture and
tech innovation leaders are embarking on pilots and more formal AI efforts.”
In the 2010s, the term AI returned to popularity as a generic label. It can in part, it may
be a result of the fact that deep learning systems such as multilayer neural networks do not
produce explanations that people can easily understand or even understand at all. Arguably,
therefore, these systems do not represent information as in the previous definition. They
certainly do not represent human knowledge. It can be argued that there are subtle differences
between expert systems and knowledge-based systems. Systems and AI systems, but some
changes in terminology certainly not more than that just "fashion". For example, in the 1980s
or 1990s often production rules are now called business rules or just rules.

6.1 Using Artificial Intelligence for Decision Making

Artificial Intelligence, Big Data, advanced algorithms and enhanced computing power
and storage, AI systems are becoming an embedded element of the digital environment
systems, and more specifically, having a profound impact on human decision-making. As a
result, there is an increasing demand for information systems researchers to study and
research. Understanding and theoretically contributing to the effects of AI on decision
making. The progress and practical success of AI applications. For early applications of AI in
business and management, Edwards (1992) states: expert systems (representing and applying
expert knowledge using AI) management and administrative practices from
scientific/technical fields
early systems were very slow. The view put forward in the article is for expert systems.
Traditional "closed world" for application to problems in management or administration the
picture of an expert system was often inadequate. The true manifestation of the expert. The
role of the system (and indeed the human expert) in management includes much more.
negotiation and interaction rather than scientific/technical fields. results for the resulting
system is much more like the traditional picture of decision support.

6.2 Human vs AI Based Decision Making

Before mentioning what organizational structures human beings have, and AI-based
decision making can be combined, we compare their features across five key decision-making
conditions: specificity of the search space, interpretability of the decision-making process and
outcome, size of the alternative set, decision speed and repeatability. Table 1 summarizes the
characteristics of human and AI-based decision making across these conditions.

14
 Data Science in Decision-Making Processes

Table 1. Comparison of AI-Based and Human Decision Making

Because AI algorithms make decisions based on computational optimization, the

"area" in which the decision is sought must be carefully specified and is limited in terms of
objective function. For example, consider an AI algorithm designed to recommend the "best"
candidate to human decision makers from a number of candidates. This process requires a
specific definition of the desired situation. Qualifications and features that need to be
optimized, for example a candidate predicted long-term productivity and sociability with
other team after hire members and a set of variables to consider for selection, such as the
candidate's education level, age and area of expertise. Artificial intelligence of today
technology is limited to well-structured (modular or "decomposable") decisions targets, and
for this reason is often referred to as "narrow" or "weak" AI. While artificial General
Intelligence, a "powerful" AI version capable of performing any type of has attracted the
attention of significant research in recent years, experts We agree that it will take a few more
years for this technology to mature and become successful. desired level of accuracy. Human
decision makers, by contrast, can exercise. reasoning and intuition in decision making and
therefore ill-constructed decision targets often with non-intuitive decision parsing. As a result,
decisions made by humans can be difficult to clearly define (code) In the Hiring example with
an objective function, a human decision maker,
"intuitively" they base their decisions on a set of implicitly held preferences (eg. candidate
with organizational culture) without explaining why and with what weights these criteria are
taken into account.

6.3 AI from human

Artificial intelligence-based algorithms in designs that include full delegation of
decision-making make decisions without human intervention similar to organization
environments where, managers delegate decision-making authority to human experts.
However, human decision makers still retain responsibility for the decision. fully loaded
authorization is particularly useful in decision-making scenarios. Search space is specific and
limited, interpretability of decision-making process if it is less important than the accuracy of
the prediction, the alternative set large, decision-making speed is critical, and repeatability of

15
 Data Science in Decision-Making Processes

decision results it is desirable. While pure forms are still limited, current implementations of
full authorization AI includes traffic planning, real-time product recommendation systems,
dynamic pricing (for example, airline and hotel pricing, high-frequency trading) and online
fraud detection. In all of these examples, algorithm designers can specify it correctly. a
concrete objective functions. For example, referral systems used to recommend products (eg
Amazon) or streaming video (eg. YouTube and Netflix), consumer engagement, sales, and
advertising revenues, while fraud detection systems are designed to detect contingencies
minimizes activity and losses. AI-based algorithms to achieve these goals instantly scan and
evaluate millions of data points for millions of user process that is practically impossible with
human participation. Stability the data generation process and the ability to determine and
constrain the decision search space is required for AI-based decision-making algorithms to
perform true. Because the collective behavior patterns of many users have not changed
radically over time (while some users can), artificial intelligence-based systems, Predict the
preferences of classes of users on a scale with high accuracy. Decision making from fully
human to AI also includes several critical
limitations. Studies have shown that machine learning algorithms can achieve. and reproduce
implicit human prejudices against race and gender from online texts the data they use to gain
insights and inform their decisions. For example, scholars have documented how popular
performing online translation systems are. natural language processing using statistical
machine translation (SMT) Construct gender-stereotypical translations from gender-neutral
languages. 4
For example, him and his colleagues state that Google Translate translates Turkish. “O
doctor, regardless of gender. O a lady.” to these English sentences: “He is a doctor. She's a
nurse." Similarly, search engine results show that Google's autocomplete function and
Facebook ads were shown to embed hateful query Advice and negative prejudices against
women of color, religious groups and women poor. When these examples are left unchecked,
AI-based decision making It not only perpetuates but reinforces cultural stereotypes and
discrimination. In addition to these concerns, fully empowered organizational structures
decision making may come under scrutiny due to the design ethics of managers and computer
engineers. Indeed, minimizing harmful consequences automated decisions require ethical
choices in algorithms' goals, data collection methods, data cleaning and preprocessing, feature
selection, simulation Algorithm behavior and data representation. As recent examples show,
Algorithms designed for the narrow purpose of maximizing user interaction and advertising
revenues can expose users and society at large to dangerous vulnerabilities and detrimental
consequences for public welfare and democracy.
Addressing these concerns requires the concerted efforts of policymakers. academic
community, business leaders, and algorithmic decision-making designer systems. Such efforts
begin with the realization that managers can delegate authority. AI has authority, but not
responsibility. Thus, while reaping the benefit Minimizing the risks of fully empowering AI-
based decision makers, business both leaders must develop an understanding of how emerging
legal frameworks emerge. such as the European General Data Protection Regulation (GDPR)
affect algorithmic quality, fairness, accountability and transparency, and proactive stance in
ensuring the ethical design of algorithmic decision making. Such efforts include adopting and
contributing new solutions to reduce data bias. developing new ways to be fair, accountable
and transparent algorithms.

7. Decision Making Response

4
(Yanqing Duan, John S. Edwards, Yogesh K Dwivedi, 2019)

16
 Data Science in Decision-Making Processes

Advances in computing hardware, especially in general processing units and tensor

processing units and efficient algorithms enabled AI based decision making will happen at an
almost instantaneous pace. This is algorithmic feature has had a major impact on decision
making in high-speed contexts, such as high-frequency currency trading. The need for quick
decision making can undermine the results of human decision making. under high time
pressure, decision makers often use heuristics to overemphasize certain things, and ignoring
other information, leading to speed accuracy trade-off. Actually, Kahneman decomposes
human decision-making into rapid Systems thinking, intuitive, automatic, unconscious and
effortless and System thinking is slow and deliberate. The system makes quick decisions,
relying on heuristics. as associative thinking. Therefore, decision making in high-speed
environments. The system enabler may be subject to significant deviations from reality, and is
prone to systematic errors. The researcher also discovered a reverse by the time it takes to
think about a decision a decision maker's confidence in that decision.

7.1 Repeatability
AI algorithms follow the standard and non-ambiguous but relatively inflexible
decision processes that provide consistent results given consistent they entered. In contrast,
the human decision-making process is reciprocal and experience, attention patterns, emotions,
and information processing that affects the information type of individuals be careful when
making decisions, encode and undo. Such quirks make reproducing results quite problematic.
For example, psychology Research has shown that decision fatigue can lead to quality
degradation.
decisions and research, as the individual's mental energy is gradually depleted. He has shown
in cognitive science and neuroscience that emotions are generated. Powerful and sometimes
unpredictable factors in decision making.

8. Hybrid Decision-Making Models

Hybrid decision-making structures are related to organizational designs. humans and
artificial intelligence-based algorithms. The structure of one decision maker provides input to
another. hybrid structures organization designers, both manpower and AI-based decision
making, but can also amplify each other's weaknesses. We will consider two hybrid structures
stylized: algorithmic decisions as inputs. Human decision making and human decisions as
inputs to algorithmic decision construction. Algorithmic decisions as input to human decision
making. This structure consists two stages. In the first stage, AI-based decision making is
applied to the first stage set of alternatives. AI acts as a filter, rejecting unnecessary or
inappropriate alternatives and a subset of viable alternatives to the second stage. AI-based
placement where a human decision maker chooses from among these alternatives Decision-
making in the first stage is the ability of human decision-makers to effectively handles
situations involving multiple alternatives. This structure is similar as long as expert advisors
offer advice on a range of alternatives to a decision maker who has authority over the final
decision. producers exercise their discretion as to whether or not to make a decision expert
advice is taken into account. Similar to full authorization design, effective the functioning of
AI in this framework requires originality in decision-seeking space. While human
involvement makes the decision more interpretable, the decision-making process loses its
repeatability and speed. Despite these limitations, researchers in machine learning It is
actively working to develop artificial intelligence systems that learn responsibly by doing.
makes decisions only when its predictions are reliably compatible with the system's goals,
considering truth and justice. Such algorithms are more reliable. Collaboration with human

17
 Data Science in Decision-Making Processes

decision makers. For example, research has shown that designing an algorithm to learn to
delay a decision or choose to make a decision
human agent can greatly improve the accuracy and fairness of an entire system. Researchers
have also designed artificial intelligence systems. The human decision maker can adaptively
learn to delay decision making depending on the situation. both confidence in the model's
accuracy and the human decision maker's expertise and weaknesses. Human decisions as
input to algorithmic decision making. In this structure human Decision makers first select a
relatively small set of alternatives from larger ones. pool of alternatives and then pass that set
to AI algorithms for evaluation and selection of the best alternative. This structure is effective
when: human decision makers have high confidence in a small set of preferred alternatives,
however, effective evaluation of this small set requires either processing. large amounts of
data and long periods of caution by decision makers time interval. This structure can
effectively take advantage of the predictive ability. Algorithms when people are undecided
about the best alternative from a small set of selected alternatives. Because this building
human decision making at the initial stage, for which environments it is appropriate the size
of the alternative set is small. AI-based decision making in the second stage requires high
specificity of the decision search space. Optional attendance as a third step, the human
decision-making process allows the final decision to be made. interpretable, but as with AI-to-
human nature, this step reduces decision making. Provides speed and repeatability.
In healthcare, this structure finds application in artificial intelligence-based
monitoring.
body functions (eg heart rate, temperature, blood pressure) in high-risk groups patients to
predict and detect the risks of patients developing acute disorders. Because the monitoring of
bodily functions requires special attention of the person. Medical professionals have long
been an attractive environment for AI. Deep learning models process and make decisions on
anonymized electronic health records potential emergencies that clinicians should attend to. In
a recent study, researchers used such an approach, using AI-based computer vision to monitor
patients an intensive care unit. Automatically notifies system maintenance providers when a
patient feels uncomfortable or falls out of bed. Despite its many potential applications and
benefits, this decision-making the structure is vulnerable to many of the limitations fully
discussed earlier, delegation structure. Also, the lack of interpretability of AI-based decision.
In the second stage, human decision-makers are potentially opportunity to learn from past
cases and events.
8.1 Collective Human Artificial Intelligence Decision Making Structures
In this structure, decisions or their aspects are first assigned to man. and AI decision
makers based on their strengths. Human and artificial intelligence-based decisions are then
aggregated into a collective decision using an aggregation rule such as majority voting or
(weighted) averaging. In this structure, artificial intelligence-based the decision maker can be
seen as a “member” of the decision-making group. the decision counts towards the conclusion
of the decision. Collective decision-making structures human decision makers and AI-based
decision-making decision makers focus on different or overlapping elements of the decision.
their strengths and weaknesses. In our recruiting example, human decision makers can focus
on harder-to-define factors like social cohesion and leave it to algorithms assessing and
estimating more objective factors such as productivity requires querying specific questions on
large amounts of data. Unlike hybrid decision-making structures where interdependence
between human and AI-based decision maker its structure allows AI-based and human
decisions to be combined independently. In in this way, the risk of increased human decision-
making errors and biases AI-based decision makers can be minimized. Also, algorithms can
find new applications alongside human decision makers to reveal the biases and errors

18
 Data Science in Decision-Making Processes

involved in past decisions. Such applications are potentially turn algorithms into a powerful
counterweight to human decision-making errors. However, combining AI-based decisions
with human decisions still remains. For example, to the transparency and credibility issues of
organizations, Investment board example above, algorithms for output decisions can be
modified in line with the preferences of those who have the power to influence algorithms are
working.
The design of organizational decision-making structures has long been is a major
concern for managers and organizational professionals. fast progress in AI, algorithmic
decision makers are gradually established as basic organizational actors. The framework
developed here provides a basis for understanding the following how human and algorithmic
decision making can be effectively combined taking advantage of each approach and enabling
better decisions. It can have the potential to improve organizations if approached cautiously
and diligence. Framework in hybrid human-AI decision making design Managers should
consider the originality of decision research domain, the interpretability of the decision-
making process and its results, its extent alternative set, speed of decision making and
repeatability of decisions. In
managers are advised to design the most appropriate decision-making structure mapping these
five dimensions to human unique strengths and weaknesses and artificial intelligence-based
algorithmic decision making in terms of human judgment and interpretability and alternative
filtering of artificial intelligence and high precision.
In addition to the more familiar limitations of human decision-makers, practitioners
and scientists need to advance the understanding of the implications of artificial intelligence.
Limitations for organizational decision making. First, there is the risk of becoming AI.
“tricked” into altering decision outcomes through manipulation data it uses as input or
through its design (for example, by changing the weights of the estimators). Discovering these
issues can be difficult due to the nature of algorithms. opacity. Therefore, to invite algorithmic
decision making to organizations, requires new regulation and procedures to audit AI
algorithms. Encouraging Advances in the AI community will likely introduce new techniques.
Increasing the robustness and defense of neural networks against bias, and hostile attacks.
Second, artificial intelligence-based decisions have been made so far. Reinforce human biases
in existing data. Bias and injustice embedded in AI decisions especially harmful to vulnerable
groups in our society. Resist these serious concerns require more emphasis on the
development of algorithms.
Data and human decision making as well as collaboration can reveal biases AI community,
lawyers, policy makers, companies and scientists to develop new measures to be fair,
accountable and fair transparent applications of artificial intelligence in organizations. Third,
the introduction of AI-based decisions to organizations is relatively It is effective when some
level of transparency or interpretability of decisions can be achieved. obtained. Managers
should follow developments in an interpretable way. and explainable AI. Finally, algorithmic
decision-making skills remain at a high level. The decision results are specialized in such a
way that interpretation is often difficult. while introducing from AI to organizational decision
making, managers must build internal capabilities the algorithm's inputs, the algorithms
themselves, and interpretation of forecasts. As artificial intelligence technologies advance
rapidly, organizations must remain alert to the strengths and limitations of artificial
intelligence in full devolution. and hybrid human AI decision-making structures.
8.2 The Data Science Life Cycle
Although Data Science Lifecycle It is a new concept, an extension. “Data Life Cycle”,
which has a long process history in information sciences and many field sciences. Data
Lifecycle defines various phases a dataset goes through a scientific collection and research

19
 Data Science in Decision-Making Processes

process and often used to guide data management decisions and practices. I stretch the idea
goes beyond focusing on data to describe the entire process of data science with the Data
Science Lifecycle. This study extends research in Data Life Cycle by focusing on the
production of scientific findings and thus including computational components, to articulate a
clear role for inferential methodology and ethics and commodities research in data science. It
can also provide a foundation pedagogical basis of data science program design. Extending
the concept of data lifecycle. Figure 1 shows a depiction of: a Data Lifecycle that monitors a
dataset by purchasing, clearing, use, publication of the resulting data set, and then until a final
conservation/destroy decision for the dataset. It is important to note that it is not. Single fixed
definition of Data Lifespan The loop is rather a thematic abstraction. whose manifestation
may vary depending on the particular dataset or collection of datasets to which it is applied.
and the purpose of data collection. A Data Science Lifecycle, area of focus beyond the
dataset, complete build package (for example, data, code, workflow and computing
environment information) and knowledge generated during data science (scientific results)
research results. The act of scientific discovery data science only produces findings like any
field of research and typically creates or exploits other works for example, the data used to
support the findings, and the code that generates the findings from the data (this It can even
produce other works such as e.g. curriculum materials, software tools and hardware
prototypes). Research findings and works are monitored at the point of dissemination to the
research community broadcast when created. Data Management Plan required by the National
Science Foundation and other sciences funders, therefore, the beginning of the Data Science
Life loop, to highlight the importance early planning for the conservation of artifacts final
publication (results as well as supporting works).

9. Big Data & Decision-making

Big data has become a popular term to describe exponential growth, availability, and
usage. contains information and is often explained by Laney's (2001) three Vs: Volume,
Diversity, and Speed. Therefore, big data is characterized by a large amount of data from
various sources. resources, with a high production rate. Such data is incorporated into our
daily life. as most people use it indirectly every day. Big data is running in the background
and It enables organizations to analyze an unprecedented amount of information. Each we
create over 2.5 quintillion bytes of data per day, it's so big that 90% of the data Today's world
was created only in the last two years (IBM, 2017). Therefore, the volume the generated data
forms the backbone of big data. such massive ability to efficiently capture and derive
produced data quantities at high-speed real-time information is vital for organizations today. It
is when organizations actually use data. decision-making processes are becoming data-driven.
Also, big data can be divided into two categories, structured and unstructured data. structured
data, data that is pre-filtered, has a predictable format, and is defined by a set of rules. With
unstructured data, no rules and therefore no predictable format. Resources unstructured data
can be social media, text, photo or video. This diversity of resources makes data
unmanageable for traditional processing. What sets big data apart analytics, information,
technology and analytical tools needed to properly use information to gain insights. Analyzing
more data streams will increase the chances of detecting patterns. and more interesting
anomalies than traditional data streams. Because big data is about 80-90% Making sense of
unstructured, unstructured data is vital for organizations. They gain insight from the
information they collect.

20
 Data Science in Decision-Making Processes

Due to the increasing competition in the business world, it is needed for quick information
and data analysis. Fast data analysis results in better understanding and therefore better to
decide. Technology It helps to use analytics to predict the level of risk for disease and
infection. It can be a great tool for decision making in the medical field. This The hospital
collects data from monitoring electronic devices. premature babies. there is a huge amount of
data cannot be analyzed by humans. So, the role of technology can be viewed here. Searches
for structured data predicting the onset of diseases and hospital stay of a person. There may
also be new algorithms Relate a patient's behavior change to infection. Business executives
are being challenged in our age high expectations from customers, high competition,
increased labor and material costs and shorter product lifecycle. Globalization is blurring the
boundaries between them. Nations. Location and distance from the market now barriers to
access to markets. In such a volatile environment, companies constantly scanning for risks
and make business decisions quickly based on opportunities and accessible information. In
this section, we discuss the role traditional “small data” and “big data” on business decisions.
9.1 Dissemination
A prerequisite for achieving the goal of relying on data in their decision making,
disseminating knowledge and insights from their analysis to organizations. Hour They have
an internal system in both web and mobile application format. Provides real-time data. It also
has weekly and daily meetings that highlight important issues. insights and how they can
improve. In their internal communication systems, round-the-clock discussions on specific
data-related topics. Also, the strategy disseminating information to employees became evident
when we entered their offices and they observed and learned that they have more boards that
offer real-time information than employees. Their goal is to disseminate more information
faster and they are currently working on automating it. as much work as possible. They collect
data from a wide variety of sources and combine them. it is a bit difficult to convert systems
into a single system. Another problem they faced is the possibility of employees being
overwhelmed by data. As internal data shares anyone can access the app, now they are trying
to customize it. since everyone entered organization, only one of the employees take a quick
look at a screen and get the answer they're looking for. Therefore, the efficiency Data sharing
is the key to success. Since everyone has the information, they have developed a tool where
journalists can focus on their own work. their own statistics, where they can see detailed
information about the articles published in their journals. the last ten days and easily spot
what has been successful or what has gone wrong. This allows them to be Ability to make a

21
 Data Science in Decision-Making Processes

quick analysis if there is an error in the attached picture article or something. Everyone has
access to some kind of master data, but at the same time targeted. As an editor you have
greater access to certain types of data and journalists another thing. Also, each department
distributes the most relevant data. for their work.
9.2 Five Vs of Big Data
While the term "big data" is relatively new, the act of collecting and storing large
amounts of information for final analysis is too old. The concept gained momentum in the
early 2000s when industry analyst Doug Laney coined the now mainstream definition of big
data as the three Vs - Volume, Velocity and Diversity. With further refinement, big data is
now characterized by five Vs as summarized in Table 1 below.

Characteristics Description
Large amount of Data in terabytes or petabytes has been
Volume
doubling every forty months
Rate of Data accumulation is increasing in every business
Velocity
or organization
There are multitude of Data Sources like enterprise
Variety systems, social media, text, video, audio, email, RFID,
web applications and other digital devices.
Quality of Data is very essential for the accuracy of
Veracity
decision.
Economic & Social Outcomes can be improved by
Value
obtaining value from the heterogeneous data.
Renowned data analytics firm SAS considers two additional dimensions when it
comes to big data, namely Variability and Complexity.
Variability - In addition to increasing speeds and data variations, data streams can be
quite inconsistent with periodic peaks. Something trending on social media? Managing peak
daily, seasonal, and event-triggered data loads can be even more challenging with
unstructured data.
Complexity - Today's data comes from multiple sources, making it difficult to connect,
map, clean, and transform data between systems. However, it is necessary to connect and
correlate relationships, hierarchies, and multiple data connections, or your data can quickly
spiral out of control. In addition to traditional information systems, big data originates from
different sources such as social networking sites, cloud applications, software, social
influencers, Data warehouse tools, government, network technologies, legacy documents,
business applications, meteorological data and sensor data. A few resources are described
below.
Transaction data combined with statistical tools such as regression analysis and
decision tree can help define a model to predict an outcome, such as a sales forecast or the
level of success of a new product launch. The model can take inputs of historical data and
predict the dependent variable. These models can be easily created using statistical tools such
as SPSS or SAS. All historical data with arguments are known as transactions, and following
these transactions is essentially referred to as the 'Transaction Processing System'. The main
purpose of Transaction Processing System is to capture information and update data for
operational decisions in an organization. There are two ways to process transactions: Batch
processing, which processes data as a single unit over a period of time, and Real-Time

22
 Data Science in Decision-Making Processes

Processing System, where data is processed immediately. Both methods are useful for making
operational decisions in any organization.
The popularity of social media in recent years has led to the collection of information
wherever possible in the world. Events are reported as they occur. Netizens are happy to share
their opinions, product or service feedback, movie reviews on Facebook, twitter or WhatsApp
within minutes. This provides a unique opportunity for decision makers to gather market
intelligence. People share their information through social media, which helps customers
make a purchasing decision by taking a look at the feedback, customer complaints, and the
various services offered with a product. Emotions of consumers are also expressed on social
media, helping companies make production decisions. Social media Analytics is also used to
gather competitive intelligence about the firm's products and services offered by competitors
in any particular market segment. This also encourages new business ideas to improve the
business lifecycle. Therefore, social media data is crucial in making marketing decision at
strategic, operational and tactical level.
With the advent of the Internet, millions of users browse various websites that
generate high volumes of clickstreams, searching the web for products or services. Numerous
online e-commerce sites (Amazon, Flipkart, Alibaba, eBay, Paytm, bookmyshow.com etc.)
search engines (Google, Yahoo, Bing etc.) or online banking applications where millions of
users log in daily and millions of users log in are available. using them. Various clickstreams
and logs are generated during their searches or transactions, which can be valuable.

9.3 Big Data Analytics

Big data analytics has emerged as an important tool to support managerial decision
making. He suggests that big data exploration efforts can reveal previously unknown findings,
resulting in insights that aid in managerial decision making. Before the invention of
computers, humans' ability to store and manipulate data was limited. There were experts who
made decisions based on their intuition. These intuitions were not always perfect, as big data
collection was not possible. In our age, big data has led to the volume, speed and diversity of
data. This made it easier to analyze the data in terms of statistical reliability and improvement
of models. Big Data Analytics is used in decision making in the fields of e-commerce, e-
government, politics, science, technology, health, security and public safety through database
segmentation, graph mining, social network analysis, text analytics, web analytics and
sentiment analytics, crime network analysis, cyber-attack analysis, multilingual text analysis,
health analysis and patient network analysis.
In general, analytics can be classified into 3 categories as descriptive, predictive and
prescriptive according to the purpose of use. Descriptive analytics explain a phenomenon
from historical data through reports, dashboards that help understand what happened.
Predictive analytics helps us understand what might happen. It supports predictions based on
historical data, correlations between variables and patterns. Prescriptive analytics is another
powerful tool to support executive decision making. It helps to understand different results
under different scenarios. It consists of various tools such as optimization, simulations,
situation analysis scenarios by changing input parameters. With an accurate understanding of
expected results, managers can make decisions and plan for contingencies in advance.
Data sources play an important role as they can be used for analysis. Analytics can be
further classified as Text, Video with Audio, Web or Network analytics based on the data
source.
9.4 Text Analysis

23
 Data Science in Decision-Making Processes

Document display, enterprise search, search engines, user models, relevance of

feedback, query processing, billions of customer searches on google for a particular product,
searches on Amazon's website provide an indication of the customer's intent to purchase the
product. Amazon, Jet Airways and many other e-commerce companies use this feature to
recommend products or flights the next time customers browse their website, thus increasing
the probability of a customer's purchase decision.
9.5 Audio and Video Analysis
Audio analytics takes seconds to process audio through technology, primarily for
security purposes in any organization, and can monitor a wide variety of sounds in the
environment. Video analytics is used to process and analyze videos from various fields and
industries. This helps in extracting events that aid in making operational decisions.
9.6 Internet analysis
Online retailer Amazon uses clickstreams, web searches, order history, online, etc. to
gain intelligence. It uses data mining techniques to extract big data such as This intelligence is
used to make decisions about product promotions and works successfully for companies like
Amazon. A correlation is obtained between previous purchase history and potential new
purchase based on past similar purchases. This correlation is used to identify potential
customers and promote different products to these customers using email, digital media such
as Facebook, or flashing messages on Amazon.com.
9.7 Network Analytics
Network analytics provides information about devices connected to the network and
how they interact with each other. This information helps make actionable decisions that help
design network policies, improve business performance, and reduce costs.

10. Role of Big Data in Decision Making

In our age, business managers face high expectations from customers, high
competition, increased labor and material costs, and shorter product life cycles. Globalization
is blurring the borders between nations. Location and distance from the market are no longer
barriers to accessing markets. In such a volatile environment, companies need to constantly
scan for risks and opportunities and make quick business decisions based on available data. In
this section, we discuss the traditional role of "small data" and "big data" for making business
decisions. Traditional decision support systems supported internal business decisions based on
data generated by transaction processing systems such as ERPs. Further evolution led to the
addition of similar systems on the supply and demand side. These systems helped integrate
the firm's internal operations with business partners such as suppliers and customers. All these
systems used well-defined structured data in relational databases. Internal operational and
tactical decisions were made from these decision support systems. This information helped
the accuracy and speed of internal decisions. Traditional data sources provided input to data
warehousing and data mining operations. The overall architecture included the core
transactional database, the data warehouse storing the extracted data, and classifying this data
into smaller databases. Other data mining tools extract business intelligence from these
datasets. Data mining from accumulated data helped analyze and identify patterns,
correlations or association rules
In the last few years, managers' information needs have changed with the advent of big
data. In addition to the traditional datasets described above, there are large datasets that come
in structured, semi-structured or unstructured forms from a variety of sources. There are
24
 Data Science in Decision-Making Processes

several ways firms can derive value from these datasets to make strategic, tactical and
operational decisions.
When extracted for attribution rules, transactional data provides important information
to decision makers in estimating demand for purchased products or specific items.
Understanding the patterns helps retailers like Wal-Mart redesign their islands and place
products together, which leads to increased sales. Forecasting demand for specific items helps
better planning ahead of major natural disasters such as hurricanes. Analyzing terabytes of
data from the aircraft engine provides indications of component failures, improving safety as
well as maintenance.
10.1 Benefits of Decision Making with Data Analytics
The following summarizes how big data-driven insights lead to informed, predictive,
and actionable decisions.
Understanding Customer Needs
Leading banks such as Wells Fargo, Bank of America, and Discover understand
customer relationships using big data gathered from a variety of sources, as described in
Figure 3 below. From call center logs, website clicks, transaction logs, ATM transactions,
click streams, etc. They create a complete profile of the customer journey using a mix of
structured, semi-structured and unstructured data originating from customer opportunities and
issues.
5

10.2 Opponent Intelligence result

There are several studies conducted to understand consumer emotions, attitudes and
opinions using social networking sites (SNSs). In addition to consumer sentiment about their
own products, company executives need to know what customers think of their competitors'
5
(F. Provost, T. Fawcett, 2013)

25
 Data Science in Decision-Making Processes

products. This intelligence will help plan innovations in future products or devise a strategy
for marketing the products. Mining a social media data can yield a comparative analysis of
consumer opinions and sales performance of a business and its competitors.
Similarly, trend analysis provided by google provides a good mechanism for
comparing product searches of two or more competing products. These analytics provide
insights into how different products, services, or people are searched on the web in different
geographies. This can provide valuable insight into product awareness and designing future
marketing strategies or new product launches. There are websites like emotion 140.com that
provide a brief overview of possible ways to leverage this intelligence.
10.3 Time & Cost Saving
There are numerous cost and time reduction opportunities by using big data. Big data
technologies such as Hadoop clusters are emerging as a significantly lower-cost option
compared to traditional databases. It can play a role in real-time decisions regarding the
promotion of offers and services to customers based on their current location. By collecting
and analyzing data from telematics sensors installed in its 46,000 vehicles, and redesigning
vehicle routes using this large dataset, UPS saves millions of dollars in fuel.
10.4 Optimization and Simulations of Supply Chains
Supply chains are becoming more and more complex with a large number of suppliers
and partners. Over the past two decades, supply chain members have implemented corporate
systems that record every transaction. With the advancement of EAI, information sharing
occurs between business partners such as suppliers and customers. Technology plays an
important role in the efficient movement of goods across supply chains. Sensors and scanning
devices like RFID, location tracking devices like GPS, video recordings, etc. - all this
churning out large volumes of data with inventory movement. Supply chain analytics
increases the capacity of decision makers by gaining an integrated view of data within the
supply chain. We can extract, transform, analyze data from data sources within the supply
chain system and run analytics to gain intelligence. It provides a variety of advanced
capabilities such as supply chain analytics, dashboards, model and trend analysis, detailed
views, forecasts, knowledge base, scenario and situation analysis, simulations, and
optimization capabilities. They develop decision-making abilities and interpretation of
situations, which are crucial for firms in competitive business environments. There are several
opportunities to use datasets to predict future outcomes. Analytical frameworks can be
developed to analyze different datasets and make predictions as listed below.
Based on historical transaction data, using forecasting models such as regression
predicts future sales for products or services for a firm.
Based on correlations found in past purchases. By looking at these correlations and a
customer's purchase history, predict which products the customer is more likely to buy and
make recommendations online. Analyze historical data, examine customer attrition patterns
from historical data, and identify which variables contribute to customer attrition. Predict
which customers are more likely to leave, take proactive steps to retain them. There are many
other applications of predictive analytics, such as predicting which customers are more likely
to purchase which products based on their past purchases. Also, based on Google searches,
outbreaks of diseases in certain geographic areas can be predicted.
A few visionary companies have developed real-time decision-making capabilities
using supply and demand side information. Using analytics, they have online real-time
decision-making capabilities that cannot be matched with traditional business models.
Transportation service providers like Uber use big data for real-time routing of vehicles to
minimize pickup times and optimize the customer experience for a journey. Ola and Uber

26
 Data Science in Decision-Making Processes

provide real-time information to both customer and taxi driver on Google map. They
constantly receive high volumes of rack demand data and the availability of racks in different
geographic regions. They develop demand management strategies based on real-time demand
information. The City of Singapore recently introduced a demand-driven, shared private
transport concept powered by data analytics and mobile technology, dubbed Beeline. This
system uses crowdsourced travel models, transportation data to identify potential travel routes
and dynamically allocates buses to routes based on demand patterns. This reduces travel time
for commuters and increases the use of shared transport services.
10.5 Process of Decision making with Big Data Analytics
In this section, we explain the conceptual framework for companies that want to
develop analytical applications for their business to support business decisions in different
areas. Based on the objectives and the literature, the conceptual framework in Figure 4 was
designed. As explained in the previous section, several leading companies apply analytics in
one or more areas of their business. There are a few small and midsize firms that have not yet
embraced analytics to gain a competitive advantage. We have developed a framework for
such firms to integrate analytics into their core business practices. We recommend that once
the analytics application is created in the company, this functionality can be managed by a
senior manager. This will enable critical insights from data analysis to be used for decision
making. It can also be incorporated into the performance measurement system such as
analytics, balanced scorecard, real-time dashboard or KPIs.

From the literature review, the authors explored the different ways big data plays a
critical role in analytics, providing insights for decision making. Figure 3 clearly shows the
path from big data to decision making. The structures are discussed in detail below. The
conceptual framework shown in Figure 4 illustrates the relationship between the five
constructs.

Develop
Data Data Decision
Data Analytics
Mining Analysis Making
Sources
Figure 4 Decision Framework
Data sources, enterprise systems, customer data, supplier data, social media data,
logistics trajectory, etc. It includes traditional data sources such as a first step, the firm must
have the information systems infrastructure and processes to collect data based on various
sources. Its business model. For example, a product company that focuses on designing new
products and putting them together will collect data on procurement, distribution and

27
 Data Science in Decision-Making Processes

logistics. A logistics company can handle fleet movement, packages and routes, etc. will
collect relevant data.
Data mining is the process of discovering patterns in large data sets using a variety of
statistical techniques, computer programs, and database systems. It helps to extract
meaningful insights from data. After the data sources are created based on the analytical
requirements of the different departments, the data warehouse will be developed to store
multidimensional data for query and analysis purposes. Mining these datasets helps to find
previously unknown patterns, correlations and relationships between different variables.
Having a wide variety of datasets is necessary but not sufficient. The firm needs to
develop talent in analytics to gain insights from data. Firm needs to develop a team of
analytics professionals with diverse interdisciplinary skills in the team- a) R software, SPSS,
SAS etc. knowledge and experience in statistical tools b) programming knowledge c) domain
knowledge of business processes d) data management skills in SQL e) data analysis
experience. Various opportunities for analytics projects can be identified with the help of
domain knowledge and data analysis.
As explained earlier, firms can benefit from three types of analytics – descriptive,
predictive, and prescriptive. The Firm may develop its requirements based on opportunities to
acquire new customers, retain existing customers, or identify risks to its business. Business
analytics uses a set of principles, statistical tools, and computer algorithms for extracting
information from data. IBM, SAP, Oracle, SAS, SPSS and R software have already developed
tools for data analysis and model development. Hadoop provides framework for developing
models from different data sources.
Business analytics is emerging as a potential tool for businesses to improve their
business performance in terms of customer service, retention and acquisition. Predictive
analytics helps us predict what might happen based on certain available information. This
gives a firm a competitive advantage to plan ahead. Models derived from data, correlations,
and associations are useful for improving sales performance, identifying the right customers
for products, or segmenting markets. Analytics has applications in every field where data can
be collected. It is available in the areas of supply chain analytics, inventory optimization,
supply planning, demand forecasting, fleet and route sizing and optimization. It provides
social media data, competitive intelligence, new product ideas, and current product reviews –
important information that helps in deciding next strategies. Big data discovery can provide
surprising and actionable findings. The main purpose of data science is to improve the ability
of managers to make better business decisions. Leading companies like Amazon, Wal-Mart,
Google or Netflix have mastered the art of using data and analytics as a tool for predictions,
simulations, or sometimes just insight. Amazon and Wal-Mart use analytics to make decisions
on all aspects of running their businesses, from generating demand to efficiently managing
their supply chains. We've come a long way since the information revolution changed the way
commercial firms operate. Big data helps companies gain competitive advantage by using
different analytical techniques. These techniques help us gain insights, patterns, correlations
and relationships that cannot be understood with traditional small data. These include social
media data, competitive intelligence, cost and time reduction strategies, supply chain
analytics, web analytics, etc. supports the decision-making process of business managers.
Realizing the importance of big data and developing products around data, companies
received huge dividends in this process. last years. Many firms use analytics in nearly all
aspects of running their business to reap the benefits of analytics-based decision making. In
this article, we present a conceptual framework for enhancing analytics capabilities and how
this emerging knowledge can help small and medium-sized firms compete using fewer
resources. It can be adopted by such companies by making changes according to the business

28
 Data Science in Decision-Making Processes

area and model. This framework can be a starting point for further analysis, development and
future research opportunities.
With the continuous digitalization of every aspect of society as well as the business
world, the production speed of high-speed high-volume data will continue. This provides a
solid opportunity to leverage the analytics space to make decisions in different business areas.
There are several unique research opportunities in different areas of business, science and
government where data is constantly generated.

Conclusion
In the decision-making phase, it is necessary to make a choice with the thought of
what we do or what we will reach by using the feedback from the previous steps. Today, the
method of making decisions based on feelings or thoughts has become a thing of the past.
Making decisions by thinking analytically rather than understanding the technique of the job
is now a sought-after feature in all business areas. You can be a very good technical service,
for example, you may understand all the functions of the machine, but if you work in a factory
or organization that produces that machine in series, you will see that only technical
knowledge is not enough to be a manager. If you don't trust the numbers, it will be inevitable
to fail if you go the old ways. Because now, management means analytical from top to
bottom, manager means the person who analyzes and makes decisions even with numbers, of
course, this is my opinion. As someone who has been a manager this year, I fully rely on my
ability to analyze numbers. In fact, I will close with an example that happened to me while
writing this article. As of February 1, 2022, I was appointed as the operations manager.
Naturally, I get various reactions from my environment. I received a different response from
our technical service manager named Serdar Çağatay, whom I love and value as "arrogant". In
our last meeting, he said that it would be difficult for me as the operations manager to decide
on technical proforma, because I did not know the technique of the job very well, but he was
not malicious, I really do not understand the technique of the job as much as they do. We are
in an automotive company and I do not know what the flywheel prices are in the market or the
routine replacement kilometers. However, I gave an answer. Yes, I do not have as much
knowledge of technical issues as you, but I can learn quickly with numbers and analysis and
decide with numbers again. I replied, "For example, by taking flywheel/brand offers from
companies in the market and comparing them, I can of course make the best purchase for the
company with analysis. Why did I give this example? This is the point we have come to in the
changing world. Old information is valuable, but nowadays you need to support this
information with analysis and decision-making features, otherwise you will fail.

29
 Data Science in Decision-Making Processes

References
Dr. Mike Cheung, D. K.-Y. (2021, 6 5). Workshop on statistical power analysis. (meeting.isdsa.org,
Interviewer)
F. Provost, T. Fawcett. (2013). Data Science and its relationship to Big Data and data-drivendecision
making. ResearchGate, 2-37.
Filip Björkman, Sebastian Franco. (2017). How big data analytics affect . Uppsala Universitet Press
decision-making, 1-30.
M. I. JORDAN, T. M. MITCHELL. (2015). Machine learning: Trends, perspectives, and prospects.
Science, 255-260.
Michael Veale, L. E. (2017). Clarity, Surprises, and Further Questions in the Article 29 Working Party.
Data Is Power: Profiling and automated decision-making in GDPR, 17.
Nada Elgendy, A. E. (2016). Big Data Analytics in Support of the Decision Making Process. Science
Direct, 1071-1084.
Nada Elgendy, Ahmed Elragal. (2016). Big Data Analytics in Support of the Decision Making Process.
ELSEVIER, 1071-1084.
Paul Brous, Marijn Janssen. (2020). Trusted Decision-Making: Data Governance for Creating Trust in
Data Science Decision Outcomes. MDPI Administrative sciences, 2-20.
Samis Passi, Phobe Sengers. (2020). Making data science systems work. Research article, 10-32.
Sarker, I. H. (2021). Data Science and Analytics: An Overview from Data-Driven Smart Computing,
Decision Making and Applications Perspective. SN Computer Science, 2-377.

30
 Data Science in Decision-Making Processes

Shahriar Akter, Ruwan Bandara, Umme Hani, Samuel Fosso Wamba, Cyril Foropon, Thanos
Papadopoulos. (2019). Analytics-based decision-making for service systems: A qualitative
study and agenda for future research. ELSEVIER, 85-95.
Shirish Jeble, Sneha Kumari, Yogesh Patil. (2018). Role of Big Data in Decision Making. ISSN 1979-
3561, 36 - 44.
Stodden, V. (2020). The Data Science Life Cycle. Illinois: Victoria Stodden.
Wieslawa Gryncewicz, M. S.-B. (2021). Data Science in Decision-Making Processes. European
Research Studies Journal, 14.
Yangyong Zhu, Y. X. (1998). Defining Data Science. School of Fudan, 1-8.
Yanqing Duan, John S. Edwards, Yogesh K Dwivedi. (2019). Artificial intelligence for decision making
in the era of Big Data. Cronfa, 63-71.
Yash Raj Shrestha, Shiko M. Ben-Menahem, Georg von Krogh. (2019). Organizational Decision-Making
Structures in the Age of Artificial Intelligence. BerkeleyHass, 1-18.

31