Introduction to Business

Analytics
Business Analytics
• Business analytics is a set of statistical and operations research techniques,
artificial intelligence, information technology and management strategies used for
framing a business problem, collecting data, and analysing the data to create value
to organizations.

• Business Analytics can be broken into 3 components:

1. Business Context
2. Technology
3. Data Science
Business Analytics
• Business Analytics can be broken into 3 components:
1. Business Context

• For many customers shopping is a habit, and most do not respond to promotions
since shopping is a routine for them. The shopping behaviour changes during
special events such as marriage and pregnancy and it becomes easy to target them
during these special events
Business Analytics
• Business Analytics can be broken into 3 components:
2. Technology

• Today most data are unstructured data; data that is not

in the form of a matrix (rows and columns) is called unstructured data. Images, texts, voice, video,
click stream are few examples of unstructured data. To analyse data, one may need to use software
such as R, Python, SAS, SPSS, Tableau, etc. Technology is also required to deploy the solution; for
example, in the
case of Target, technology can be used to personalize coupons that can be sent to individual
customers.
Business Analytics
• Business Analytics can be broken into 3 components:
2. Technology

• Today most data are unstructured data; data that is not

in the form of a matrix (rows and columns) is called unstructured data. Images,
texts, voice, video, click stream are few examples of unstructured data. To analyse
data, one may need to use software such as R, Python, SAS, SPSS, Tableau, etc.
Technology is also required to deploy the solution; for example, in the
case of Target, technology can be used to personalize coupons that can be sent to
individual customers.
Business Analytics
• Business Analytics can be broken into 3 components:
3. Data Science

• Classification .. Likely to be defaulter or not in case of Loan

Analytics Capability Maturity Model

• Type and stages

DESCRIPTIVE ANALYTICS

• Descriptive analytics is the simplest form of analytics that mainly uses simple
descriptive statistics, data visualization techniques, and business related queries to
understand past data.
PREDICTIVE ANALYTICS

• It aims to predict the probability of occurrence of

a future event such as forecasting demand for products/services,
customer churn, employee attrition, loan defaults, fraudulent
transactions, insurance claim, and stock market fluctuations.
• While descriptive, analytics is used for finding what has happened in
the past, predictive analytics is used for predicting
what is likely to happen in the future.
PREDICTIVE ANALYTICS
PRESCRIPTIVE ANALYTICS

• Prescriptive analytics is used for choosing optimal actions once an

organization gains insights through descriptive and predictive
analytics.
• In many cases, prescriptive analytics is solved as a separate
optimization problem. Prescriptive analytics assists users in finding
the optimal solution to a problem or in making the right
choice/decision among several alternatives.
PRESCRIPTIVE ANALYTICS
• Operations Research (OR) techniques form the core of prescriptive
analytics. Apart from operations research techniques, machine learning
algorithms, metaheuristics, and advanced statistical models are used in
prescriptive analytics.
• Note that actionable items can be derived directly after descriptive and
predictive analytics model development; however, they may not be the
optimal action.
• Inventory management is one of the problems that is most frequently
addressed using prescriptive analytics. Samsung implemented a set of
methodologies under the title ‘Short Life and Low Inventory in
Manufacturing’ (SLIM) to manage all the manufacturing and supply
chain problems.
PRESCRIPTIVE ANALYTICS
ANALYTICAL CAPABLITY
Analytics Techniques
• Regression
• Regression is the most frequently used predictive analytics tool. It is a
supervised learning algorithm. In management and social sciences,
almost all hypotheses are validated using regression models.
• In business, irrespective of the sector, the decision maker would like to
know how the key performance indicators (KPIs) of the business are
related to macro-economic parameters and other internal process
parameters.
Analytics Techniques
• Regression
• Regression is an excellent tool for establishing the existence of an
association relationship between a response variable (KPI) and other
explanatory variables.
Analytics Techniques
• Logistic and Multinomial Regression
• Logistic and multinomial logistic regression techniques are used to
find the probability of occurrence of an event. Logistic regression is a
supervised learning algorithm.
• Logistic regression is used for solving classification and discrete
choice problems. Classification problems are common in many
businesses. For example, banks and financial institutions would like to
classify their customers into several risk categories.
Analytics Techniques
• Logistic and Multinomial Regression
• Companies would like to predict which customer is highly likely to
churn in the next quarter. Marketers would like to know which brand a
customer is likely to buy and whether promotions can make a
customer change his/her brand
• loyalty. Credit scoring and fraud detection are other popular
applications of logistic regression
Analytics Techniques
• Decision Trees
• Decision trees or classification trees are usually used for solving
classification problems. Although the decision trees are usually used
for solving classification problems (in which the outcome variable is
discrete), they can also be used when the outcome variable is
continuous.
Analytics Techniques
• Markov Chains
• Olle Haggstrom (2007) wrote an article stating that problem solving is
often a matter of cooking up an appropriate Markov chain.
• One of the initial applications of Markov chains was implemented by
the American retail giant Sears. They used a Markov Decision Process
to decide the optimal mailing policy for their catalogues (Howard,
2002).
• Today, Markov chains are one of the key analytics tools in marketing,
finance, operations, and supply chain management.
Analytics Techniques
• Markov Chains
Analytics Techniques
• Markov Chains
BIG DATA ANALYTICS
• Big data is a class of problems that challenge existing IT and
computing technology and existing algorithms. Traditionally, big data
is defined as a big volume of data (in excess of 1 terabyte) generated
at high velocity with high variety and veracity. That is, big data is
identified using 4 Vs, namely, volume, velocity,
variety, and veracity which are defined as follows:
• Volume is the size of the data that an organization holds. Typically,
this can run into several petabytes (1015 bytes) or exabytes (1016
bytes). Organizations such as telecom and banking collect and store a
large quantity of customer data.
BIG DATA ANALYTICS
• Data collected using satellite and other machine generated data such
as data generated by health and usage monitoring systems fitted in
aircrafts, weather and rain monitoring systems can run into several
exabytes since the data is captured minute by minute

• Velocity is the rate at which the data is generated. For example, AT&T
customers generated more
than 82 petabytes of data traffic on a daily basis (Anon, 2016)
BIG DATA ANALYTICS
• Variety refers to the different types of data collected. In the case of
telecom, the different data types are voice calls, messages in different
languages, video calls, use of Apps, etc

• Veracity of the data refers to the quality issues. Especially in social

media there could be biased or incorrect data, which can result in
wrong inferences.
BIG DATA ANALYTICS
• For example, Google processes 24 petabytes of data every day
(Mayer-Schonberger and Cukier, 2013). Google was the first to exploit
big data for targeted advertising using clickstream data. Google also
predicted the spread of H1N1 flu based on the search terms entered by
Google users (Ginsberg et al., 2009).
• According to Richard Kellet, Director of Marketing at SAS, we
(human) created 500 times more data in the last 10 years than what we
had done prior to that, since the beginning of humanity (Scott, 2013).
BIG DATA ANALYTICS
• Every B787 flight creates half a terabyte of machine-generated data.
For large banks, the automatic teller machine (ATM) transactions
themselves will run into several billion transactions per month.
• Telecom call data, social media data, banking and financial
transactions, machine-generated data, and healthcare data are a few
examples of the sources of big data.
BIG DATA ANALYTICS
• Innovative parallel processing capabilities such as Apache Hadoop
(that comes with the ability to process large-scale data sets in multiple
clusters using the Hadoop Distributed File System) and MapReduce
(which enables parallel processing of large data sets) are used by
organizations to handle big data.
WEB AND SOCIAL MEDIA ANALYTICS
• Social media and mobile devices such as smart phones are becoming
an important source of data for all organizations, small and big. Social
media is also an important marketing channel for marketers since it
helps to create a buzz or electronic word-of-mouth (WoM) effectively.
Stelzner (2013) claimed that 86% of the marketers indicated that
social media is important for their business.
• Social media is measurable in terms of impressions, visits, views,
clicks, comments, shares, likes, followers, fans, subscribers, etc.
Impact of conventional media cannot be measured, for example, views
of a hoarding or newspaper ad cannot be measured.
MACHINE LEARNING ALGORITHMS
• Machine learning algorithms are part of artificial intelligence (AI) that
imitates the human learning process. Humans learn through multiple
experiences to perform a task.
• Similarly, machine learning algorithms usually develop multiple
models and each model is equivalent to an experience.
• For example, consider someone trying to learn tennis. Getting the
service right requires much practice especially, to learn to serve an ace
(serve such that the opponent player is unable to reach the ball).
MACHINE LEARNING ALGORITHMS
• To master the service in tennis (especially ace), a player probably has
to practice several thousand times; each practice session is a learning.
AI is still a developing field and nowhere near the human cognitive
process. In machine learning algorithms, we develop several models
which can run into several hundreds and each model is treated as a
learning opportunity.
• Mitchell (2006) defined machine learning as follows:
“Machine learns with respect to a particular task T, performance
metric P following experience E, if the system reliably improves its
performance P at task T, following experience E”.
MACHINE LEARNING ALGORITHMS
• Machine learning algorithms are classified into the following four
categories:
• Supervised Learning Algorithms: When the training data set has
both predictors (input) and outcome (output) variables, we use
supervised learning algorithms. That is, the learning is supervised by
the fact that predictors (X) and the outcome (Y) are available for the
model to use.
• Techniques such as regression, logistic regression, decision tree
learning, random forest, and so on are supervised learning algorithms
MACHINE LEARNING ALGORITHMS
• Machine learning algorithms are classified into the following four
categories:
• Unsupervised Learning Algorithms: When the training data has only
predictor (input) variables (X), but not the outcome variable (Y), then
we use unsupervised learning algorithms.
• Techniques such as K-means clustering and Hierarchical clustering are
examples of unsupervised learning algorithms.
MACHINE LEARNING ALGORITHMS
• Machine learning algorithms are classified into the following four
categories:

• Reinforcement Learning Algorithms: In many cases, the input

variable X and the output variable Y are uncertain (predictive
keyboards and/or spell check). The algorithms are also used in
sequential decision-making scenarios; techniques such as dynamic
programming and Markov decision process are examples of
reinforcement learning algorithms.
MACHINE LEARNING ALGORITHMS
• Machine learning algorithms are classified into the following four
categories:
• Evolutionary Learning Algorithms: These are algorithms that
imitate human/animal learning process. They are most frequently used
to solve prescriptive analytics problems. Techniques such as genetic
algorithms and ant colony optimization belongs to this category.