Copyright © 2020 Statswork. All rights reserved 1

Introduction to Business Analytics and Operational Research Solution

Methods, Including Decision Analysis, Linear Programming, Inventory

Control, Simulation and Markov Chains

Dr. Nancy Agens, Head,

Technical Operations, Statswork

info@statswork.com

I. INTRODUCTION

In modern years, there is a growing demand

in the field of business analytics. It actually

means that what outcome we should get in

business from the data to make better

decisions. This is often sound like relating a

business problem to an operation research

problem. However, there is often a question

arise in connecting the business analytics to

the operation research problem. In this blog,

I will explain you the meaning of business

analytics and how it is related and useful in

the operation research methods or decision

making including linear programming,

inventory management, simulation and

II. MARKOV CHAINS

Analytics are used to identify (i)

what has happened? (ii) What should

happen? And (iii) what will happen? In the

business. These three forms of question are

categorized into Descriptive, Prescriptive

and Predictive analytics respectively.

However, business analytics is the study of

data via statistical techniques, constructing

predictive models, implementing the

optimizing rule and draw a valid inference

according to the business needs. Thus,

business analytics uses a huge amount of

data or simply big data to make a profitable

conclusion.

There is a different approach to business

analytics, which in turn delivers profitable

benefits (Budnick et al., 1994). I will list out

a few uses of business analytics for the

betterment of the business.

If a business company wants to identify

the pattern of the sales of a product or to

find a new pattern to promote the growth

of the business, then business analytics

is used to implement the data mining

techniques such as classification,

regression analysis, clustering analysis,

etc., and to understand the complex data

using neural networks, deep learning and

machine learning techniques.

Business analytics is used to do

quantitative statistical analysis or

solving a mathematical model to deliver

justifications for the occurrence of the

problem

It can be used as a supporting tool for

conducting any multivariate testing and

A/B testing to find the relationship or

test the relationship with past decisions.

It can be used for predictive modelling to

improve business standards.

Apart from the benefits and uses of

business analytics, the main goal of business

analytics is to identify which dataset will be

useful and how it can be taken forward to

solve the business problems and increase the

profit, productivity, and efficiency. So far, I

explained to you about the meaning and

benefits of business analytics. However, in

recent years, business analytics in

operational practice has become a great

interest among researchers. With the growth

of technologies, and with the large amount

of data at hand, it is important to make use

of analytics and the operation research

approach to solve many complex business

problems (Choi et al., 2017; Hillier &

Lieberman, 2015). Thus, in the coming

Copyright © 2020 Statswork. All rights reserved 2

years, business analytics tools are the most

powerful tool to take the business standard

to the next level. Now, let look at how a

simple Markov chain is used to solve a

business problem.

Consider a bank which deals with

both asset and liability products, and it is

obvious that loans taken from the bank play

a vital role in the revenue. Hence, the bank

executive decided to hire a consultant to find

whether they end up in good loans, risky

loans, paid-up loans or bad loans.

In this example, the bad loans and

the paid-up loans are the absorbing nodes or

the end state in a Markov chain. The

absorbing node is that it has no transition

probability to any other nodes. So, as a

statistical consultant, the first step is to

understand the trends in the loan cycle with

the previous study. Let's say; the following

Markov chain represents the pattern of loans

for the previous year

Fig 1. Markov Chain for pattern of loans

From the above transition diagram, it

is clear that the bad loans and paid-up loans

are the absorbing states; that is, the process

end and stays in these states forever.

Otherwise, paid-up loans cannot be a bad

loan or risky or good and similarly, the bad

loans cannot be a paid-up or risky or good.

Next step is to calculate the

transition probability matrix with the

previous probability. That, it with the

previous probability, estimate the number of

loans belongs to each category. From the

diagram, it is clear that 60% has good loans,

and 40% has bad loans. Thus, the calculation

becomes,

Copyright © 2020 Statswork. All rights reserved 3

From the final output, it is expected

that 15% of the loans are going to be paid-up

loans for the current year and 16% becomes

a bad loan. Thus, from this Markov chain

example, the retail industry can develop

their business insights to decrease the

percentage of bad loans in the future. In

addition, if you want to predict the same for

2 years, then with the same transition matrix,

it is calculated as

Similarly, the process is repeated until the convergence is achieved. That is,

Copyright © 2020 Statswork. All rights reserved 4

From the convergence result, it is identified

that 54% of the present loan will be paid

fully, and 46% will be a bad loan. This is

useful in identifying the risk of banks in

issuing loans to customers.

Suppose, if you want to identify the

proportion of good loans becoming a paid

loan, then you should start with 100% of

good loans and others as 0% in the initial

stage and repeat the process until

convergence is achieved.

From the results, it is identified as

only 23% becomes a bad loan whereas in the

previous case it was recorded as 46%.

Similarly, if you are interested in identifying

the proportion of risk loans ending as paid-

up or bad loans then assign 100%

probability to risk loans and others with 0%

probability the do the process until

convergence and deliver a valid conclusion.

The previous case deals with the

Markov process into business insights.

However, there is still a question persists

where the analytics relate to operation

research? An operation research scientist is

everywhere in the process and few having

developed these kinds of tools to solve a

business problem and few have developed a

robust model for the same. In practice,

Operational Analytics or business analytics

involves building a suitable model or

developing a predictive model to make

meaningful business decisions. It may be a

transportation model, or the Markov model,

or the Linear programming model or a

simulation model; the objective is to satisfy

the business needs and do a profitable

business.

III. SUMMARY

I presented an informal description

of business analytics and Operations

Research in this blog with an application to

a retail bank industry using Markov chains. I

personally feel that if I want to understand

anything, it is better to dig deeper into the

topic and go for details.

REFERENCES

[1] Budnick, F.S., McLeavey, D.W. & Mojena, R. (1994).

Principles Of Operations Research For Management

(2nd Edition). Irwin series in quantitative anlysis for

business. [Online]. A.I.T.B.S. Publishers. Available

from:

https://books.google.co.in/books?id=wBMVYAAA

CAAJ.

[2] Choi, T.-M., Chan, H.K. & Yue, X. (2017). Recent

Development in Big Data Analytics for Business

Operations and Risk Management. IEEE

Transactions on Cybernetics. [Online]. 47 (1). pp.

81–92. Available from:

http://ieeexplore.ieee.org/document/7378465/.

[3] Hillier & Lieberman, J. (2015). Introduction to

Operations Research. [Online]. Available from:

https://notendur.hi.is/kth93/3.20.pdf.