FINM4100

Analytics in Accounting,

Finance and Economics

Week 8

Data analytics techniques and applications in

accounting, finance and economics

Lesson Learning Outcomes

1 Explore and apply some of the widely used data

analytics techniques which are used to extract

insights in accounting, finance and economics, e.g.

• Association rule learning

• Classification tree analysis

• Genetic algorithms

• Machine learning

• Regression analysis

Software for today

1. Google Colab

• Either

A. watch the teacher demonstrate analytics and accounting in python OR

B. you can run the python scripts yourself in Google Colab

• If you want to run the code provided, make sure you have access (signed in) to Google Colab https://colab.research.google.com

2. Exploratory

A. watch the teacher demonstrate analytics and accounting in Exploratory OR

B. run each step yourself

Data for today

1. GroceryStoreDataSet.csv

2. Churn_Modelling.csv

3. Salary_Data.csv

This Photo by Unknown Author is licensed under CC BY-SA-NC

A Vital Commodity

“It is a capital mistake to

theorize before one has

data.”

Sir Arthur Conan Doyle

Author

Sherlock Holmes

The Big Data Environment

216,000TB Amount of new information

generated per person per year

90% Proportion of the world’s total

big data created in the past 3

years.

$65 million Boost in net income for every

Fortune 1000 company (if

data access is boosted 10%)

83% Proportion of surveyed

businesses (Accenture)

investing in Big Data

initiatives.

Inevitable Transition

Force multiplier – Big data analytics and analytics

infrastructure is the means by which institutions apply force to

achieve geo-economic advantage.

Commercial activities will increasing relay on sophisticated

network-based logistics, communications systems and a big

data ecology to recommend products, retain customers and

mitigate churn.

The goal is to turn data into information, and information into

insight.

Techniques

There are a number of widely used analysis techniques to

extract valuable insights from data.

• Association rule learning

• Classification tree analysis

• Genetic algorithms

• Machine learning

• Regression analysis

This Photo by Unknown Author is licensed under CC BY-SA-NC

Association Rule Learning

Association rule learning is a method for discovering interesting

correlations between variables in large databases. It was first used by

major supermarket chains to discover interesting relations between

products, using data from supermarket point-of-sale (POS) systems.

“Are people who purchase tea more or less

likely to purchase carbonated drinks?”

Association Rule Learning Association rule learning is used to:

• place (correlated) products in better proximity to each other

in order to increase sales

• Determine data quality in accounting

• Help in investment planning

• monitor system logs to detect intruders and malicious

activity

• provide insight in revenue analysis

T h is

P h o to

b y U

n k n o w

n A

u th

o r is

lic e n s e d u

n d e r C

C B

Y

This Photo by Unknown Author is licensed under CC BY-NC-ND

Association coding concepts

“The Apriori Algorithm, used for the first phase of the Association Rules, is the most popular and classical algorithm in the frequent old parts. These algorithm properties and data are evaluated with Boolean Association Rules. In this algorithm, there are product clusters that pass frequently, and then strong relationships between these products and other products are sought. Three main parameters that are used to identify the strength of the algorithm are

Activity 2: Python in Colab

• Make sure you have access (signed in) to Colab https://colab.research.google.com

• Click on the ‘File’ menu and select ‘New notebook’

Activity 2: Python in Colab

We have grocery store data for you to analyse

• The code is given below. All you have to do is click on the arrows and run the

code

• NOTE: you don’t need to run the interpretation text at the end it is just to help

you interpret the results

• https://colab.research.google.com/drive/1Qg0qokW_oDUI6xU8gvmZeV6AiMo

6bhxu?usp=sharing

• We start by getting you to choose to upload the GroceryStoreDataSet.csv file

on MyKBS

(You will be prompted to Choose (find) the data file from where it is

stored on your device)

Activity 2: Output

Interpretation

# The probability of seeing sugar sales is seen as 30%.

# Bread intake is seen as 65%.

# We can say that the support of both of them is measured as 20%.

# 67% of those who buys sugar, buys bread as well.

# Users who buy sugar will likely consume 3% more bread than users who don't buy sugar.

# Their correlation with each other is seen as 1.05.

# As a result, if item X and Y are bought together more frequently, then several steps can be take

n to increase the profit.

Glossary 1: What are Bonds and

mortgage-backed security (MBS) ?

• Securitisation is about pooling debt (such as mortgages) and selling their cash flows, as securities, to third party investors

• A bond is a fixed income security that provides a return in the form of fixed interest payments made at regular intervals over time

• A mortgage-backed security (MBS) is an investment similar to a bond. A MBS consists of a bundle of loans sold to investors.

• The bundles are rated between AAA (best, debts most likely to be paid back) through to “not rated” (worst)

• The bank effectively becomes an intermediary between a person with a mortgage and investors. See next slide

Risk Ratings

Can machine learning help classify items for investment?

Classification Tree Analysis YES! Classification, a machine learning method can be used to classify debt

• Statistical classification is a method of identifying categories that a

new observation belongs to. It requires a training set of correctly

identified observations – historical data in other words.

• Classifying customers correctly will maximise sales and minimise

expenses (cost of acquisition, discounts, bad debt etc).

“Are these mortgages investment grade or sub-prime?”

AAA BBB D

Classification Tree Analysis

Statistical classification is also being used to:

• automatically assign financial documents to

categories;

• categorize customers into groupings (e.g.

insurance);

• classify transactions This Photo by Unknown Author is licensed under CC BY-NC

Activity 3: Decision Trees

• Decision trees that classify items into categories are called “Classification tree”

• Decision trees that predicts numerical values is called “Regression tree”

Watch the video at https://www.youtube.com/watch?v=zs6yHVtxyv8

From groups,

• Suppose that you are an analyst at the tax office. You wish to identify which of

your clients is most likely to avoid lodging a tax return form and thus avoid

paying tax (or even recouping funds after paying too much tax)

1. Discuss the idea of using a classification tree for this purpose

2. How would you limit so-called “overfitting”?

3. What kind of data would you collect for the classification tree?

Genetic Algorithms

Genetic algorithms are inspired by the way evolution works – that is,

through mechanisms such as inheritance, mutation and natural selection.

These mechanisms are used to “evolve” useful solutions to problems that

require optimization.

“Which TV programs should we offer viewers,

and in what time slot, to maximize viewership?”

Genetic Algorithms

• A biology- inspired algorithm which reflects natural selection (the fittest

individuals survive)

• Technically an optimisation method

• It has three main rules: selection

crossovermutation

evaluation

This Photo by Unknown Author is licensed under CC BY-SA

1. “Selection rules select the

individuals, called parents, that

contribute to the population at the

next generation.”

2. Crossover rules represent

reproduction, i.e. combining two

parents to form children.

3. Mutation rules apply random

changes to individual parents to

create genetic diversity in children.

Genetic Algorithms Genetic algorithms are being used in:

• Finance:

– Algorithmic trading;

– Financial statement fraud

• In accounting

– Distribution problems assigning sources to destinations

– Bankruptcy predictions

• The cobweb model in economics which explains

why prices may fluctuate in certain markets.

This Photo by Unknown

Author is licensed under

CC BY

Activity 4: Genetic Algorithms

• Here is a video with a real-world examples of a genetic algorithms. Watch the video at

Form groups and answer the following,

Q1. What issues do genetic algorithms appear to have at the start?

Q2. What are the three rules used here?

Q3. What applications are shown here?

Q4. How could this be used in accounting and finance?

Machine Learning

Machine learning includes software that can ‘learn’ from data and generate

adaptive solutions. It gives computers the ability to compute solutions

without being explicitly programmed along a strict instruction set.

Applications are primarily focused on making predictions based on known

properties learned from sets of ‘training data’.

“What other products would this customer likely

purchase, based on their transaction history?”

Extract Transform Test Validate

Machine Learning

Machine learning is being used to:

• distinguish between spam and non-spam email

messages;

• learn invoice coding behaviours for allocation

purposes

• determine the best content for engaging

prospective customers;

• run AI chatbots for customer enquiries

This Photo by Unknown Author is licensed under CC BY-NC-ND

Activity 5: Customer churn example

Source: https://www.kaggle.com/kmalit/bank-customer-churn-prediction

• Watch the demo by your teacher or run the code for analysis of

customer churn at

https://colab.research.google.com/drive/1Sgro8G9o2UtErsiEMG-

UOe7yS-JQMqUU?usp=sharing

• Data for this script is Churn_Modelling.csv

• NOTE: This is a part of a project on Kaggle, so we took a small section

of it to give you an appreciation of this technique

• Interpret your findings. For example, regarding churn, is there any

difference depending on the country of origin of customers, gender,

ownership of a credit card or whether or not a member is active?

Regression Analysis

• Regression analysis involves manipulating one or more independent

variables (i.e. number of customers) to see how they influence a

dependent variable (i.e. weekly sales).

• The dependent variable is also called a target variable

• The independent variable is also called a predictor variable

“How would social, biological, demographic and

lifestyle factors affect health insurance premiums?”

Social Biological Demography Validate

Copyright © 2013 Pearson Australia (a division of Pearson Australia Group Pty Ltd) – 9781442549272/Berenson/Business Statistics /2e

The simple linear regression equation (derived from a sample) looks like a

straight line. The mathematical representation is shown below.

Estimate of

the

regression

intercept

Estimate of the regression

slope

Estimated (or

predicted) Y value for

observation i

Value of X for observation

i෡𝒀𝒊 = 𝒃𝟎 + 𝒃𝟏 𝑿𝒊

Simple linear regression equation

for estimating values

• Example: ෣𝑊𝑒𝑒𝑘𝑙𝑦 𝑠𝑎𝑙𝑒𝑠 = 98.248 + 0.110 Number of customers

• Weekly sales is the target variable,

• Number of customers is a predictor variable

Simple linear regression equation

for estimating values • Example: ෣𝑊𝑒𝑒𝑘𝑙𝑦 𝑠𝑎𝑙𝑒𝑠 = 98.248 + 0.110 Number of customers

• Weekly sales is the target variable,

• Number of customers is a predictor variable

0

50

100

150

200

250

300

0 500 1000 1500 2000

W e

e k ly

S a

le s

Number of Customers

slopeintercept x෡𝒀

Regression Analysis Applications

Regression analysis is being used to determine how:

• In Economics:

– Demand curves

– Predicting economic growth rate

• In Finance:

– Forecasting, e.g. revenues from Ads

– Bank performance given multiple variables

– levels of customer satisfaction affect customer loyalty

• In accounting:

– to estimate fixed and variable costs

– Cost versus hours worked

T h e s e P

h o to

s b

y U

n k n o w

n A

u th

o r is

lic e

n s e d u

n d e r C

C B

Y

This Photo by Unknown Author is licensed under CC BY-SA

Glossary: What is Beta?

• Beta is a measure of volatility of returns of stock relative to the overall

market.

• If we plot returns of an individual stock against market returns, e.g. S&P

500 Index, Beta is equal to the slope of the line (see next page)

Glossary: What is Beta?

y = 0.7808x – 0.004

-5.0%

-4.0%

-3.0%

-2.0%

-1.0%

0.0%

1.0%

2.0%

3.0%

-6.0% -4.0% -2.0% 0.0% 2.0% 4.0% 6.0%

M a rk

e t

Indiv Stock

Field: Indiv Stock and Field: Market appear highly correlated.

Other types of regression

This Photo by Unknown Author is licensed under CC BY-SA

T h is

P h o to

b y U

n k n o w

n A

u th

o r is

lic e n s e d u

n d e r C

C B

Y -S

A

Polynomial regression

3-D regression movie

Activity 6: Salary regression model

• We will look at a simple model of how salary is related to years of work

experience.

• Data for this activity in Exploratory is Salary_Data.csv

• Open Exploratory and create a new project called Salary analysis

• Use the Data Frames menu to load the Salary_Data.csv file and save it

Activity 6: Salary regression model

• The Summary in Exploratory shows the distribution of the two variables

• Click on the Analytics menu (in Green)

• Go to the model ‘Type’ menu

• Choose ‘Linear regression’ as the type

of model you want

• Choose ‘Salary’ as the Target variable

• Choose ‘YearExperience’ as the

predictor variable and run

Activity 6: Salary regression model

• Interpret the output in a general sense

• Click on ‘Coef. Table’ to see the values

of the coefficients for the regression

equation

• The equation is

• ෣𝑆𝑎𝑙𝑎𝑟𝑦 = 25,792 + 9,449 YearsExperience

• You can make estimates from this by

substituting numbers for Years of

experience, e.g. 5 years of experience

gives you an estimate of

• ෣𝑆𝑎𝑙𝑎𝑟𝑦 = 25,792 +9,449*5 = $73,037

• You will learn more detail on this in week 9 of

STAM4000

Finance applications of big data and

predictive analytics: risk & return

FINM4100

Analytics in Accounting,

Finance and Economics

Week 10

Lesson Learning Outcomes

1 Define risk and return

2 Explore different ways of measuring risk and return

3 Investigate factors influencing risk and return

4 Performing portfolio analytics and optimisation

Why Build Models?

“Just because you

have more data

doesn’t mean that

you’re going to make

better decisions.”

Models encapsulate

patterns that exist in

data, helping us make

sense of them.Christina Zhu Assistant Professor of Accounting Wharton School of the University of Pennsylvania

Software for today

1. Google Colab

• Either

A. watch the teacher demonstrate analytics and accounting in python OR

B. you can run the python scripts yourself in Google Colab

• If you want to run the code provided, make sure you have access (signed in) to Google Colab https://colab.research.google.com

2. Exploratory

A. watch the teacher demonstrate analytics and accounting in Exploratory OR

B. run each step yourself online (access is explained on the next slide)

The risk return relationship is one of

the most fundamental relationships in

all of finance

• Return is a measure of the amount

earned by owning an asset

• Risk is a measure of the variability of

that return

To earn more return, an asset owner

must be prepared to accept more risk

The Risk Return Relationship

Photo by Parker Johnson on Unsplash

All investments carry risk, some more than others.

Risk & Return

Cash is generally low

risk. Suitable for investors

who have a short-term

investment outlook or low

tolerance for risk.

Shares are the most

volatile asset class, but

historically over long

periods of time have

achieved on average the

highest returns.

Risk and return in Australia

Risk and Return for Australian Shares & Bonds from 1974 to 2009

High return, high risk

Medium return, medium risk

Low return, low risk

Average

return

Std

14.34% 21.89%

10.14% 7.66%

9.73% 4.33%

How do we measure risk and return?

Return is a

measure of the

earnings made on

an asset

Risk is a measure

of the variability in

earnings made on

an asset

Dollar terms ($)

Percentage terms

(%)

Standard deviation

Coefficient of

variation

Beta

Dollar terms ($)

Percentage terms

(%)

• Let’s review the measures of standard deviation and

coefficient of variation

• We saw Beta in week 8

Glossary 1: Variance and Standard

deviation as measures of variability

• Measures the squared difference of a data set relative to its mean.

Variance

• Measures the spread of a data set relative to its mean.

Standard deviation

Recall from STAM4000 that

Hence, standard deviation is used a

measure of financial risk

Formulas for the variance &

standard deviation

N = population size

n = sample size

𝜇 = population mean (average)

ҧ𝑥 = sample mean (average)

Population Sample

Variance 𝜎2= σ x−𝜇 2

𝑁

𝑠2= σ x− ҧ𝑥 2

(n−1)

Standard deviation σ = 𝜎2 s = 𝑠2

11

Use 𝑠2 and s, respectively, as we have a sample.

First, we need ҧ𝑥 = σ 𝑥

𝑛 =

6.9−4.8+2.3+2.2+0.6

6 = 1.68%

𝑠2= σ 𝑥− ҧ𝑥 2

(𝑛−1) so we have

Example of STDEV of returns for the

S&P 500

Month Return

October 2021 6.9%

September 2021 -4.8%

August 2021 2.9%

July 2021 2.3%

June 2021 2.2%

May 2021 0.6%

Returns for S&P 500, May 2021-October 2021

𝑠2= 6.9−1.68 2+ −4.8 −1.68 2+ 2.9−1.68 2+ 2.3−1.68 2+ 2.2−1.68 2+ 0.6−1.68 2

(6 −1) =14.5

Standard deviation, s = 14.5 = 3.8%

https://www.businessinsider.com.au/what-is-standard-deviation

Standard deviation measu