MSBA 7027 Machine Learning

Overview

Zhengli Wang

Faculty of Business and Economics

The University of Hong Kong
2024

1
Welcome to Machine Learning

https://www.smartdatacollective.com/5-ways-machine-learning-is-changing-the-way-businesses-operate/
2
Overview of ML

• ML Trend
• Google, McKinsey, Harvard Business Review, Forbes

• ML Applications in different industries

• What is ML

• Class Logistics

3
Machine Learning Trend: Google
Google Trends
Term: Machine Learning
Time Range: 2004 – 2023

4
Fundamental Reason of Popularity
ML/AI Decreases Cost & Increases Revenue

Source: Mckinsey

5
Firms’ Reaction

• 76% of organization: Prioritize AI/ML over other IT initiatives

• 83% of enterprises: Have increased their budgets for AI and machine

learning year-over-year

• 83% of IT leaders: AI & ML is transforming customer engagement and

69% say it is transforming their business.

Source: Forbes

6
If you master ML, how you can contribute
• Improving customer loyalty and retention.
• Automating finance.
• Detecting fraud.
• Smoother supply chains

Source: Harvard
Business Review
7
Machine Learning Applications
Across Industries

• Online Services
• Finance
• Retail
• Healthcare
• Supply Chain/Logistics
• Real Estate
• Others (Education, Manufacturing, etc.)

8
Machine Learning Applications
Online Services: Recommendation

9
Machine Learning Applications
Online Services: Product Bundling Dynamic Pricing

https://www.price2spy.com/blog/dynamic-pricing-explained-benefits-strategies-and-examples/

10
Machine Learning Applications
Finance: Fraud Detection Trading Advisory / Algo Trading

https://www.bluegranite.com/blog/detecting-financial-fraud-with-machine-learning
https://www.datacamp.com/community/tutorials/finance-python-trading 11
Machine Learning Applications
Retail: Customer Churn Modeling Customer Segmentation & Market Research

Source: https://cio-wiki.org/wiki/Customer_Churn
Source: https://gritglobal.io/blog/5-common-customer-segmentation-mistakes/
12
 Machine Learning Applications
Healthcare: Disease Identification / Diagnosis Disease Prediction

Source: https://ai.myesr.org/education/
13
Machine Learning Applications
Supply Chain/Logistics: Demand Estimation Delivery Prediction

https://www.forbes.com/sites/louiscolumbus/2018/06/11/10-ways-machine-learning-is-revolutionizing-supply-chain-management/?sh=787ad2303e37
https://medium.com/@ODSC/ai-as-the-ultimate-disrupter-in-logistics-how-to-manage-last-mile-costs-c4874e8f2ea0 14
Machine Learning Applications
Real Estate: Housing Price Prediction

Source: https://github.com/RohitLearner/House-Prices-Visualization-Prediction
15
What is Machine Learning?
Making the computer able to learn without being explicitly programmed.

data contains outcome Supervised Unsupervised observe data without outcomes

variable to guide the and describe how the data are
learning process
Learning Learning
organized or clustered

Reinforcement
Learning

The algorithm learns in an interactive

environment by trial and error
16
Supervised Learning
Given: A dataset that contains n samples
𝑥 1 ,𝑦 1 ,… 𝑥 𝑛 ,𝑦 𝑛

Goal: given 𝑥 𝑛+1 , predict 𝑦 𝑛+1

• Regression: Outcomes are continuous values

• Tomorrow’s temperature / Stock price / House price
• Linear Regression, Ridge/LASSO Regression
• Classification: Outcomes are categorical values
• Customer Purchase or No purchase / Diabetes or Not / Spam or No Spam (Email)
• Logistic Regression, LDA/QDA, Linear SVM

17
Unsupervised Learning
Given: A dataset with no labels
𝑥 1 ,𝑥 2 ,…,𝑥 𝑛

Goal: find interesting structures / patterns in the data

• PCA
• Clustering
• K-means
• Hierarchical

18
Reinforcement Learning
The algorithm collects data interactively

Try the strategy Improve the strategy

and collect Data based on the
Training
feedback
collection feedback

19
Most relevant in business: Supervised Learning

This is what we will focus on in this course

20
In MSBA7002 Topics in This Course (MSBA7027)

Linear/Logistic Regression K nearest neighbors (KNN)

Linear SVM Nonlinear Regression: Splines
LDA/QDA Nonlinear SVM
PCA Machine Learning System Design
Clustering Tree, Bagging, Random Forest,
Gradient Boosting Machine
Stacking
Interpretable Machine Learning
Recommender System

21
Class Logistics

22
Pre-requisite

MSBA7002 Business Statistics

Linear algebra

Statistics and probability

Programming experiences with R or Python

23
Tentative Schedule
Topic Content Lecture
1 ML Overview, Class Logistics ~1-4
2 K-nearest neighbors
3 Nonlinear Methods in Regression: Splines
4 Nonlinear Methods in SVM
5 Machine Learning System Design ~5-7
6 Decision Tree, Bagging, Random forest
7 Gradient Boosting Machines
8 Stacking / Ensemble Learning ~8-10
9 Interpretable Machine Learning
10 Recommender Systems
11 Preview of Deep Learning

24
Tentative Schedule

TAs:
Zhao Yuwei (Vera): [email protected]
Zhang Mingming (Eric): [email protected]
Hu Yue: [email protected]
Lyu Xian: [email protected]

If you have any questions, please email the TAs for fast response.

If you need to talk to the instructor, please do so before or during class break / during
instructor’s office hour.

25
Tentative Schedule

Two Tutorials (Optional):

1st Tutorial (after Lec 4):
Dec 10, 2024 (Tue) 2:00 - 4:00pm or Dec 11, 2024 (Wed) 6:30 - 8:30pm

2nd Tutorial (after Lec 8):

Jan 7, 2024 (Tue) 2:00 - 4:00pm or Jan 8, 2024 (Wed) 6:30 - 8:30pm

Weekly Office Hours: Wed 3:30-6:00pm

Note: 1st class for Section A/B: Nov 28; 1st class for Section C: Nov 27

26
Programming Language
• Why we use R in this course
• Smooth transition and learning experience from MSBA 7002

• Used by the best data scientists in the world.

• Most popular tool among professional ML practitioners

• Powerful: breadth of techniques it offers in third-party packages.

• Has more techniques than any other platforms

• Advanced: state-of-the-art
• New algorithms developed & released by academics

• Free & open-source

• Note: Python is also allowed in this course.

• But R is encouraged because we will be doing many demonstrations using R
27
Reference Textbooks

ISLR HMLR

28
Grades

Attendance & Participation 20%

Assignment 1 20% Due Dec 13, released by the week of Dec 4
Assignment 2 20% Due Jan 13, released by the week of Dec 24
Final Exam 40% On Jan 17, tentative 6:30PM – 9:00PM

Absence of Final Exam will automatically result in an F grade.

The TAs and the instructor reserve the rights to award bonus points for students
who demonstrate exceptional understanding of the class material

29
Assignments: General Rules

• Plagiarism will NOT be tolerated, it will automatically result in an F grade

• Don’t look at solutions or code that are not your own; everything you submit
should be your own work

• Don’t share your solution code with others; however discussing ideas or general
strategies is fine. Indicate in your submissions anyone you discussed/worked with

30
Expectation for this course
• Lots of useful tools / material

• Will be challenging
• Lots of coding involved
• Self-study, adapt to new methods / new packages
• Patience in debugging

• To facilitate learning, course designed to

• Contain basic & optional content
• Will do a lot of coding demonstrations in class

• Be mentally prepared
• Most Important: Review material after every lecture, put in your best effort –
You will learn lots of useful material and you won’t regret putting in the effort!
31
Welcome to MSBA 7027

Learn Stuff & Have Fun!

32