MACHINE LEARNING FOR BUSINESS
INHOUD
Lecture 1: Introduction to Machine Learning ..................................................................................................................... 2
Introduction to Machine Learning ................................................................................................................................. 3
Data types.................................................................................................................................................................... 5
Modeling ...................................................................................................................................................................... 6
Lecture 2: The CRISP-DM Process .................................................................................................................................... 8
Business Understanding............................................................................................................................................... 8
Data Understanding ..................................................................................................................................................... 9
Data Preparation .......................................................................................................................................................... 9
Modeling .................................................................................................................................................................... 14
Evaluation .................................................................................................................................................................. 15
Deployment ............................................................................................................................................................... 15
Lecture 3: Classification and Decision Trees .................................................................................................................. 16
Classification ............................................................................................................................................................. 16
Decision Trees ........................................................................................................................................................... 18
Lecture 4: Overfitting & Generalization ........................................................................................................................... 22
Lecture 5: Evaluating Classifiers .................................................................................................................................... 28
Logistic Regression .................................................................................................................................................... 28
Evaluating Classification Models ................................................................................................................................ 30
Visualizing Model Performance................................................................................................................................... 33
Lecture 6: Naive Bayes & SVMS ...................................................................................................................................... 36
Naive Bayes ............................................................................................................................................................... 36
Support Vector Machines ........................................................................................................................................... 40
Random Forest........................................................................................................................................................... 42
Lecture 7: Unsupervised Learning: Similarity, Neighbors and Clusters ............................................................................ 45
Lecture 8: Natural Language Processing ......................................................................................................................... 51
Text Mining ................................................................................................................................................................. 51
Association Rule Mining ............................................................................................................................................. 56
Lecture 9: Recommender Systems ................................................................................................................................. 60
Recommender Systems ............................................................................................................................................. 60
Recommendation as a ML problem ............................................................................................................................ 61
Recommendation Algorithms: Data Perspective ......................................................................................................... 65
Lecture 10: Deep Learning ............................................................................................................................................. 69
Neural Networks ........................................................................................................................................................ 69
Deep Learning ............................................................................................................................................................ 75
Lecture 11: Ensembling.................................................................................................................................................. 83
1
,LECTURE 1: INTRODUCTION TO MACHINE LEARNING
PRACTICAL INFORMATION
• Who?
o Lien Michiels ()
▪ PhD in Computer Science @ University of Antwerp
• On “Evaluation of Recommender Systems”
▪ Previously worked at various Tech Startups as a Machine Learning Engineer/Data Engineer/Data
Scientist/Solutions Architect
▪ Guest Professor @ FBE, University of Antwerp
• Senior Researcher @ imec-SMIT Vrije Universiteit Brussel and FWET University of
Antwerp
o Yifan He ()
• What?
o Lectures - ML Theory (Tuesdays)
▪ ML Basics
• Supervised Learning - Classification (Models & Evaluation)
• Unsupervised Learning - Clustering and Association Rule/Frequent Set Mining
▪ Advanced ML Topics
• Artificial Neural Networks & Deep Learning
• Resembling
▪ Applications & Application-specific ML
• Natural Language Processing (NLP) = text mining
• Recommender Systems (RecSys)
• …
o Tutorials - Hands-on ML (Thursdays)
▪ Programming Basics in Python
• 2-3 tutorials, goes very quick so make sure you are up to date with the course material
▪ Machine- Learning in Python
▪ Tips & Tools for ML in Python
• Why? At the end of this course, you should be able to…
o solve business problems by applying data-analytical thinking
o decide on the appropriate data modeling technique
o set up an appropriate model evaluation
o implement proof-of-concept (POC) data science solutions
• How? We will have…
o 10 lectures on ML theory
o 8 tutorial sessions on Python & ML in practice
▪ Make sure to install Python!
o The Data Science Challenge
▪ Real-world classification dataset
▪ Opportunity to demonstrate the theoretical and practical knowledge you gained in this course
▪ Previous year, this was all classification of a banking dataset into fraud actions
o Exam: know what the key concepts mean and be able to apply these!
2
, • Contact Policy - How to Contact the Teaching Staff
o Lectures & Tutorials
▪ Ask questions during lectures (ML theory & logistics) and tutorials (Python, ML in Python, Data
Science Challenge). These are your primary support channels.
o Blackboard Forum (FAQs)
▪ Check the FAQs before emailing—many questions will already be answered there. It helps
everyone and avoids overtaxing the teaching staff.
o Email
▪ Only use email if you cannot use the two options above.
▪ Email replies may be delayed due to other responsibilities.
▪ If you must
• Send to both Prof. Michiels (me) & Yifan only => No emails to other ADM members
• Use your student email => No personal email accounts
• Include MLB or PML in the subject
• Keep it brief
• Data Science Challenge
o Again partnering with Crelan, More details soon!
o Time to start forming groups! Go to “Groepen” on Blackboard and enroll yourself into a group of 3!
▪ Only in exceptional cases will we allow groups of 2 (or individual assignments).
▪ Contact us if you think you need an exception!
INTRODUCTION TO MACHINE LEARNING
Loosely based on Chapter 1 & 2 Data Science for Business, Provost and Fawcett
You can buy the book if you like, but the book is from 2013, so a bit outdated. It’s showing its age
Slides will be enough to pass, but if you like it more to read a book, you can buy it
All these terms are used interchangeably, hence the image of a kind of spaghetti.
There is also no single definition for all these terms that everyone agrees on. The
table below discusses how we will use these terms in this course.
Key concept (Term) Explanation
Data Science • The process of extracting insights and knowledge from data using techniques from
statistics, computer science, and domain expertise.
• The broadest of the terms → for example also just a chart you make in Excel from a
dataset
• Used by people with backgrounds in statistics, computer science, mathematics,
economics, engineering, or business analytics. (So everyone, essentially )
Data Analytics • The process of examining, cleaning, transforming, and modeling data to discover
useful information and support decision-making.
• The most manual of the “data sciences”
o Hands on with data to discover useful information to present to the stakeholders
o Make sense of the world they are operating in
• Used by people with backgrounds in business, economics, marketing, finance, or
statistics.
3
, Data Mining • The process of discovering patterns, relationships, or anomalies in large datasets
using statistical and computational methods.
• The oldest of the “data sciences”
o Seeking for meaning, patterns, relationships, …
o Literally mine all of the data
• Used by people with backgrounds in business intelligence, marketing, statistics, or
computer science.
You finally make the decision
Machine Learning • The process of building algorithms that can learn from data and make predictions or
decisions without being explicitly programmed.
• The highest value of the “data sciences” (in my personal opinion), because of its
scalability
• Used by people with backgrounds in computer science, data science, (digital business)
engineering, mathematics, or physics.
o Less about business understanding
o About data understanding
o For example predict what the user needs, or decide if AI can answer the
question based on the existing database, or if you need a human to answer it
Algorithm finally makes the decision
Artificial • The art of creating systems that can perform specific tasks typically requiring human
Intelligence intelligence, such as reasoning, learning, and perception.
• Currently the hippest of the “data sciences”
• Used by people with backgrounds in computer science, robotics, cognitive science, or
philosophy of mind.
• Artificial General Intelligence
o The quest for a theoretical form of AI that can perform any intellectual task a
human can, with general reasoning and learning abilities across domains.
o = AI with general capabilities, not task or domain specific tasks (→ ALL tasks)
o The absolute hippest of the “data sciences”
o Most used by Sam Altman, Ilya Sutskever, and other AI prophets.
Big Data • Extremely large and complex datasets that require advanced tools and infrastructure to
store, process, and analyze.
• A bit of an outdated term in 2025.
o It was a big term in 2000-2010, but nowadays the majority of the data we story is
garbage, but we have the storage to do that, so we don’t have an incent to clean
it and it’s ‘normal’ to have that storage
• Used by people with backgrounds in computer science, information systems, cloud
computing, or data engineering.
Data Engineering • A discipline of designing and building systems for collecting, storing, and processing
(big) data efficiently and reliably.
• Underrated but in my experience, the number one reason why data science projects
succeed or fail.
• Used by people with backgrounds in software engineering, computer science, or
information technology.
• For example a classifier based on GPS data which failed because it was tested all on
Apple devices, but Samsung doesn’t have such a great GPS signal → Shows the
importance of data engineering
Machine Learning • The practice of implementing, deploying, and maintaining machine learning models
Engineering in production environments.
• One of the hardest things to do well, in my experience.
• Used by people with backgrounds in computer science, software engineering, or applied
mathematics.
4
INHOUD
Lecture 1: Introduction to Machine Learning ..................................................................................................................... 2
Introduction to Machine Learning ................................................................................................................................. 3
Data types.................................................................................................................................................................... 5
Modeling ...................................................................................................................................................................... 6
Lecture 2: The CRISP-DM Process .................................................................................................................................... 8
Business Understanding............................................................................................................................................... 8
Data Understanding ..................................................................................................................................................... 9
Data Preparation .......................................................................................................................................................... 9
Modeling .................................................................................................................................................................... 14
Evaluation .................................................................................................................................................................. 15
Deployment ............................................................................................................................................................... 15
Lecture 3: Classification and Decision Trees .................................................................................................................. 16
Classification ............................................................................................................................................................. 16
Decision Trees ........................................................................................................................................................... 18
Lecture 4: Overfitting & Generalization ........................................................................................................................... 22
Lecture 5: Evaluating Classifiers .................................................................................................................................... 28
Logistic Regression .................................................................................................................................................... 28
Evaluating Classification Models ................................................................................................................................ 30
Visualizing Model Performance................................................................................................................................... 33
Lecture 6: Naive Bayes & SVMS ...................................................................................................................................... 36
Naive Bayes ............................................................................................................................................................... 36
Support Vector Machines ........................................................................................................................................... 40
Random Forest........................................................................................................................................................... 42
Lecture 7: Unsupervised Learning: Similarity, Neighbors and Clusters ............................................................................ 45
Lecture 8: Natural Language Processing ......................................................................................................................... 51
Text Mining ................................................................................................................................................................. 51
Association Rule Mining ............................................................................................................................................. 56
Lecture 9: Recommender Systems ................................................................................................................................. 60
Recommender Systems ............................................................................................................................................. 60
Recommendation as a ML problem ............................................................................................................................ 61
Recommendation Algorithms: Data Perspective ......................................................................................................... 65
Lecture 10: Deep Learning ............................................................................................................................................. 69
Neural Networks ........................................................................................................................................................ 69
Deep Learning ............................................................................................................................................................ 75
Lecture 11: Ensembling.................................................................................................................................................. 83
1
,LECTURE 1: INTRODUCTION TO MACHINE LEARNING
PRACTICAL INFORMATION
• Who?
o Lien Michiels ()
▪ PhD in Computer Science @ University of Antwerp
• On “Evaluation of Recommender Systems”
▪ Previously worked at various Tech Startups as a Machine Learning Engineer/Data Engineer/Data
Scientist/Solutions Architect
▪ Guest Professor @ FBE, University of Antwerp
• Senior Researcher @ imec-SMIT Vrije Universiteit Brussel and FWET University of
Antwerp
o Yifan He ()
• What?
o Lectures - ML Theory (Tuesdays)
▪ ML Basics
• Supervised Learning - Classification (Models & Evaluation)
• Unsupervised Learning - Clustering and Association Rule/Frequent Set Mining
▪ Advanced ML Topics
• Artificial Neural Networks & Deep Learning
• Resembling
▪ Applications & Application-specific ML
• Natural Language Processing (NLP) = text mining
• Recommender Systems (RecSys)
• …
o Tutorials - Hands-on ML (Thursdays)
▪ Programming Basics in Python
• 2-3 tutorials, goes very quick so make sure you are up to date with the course material
▪ Machine- Learning in Python
▪ Tips & Tools for ML in Python
• Why? At the end of this course, you should be able to…
o solve business problems by applying data-analytical thinking
o decide on the appropriate data modeling technique
o set up an appropriate model evaluation
o implement proof-of-concept (POC) data science solutions
• How? We will have…
o 10 lectures on ML theory
o 8 tutorial sessions on Python & ML in practice
▪ Make sure to install Python!
o The Data Science Challenge
▪ Real-world classification dataset
▪ Opportunity to demonstrate the theoretical and practical knowledge you gained in this course
▪ Previous year, this was all classification of a banking dataset into fraud actions
o Exam: know what the key concepts mean and be able to apply these!
2
, • Contact Policy - How to Contact the Teaching Staff
o Lectures & Tutorials
▪ Ask questions during lectures (ML theory & logistics) and tutorials (Python, ML in Python, Data
Science Challenge). These are your primary support channels.
o Blackboard Forum (FAQs)
▪ Check the FAQs before emailing—many questions will already be answered there. It helps
everyone and avoids overtaxing the teaching staff.
o Email
▪ Only use email if you cannot use the two options above.
▪ Email replies may be delayed due to other responsibilities.
▪ If you must
• Send to both Prof. Michiels (me) & Yifan only => No emails to other ADM members
• Use your student email => No personal email accounts
• Include MLB or PML in the subject
• Keep it brief
• Data Science Challenge
o Again partnering with Crelan, More details soon!
o Time to start forming groups! Go to “Groepen” on Blackboard and enroll yourself into a group of 3!
▪ Only in exceptional cases will we allow groups of 2 (or individual assignments).
▪ Contact us if you think you need an exception!
INTRODUCTION TO MACHINE LEARNING
Loosely based on Chapter 1 & 2 Data Science for Business, Provost and Fawcett
You can buy the book if you like, but the book is from 2013, so a bit outdated. It’s showing its age
Slides will be enough to pass, but if you like it more to read a book, you can buy it
All these terms are used interchangeably, hence the image of a kind of spaghetti.
There is also no single definition for all these terms that everyone agrees on. The
table below discusses how we will use these terms in this course.
Key concept (Term) Explanation
Data Science • The process of extracting insights and knowledge from data using techniques from
statistics, computer science, and domain expertise.
• The broadest of the terms → for example also just a chart you make in Excel from a
dataset
• Used by people with backgrounds in statistics, computer science, mathematics,
economics, engineering, or business analytics. (So everyone, essentially )
Data Analytics • The process of examining, cleaning, transforming, and modeling data to discover
useful information and support decision-making.
• The most manual of the “data sciences”
o Hands on with data to discover useful information to present to the stakeholders
o Make sense of the world they are operating in
• Used by people with backgrounds in business, economics, marketing, finance, or
statistics.
3
, Data Mining • The process of discovering patterns, relationships, or anomalies in large datasets
using statistical and computational methods.
• The oldest of the “data sciences”
o Seeking for meaning, patterns, relationships, …
o Literally mine all of the data
• Used by people with backgrounds in business intelligence, marketing, statistics, or
computer science.
You finally make the decision
Machine Learning • The process of building algorithms that can learn from data and make predictions or
decisions without being explicitly programmed.
• The highest value of the “data sciences” (in my personal opinion), because of its
scalability
• Used by people with backgrounds in computer science, data science, (digital business)
engineering, mathematics, or physics.
o Less about business understanding
o About data understanding
o For example predict what the user needs, or decide if AI can answer the
question based on the existing database, or if you need a human to answer it
Algorithm finally makes the decision
Artificial • The art of creating systems that can perform specific tasks typically requiring human
Intelligence intelligence, such as reasoning, learning, and perception.
• Currently the hippest of the “data sciences”
• Used by people with backgrounds in computer science, robotics, cognitive science, or
philosophy of mind.
• Artificial General Intelligence
o The quest for a theoretical form of AI that can perform any intellectual task a
human can, with general reasoning and learning abilities across domains.
o = AI with general capabilities, not task or domain specific tasks (→ ALL tasks)
o The absolute hippest of the “data sciences”
o Most used by Sam Altman, Ilya Sutskever, and other AI prophets.
Big Data • Extremely large and complex datasets that require advanced tools and infrastructure to
store, process, and analyze.
• A bit of an outdated term in 2025.
o It was a big term in 2000-2010, but nowadays the majority of the data we story is
garbage, but we have the storage to do that, so we don’t have an incent to clean
it and it’s ‘normal’ to have that storage
• Used by people with backgrounds in computer science, information systems, cloud
computing, or data engineering.
Data Engineering • A discipline of designing and building systems for collecting, storing, and processing
(big) data efficiently and reliably.
• Underrated but in my experience, the number one reason why data science projects
succeed or fail.
• Used by people with backgrounds in software engineering, computer science, or
information technology.
• For example a classifier based on GPS data which failed because it was tested all on
Apple devices, but Samsung doesn’t have such a great GPS signal → Shows the
importance of data engineering
Machine Learning • The practice of implementing, deploying, and maintaining machine learning models
Engineering in production environments.
• One of the hardest things to do well, in my experience.
• Used by people with backgrounds in computer science, software engineering, or applied
mathematics.
4
