Machine Learning Types: Hands-On Exercises

Exercise 1: Basic Identification

Instructions: For each scenario below, identify:

Is this supervised or unsupervised learning?
If supervised, is it classification or regression?

Scenario 1.1

A bank wants to automatically approve or reject loan applications based on applicant's credit score, income, and employment history. They have 10,000 historical applications with known outcomes.

Type:

Sub-type (if applicable):

Hint: Do we have target labels (known outcomes) from the past?

Scenario 1.2

An e-commerce company wants to understand different types of customers based on their browsing behavior, purchase patterns, and demographics. They don't know what customer groups exist.

Type:

Sub-type (if applicable):

Hint: Are we trying to discover hidden patterns without known labels?

Scenario 1.3

A real estate company wants to estimate the selling price of houses based on size, location, age, and number of bedrooms. They have sales data from the past 5 years.

Type:

Sub-type (if applicable):

Hint: Are we predicting a continuous numerical value?

Exercise 2: Making Predictions with Simple Rules

Instructions: You will use existing data to make predictions about new cases using simple rules. No complex algorithms needed - just logical thinking!

🎯 Example: How to Make Rule-Based Decisions

Sample Data - Student Grade Prediction:

Student	Study Hours	Attendance %	Previous Test	Final Grade
Alice	8	95%	85	A
Bob	3	70%	60	C
Carol	6	88%	78	B
David	2	60%	45	D

Example Rules I can create:

Rule 1: If Study Hours ≥ 7 AND Attendance ≥ 90% → Grade A
Rule 2: If Study Hours ≤ 3 AND Previous Test ≤ 60 → Grade C or D
Rule 3: If Study Hours = 5-6 AND Attendance ≥ 80% → Grade B

New Student to Predict: Emma: 7 Study Hours, 92% Attendance, Previous Test: 82

My Prediction: Grade A (because she meets Rule 1: Study Hours ≥ 7 AND Attendance ≥ 90%)

Dataset 2.1: Sales Performance Classification

Business Goal: Predict whether a new product will have High, Medium, or Low sales performance.

Product ID	Season	Marketing Spend	Price	Competitor Price	Sales Performance
P001	Summer	$5,000	$29.99	$34.99	Medium
P002	Winter	$8,000	$49.99	$45.99	Low
P003	Fall	$3,000	$19.99	$22.99	High
P004	Spring	$6,000	$39.99	$41.99	Medium
P005	Summer	$7,500	$24.99	$28.99	High
P006	Winter	$2,000	$59.99	$55.99	Low

Step 1: What patterns do you notice that affect sales performance?

Step 2: Create a simple rule for predicting sales performance

NEW PRODUCT: P007 - Season: Spring, Marketing: $4,500, Price: $34.99, Competitor Price: $38.99

Step 3: Predict sales performance for the new product

Explain your prediction using your rule:

Dataset 2.2: Customer Segmentation

Business Goal: Group customers into 2 types and predict which group a new customer belongs to.

Customer ID	Age	Income	Purchases/Year	Time on Site
C001	25	$45,000	12	15 min
C002	34	$75,000	8	22 min
C003	19	$25,000	3	5 min
C004	42	$95,000	15	18 min
C005	28	$55,000	6	8 min

Step 1: Create 2 customer groups from the data above

Group 1 Name: Which customers belong here?

Group 2 Name: Which customers belong here?

Step 2: Create a simple rule to distinguish the groups

NEW CUSTOMER: C006 - Age: 38, Income: $68,000, Purchases/Year: 11, Time on Site: 20 min

Step 3: Which group does the new customer belong to?

Explain why using your rule:

Dataset 2.3: Email Classification

Business Goal: Predict the category of a new email.

Email ID	Contains "Free"?	Contains "Click"?	From Known Sender?	Category
E001	Yes	Yes	No	Spam
E002	No	No	Yes	Important
E003	Yes	No	Yes	Promotional
E004	No	Yes	No	Spam
E005	No	No	Yes	Important
E006	Yes	Yes	Yes	Promotional

Step 1: Create rules for each email category

Spam Rule:

Important Rule:

Promotional Rule:

NEW EMAIL: E007 - Contains "Free": No, Contains "Click": Yes, From Known Sender: Yes

Step 2: Predict the category of the new email

Explain which rule applies:

Exercise 3: Business Decision Framework

Decision Framework

Scenario 3.1: Healthcare Analytics

A hospital wants to improve patient care and operational efficiency. They have patient data including demographics, medical history, treatment costs, length of stay, and patient satisfaction scores.

Multiple Questions to Explore:

"What types of patients do we treat?" (Unknown patient groups)
"Will this patient be readmitted within 30 days?" (Yes/No)
"How much will this patient's treatment cost?" (Dollar amount)

Question 3.1a Analysis: "What types of patients do we treat?"

Target Variable Present?

ML Approach:

Business Value:

Question 3.1b Analysis: "Will this patient be readmitted within 30 days?"

Target Variable Present?

ML Approach:

Business Value:

Question 3.1c Analysis: "How much will this patient's treatment cost?"

Target Variable Present?

ML Approach:

Business Value:

Machine Learning Types: Hands-On Exercises

Supervised vs Unsupervised & Classification vs Regression

Learning Objectives

Scenario 1.1

Scenario 1.2

Scenario 1.3

🎯 Example: How to Make Rule-Based Decisions

Dataset 2.1: Sales Performance Classification

Dataset 2.2: Customer Segmentation

Dataset 2.3: Email Classification

Decision Framework

Scenario 3.1: Healthcare Analytics

Question 3.1a Analysis: "What types of patients do we treat?"

Question 3.1b Analysis: "Will this patient be readmitted within 30 days?"

Question 3.1c Analysis: "How much will this patient's treatment cost?"