Computer Science · Grade 10 · Data and Information Systems · Term 2

Machine Learning Basics

Explore fundamental concepts of machine learning, including supervised and unsupervised learning.

• Ontario Curriculum Expectations
• CS.HS.D.9
• CS.HS.D.10

About This Topic

Machine learning basics introduce students to algorithms that improve from experience with data. Supervised learning relies on labeled training data to predict outcomes, such as identifying handwritten digits from examples marked correct. Unsupervised learning finds patterns in unlabeled data, like grouping similar news articles by topics. These ideas fit Ontario's Grade 10 Computer Science curriculum in the Data and Information Systems unit, meeting standards CS.HS.D.9 and CS.HS.D.10 on data processing and analysis.

Students differentiate the learning types, examine prediction examples, and explain how training data shapes models. This topic connects data handling to real applications, from spam filters to customer segmentation, while building skills in critical evaluation of algorithms and data quality. It prepares students for advanced topics like ethics in AI.

Active learning works well for machine learning basics because students handle tangible datasets and see immediate results from changes. Sorting scenarios in pairs or adjusting mock training sets reveals how labels and data quality affect predictions, turning abstract processes into practical insights students retain.

Key Questions

1. Differentiate between supervised and unsupervised machine learning.
2. Analyze simple examples of how machine learning algorithms make predictions.
3. Explain the role of training data in machine learning models.

Learning Objectives

• Classify machine learning problems as either supervised or unsupervised learning based on the presence or absence of labeled data.
• Analyze simple datasets to predict an outcome using a given supervised learning algorithm, such as predicting house prices based on size.
• Explain the impact of data quality and quantity on the performance of a machine learning model.
• Compare and contrast the goals and methods of supervised and unsupervised learning algorithms.
• Design a basic training dataset for a simple classification task, identifying necessary features and labels.

Before You Start

Data Representation and Organization

Why: Students need to understand how data is structured in tables and lists to comprehend training datasets and features.

Basic Programming Concepts (Variables, Loops, Conditionals)

Why: Understanding fundamental programming constructs helps students grasp how algorithms process data and make decisions.

Key Vocabulary

Machine Learning	A field of artificial intelligence where computer systems learn from data without being explicitly programmed. The system improves its performance on a task with more experience.
Supervised Learning	A type of machine learning that uses labeled datasets to train algorithms. The algorithm learns to map inputs to outputs based on example input-output pairs.
Unsupervised Learning	A type of machine learning that uses unlabeled datasets to find patterns or structures. Algorithms identify relationships in data without predefined outcomes.
Training Data	The dataset used to train a machine learning model. It consists of input features and, for supervised learning, corresponding correct output labels.
Algorithm	A set of rules or instructions followed by a computer to solve a problem or perform a calculation. In machine learning, algorithms learn from data.

Watch Out for These Misconceptions

Common Misconception

Machine learning models think and reason like humans.

What to Teach Instead

Models detect statistical patterns in data, not understand context. Hands-on sorting activities let students manipulate data to see predictions stem from examples, not intelligence, building accurate mental models through trial and error.

Common Misconception

Supervised learning always outperforms unsupervised.

What to Teach Instead

Each suits different goals; unsupervised reveals hidden structures. Group challenges with unlabeled data show its value for exploration, helping students appreciate context via peer debate on task fit.

Common Misconception

Any data works as training data.

What to Teach Instead

Quality and relevance matter; poor data leads to bad predictions. Dataset tweaking in small groups demonstrates bias or noise effects, with discussions reinforcing data cleaning needs.

Active Learning Ideas

Pairs Sort: Learning Type Scenarios

Provide cards describing real-world tasks, such as 'predict house prices from sizes and locations' or 'group songs by listener habits'. Pairs sort cards into supervised or unsupervised piles and write one-sentence justifications for each. Follow with whole-class share-out to refine categories.

25 min·Pairs

Small Groups: Mock Training Data

Give groups a simple dataset, like animal features without labels. First, have them predict categories intuitively, then add labels for supervised practice and cluster without for unsupervised. Groups compare prediction accuracy and discuss training data impact.

35 min·Small Groups

Individual: Prediction Journal

Students receive printed examples of input data and model outputs. Individually, they journal how changing one training example alters predictions, then pair up to verify entries. Collect journals for feedback.

20 min·Individual

Whole Class: Visual Algorithm Demo

Use slides or free online tools like Teachable Machine to demo live predictions. Class votes on inputs, observes model updates with new training data, and notes supervised versus unsupervised shifts.

40 min·Whole Class

Real-World Connections

• Data scientists at Netflix use supervised learning algorithms trained on viewing history to recommend movies and shows to users, personalizing the viewing experience.
• Financial analysts employ unsupervised learning techniques to segment customers into groups with similar purchasing behaviors, allowing for targeted marketing campaigns by companies like Amazon.
• Medical researchers utilize machine learning to analyze patient data, identifying patterns that could predict disease outbreaks or personalize treatment plans for conditions like diabetes.

Assessment Ideas

Exit Ticket

Provide students with two scenarios: one describing a system that predicts house prices based on square footage and number of bedrooms, and another describing a system that groups customers by shopping habits. Ask students to identify which scenario uses supervised learning and which uses unsupervised learning, and to briefly explain why.

Quick Check

Present students with a small, simplified dataset (e.g., fruit images labeled 'apple' or 'orange'). Ask them to explain what 'training data' means in this context and how they would use it to teach a computer to identify apples. Then, ask them to describe a scenario where they might use unlabeled data to find patterns in fruit types.

Discussion Prompt

Pose the question: 'Imagine you are building a spam email filter. What kind of data would you need for training, and would this be supervised or unsupervised learning? Explain your reasoning.' Facilitate a class discussion where students share their answers and justify their choices.

Frequently Asked Questions

How do you differentiate supervised and unsupervised learning for grade 10?

Supervised uses labeled data for direct predictions, like teaching a model spam emails with 'spam' tags. Unsupervised finds patterns without labels, such as clustering fruits by color and size alone. Use everyday examples: supervised for quizzes with answers, unsupervised for sorting laundry by type. Visual charts and card sorts clarify the distinction quickly.

What role does training data play in machine learning models?

Training data teaches the model patterns for predictions. In supervised learning, labels guide outputs; more diverse, accurate data improves reliability. Poor data causes errors or bias. Students explore this by altering mock datasets and testing predictions, seeing direct links between input quality and model success across 50-100 examples.

What are simple examples of machine learning predictions?

Email spam detection (supervised: learns from labeled spam) or movie recommendations (unsupervised: groups similar viewers). Image recognition apps predict objects from photo training sets. For class, use fruit classification: supervised labels apples, unsupervised clusters by shape. These tie to daily tech, making concepts relatable without complex math.

How can active learning help students understand machine learning basics?

Active methods like pair sorts and group data challenges make abstract ideas concrete. Students physically categorize scenarios or tweak datasets, observing how labels change outcomes. This builds intuition faster than lectures, with collaboration exposing errors. Follow-up discussions connect experiences to key questions, boosting retention by 30-50% per studies on kinesthetic CS learning.