Lesson 2 · Foundations

What is Machine Learning?

Machine Learning (ML) is a way to build systems that learn patterns from examples and use those patterns to make predictions or decisions on new data.

Why do we need Machine Learning?

Many real-world problems contain too many patterns to describe with a fixed list of rules. Consider an email spam filter. A message can be spam because of its words, sender, links, formatting, timing, and many combinations of these signals.

We could manually write hundreds or thousands of rules, but those rules would be difficult to maintain. Machine learning takes a different approach: learn from examples of messages that people have already labeled as spam or not spam.

How Machine Learning works

At a high level, an ML system follows a simple loop: collect examples, represent the examples as data, train a model to find patterns, then use the trained model to make predictions on data it has not seen before.

A concrete example

Imagine that we want to predict whether a support ticket should go to the billing team. We start with historical tickets that have already been routed. The examples teach the model which words and patterns are associated with billing requests.

# A tiny rule-based example to understand the idea

def route_ticket(message):
    text = message.lower()

    if "payment" in text or "refund" in text:
        return "Billing"

    if "password" in text or "login" in text:
        return "Account Support"

    return "General Support"


ticket = "I need a refund for my payment"

print(route_ticket(ticket))

This first example is intentionally rule-based, not machine learning. That distinction is important. We wrote the conditions ourselves. ML changes the learning step: instead of manually specifying every condition, an algorithm learns parameters from training examples.

From rules to learning

Here is the key difference. In a traditional program, a developer writes the rules and the computer applies them. In machine learning, we provide data and a learning algorithm; the resulting model captures patterns that can be applied to new inputs.

Training data, model, and prediction

1. Training data

Training data is the collection of examples used to teach a model. In supervised learning, each example usually includes an input and a known answer called a label.

2. Model

A model is the learned mathematical representation of patterns in the training data. Different algorithms learn different kinds of patterns.

3. Prediction

After training, we pass new input into the model. The model produces a prediction, classification, score, or another useful output.

Three common types of Machine Learning

Recap

• Machine learning lets systems learn useful patterns from data.
• Training examples are used to create a model.
• The trained model can make predictions on new data.
• Supervised, unsupervised, and reinforcement learning are three major ML approaches.
• Machine learning is one important part of the broader field of AI.