Introduction
Machine-learning models are computer algorithms that use data to make estimations (educated guesses) or decisions. Machine-learning models differ from traditional algorithms in how they're designed. When traditional computer software needs to be improved, people edit it. In contrast, a machine-learning algorithm uses data to get better at a specific task.
For example, spam filters use machine learning. Twenty years ago, spam filters didn't have many examples to learn from and weren't good at identifying what is and isn’t spam. As more spam has arrived and been labeled as junk by human users, the machine-learning algorithms have gained more experience and become better at their job.
Boots that fit
Throughout this module, we use an example scenario to explain key machine-learning concepts.
In this scenario, you own a shop that sells harnesses for avalanche-rescue dogs, and you’ve recently expanded to also sell doggy boots. Customers all seem to pick the correct harness sizes, but are constantly ordering doggy boots that are the wrong size. You know most customers buy harnesses and boots in the same transaction, which gives you an idea: perhaps you could approximate which doggy boots are the correct size, depending on the harness chosen. Then, you could warn customers if the boots they have selected are likely to be the correct size before they make the purchase.
During this module, we create a machine-learning model that implements this idea. Along the way, we use this scenario to introduce you to some basic machine-learning concepts and demonstrate how to use them in a practical setting.
Learning objectives
In this module, you'll:
- Explore how machine learning differs from traditional software.
- Create and test a machine-learning model.
- Load a model and use it with new data.
Prerequisites
None