Introduction

Introduction

A machine learning project has two main ingredients: a learning algorithm and data. So there are only two things that can really go wrong — a bad algorithm or bad data. This topic is about the second, which is by far the more common problem in practice.

A useful saying in ML is "garbage in, garbage out." A model can only learn what the data show it. If the examples are too few, unfair, messy or about the wrong things, even the smartest algorithm will learn the wrong lessons.

We will look at four kinds of bad data:

1. not enough data;
2. data that don't represent the real world;
3. poor-quality (noisy, wrong, missing) data;
4. irrelevant features.