The Machine Learning Workflow

The big idea: A machine learning (ML) project is not a single magic button you press. It is a loop of clear stages that you walk through in order — and often walk through again and again — to turn a real-world question into a working, useful prediction system.

In this very first lesson we look at that whole loop from far above, like a map before a road trip. Every later lesson in this course zooms into one part of this map. Once you truly understand the shape of the journey, everything that follows will feel like filling in details you already expected. We will also meet the example project we reuse for the rest of the course: predicting apartment prices in Bishkek.

• What the seven stages of a machine learning project are, and the order they happen in.
• For each stage: what goes in, what comes out, and what commonly goes wrong.
• How to look at a real-world problem and decide whether it is regression (predict a number), classification (predict a category), or something else.
• When machine learning is the wrong tool, and a plain set of rules is better.
• Who does what on a team: the data scientist, the ML engineer, and the data analyst.
• Our running course example: predicting apartment prices in Bishkek.

When people first hear about machine learning, they often picture the middle step only: "the computer learns from data." That is real, but it is a tiny slice of the actual work. In practice, professionals spend most of their time around that step — deciding what problem to solve, hunting for data, cleaning up messy data, checking whether the result is any good, and keeping the system healthy after it goes live.