The Machine Learning Workflow

Beginners think machine learning means choosing a clever algorithm. Professionals know that choosing the algorithm is often the smallest part of the job. Every serious project, at every company, follows the same seven steps. Learn them now, and every lesson after this will fit neatly into place.

• The algorithm is the small part
• Every project has seven steps
• This map guides the whole course

Seven steps, one loop

Step one, the question. What exactly are we predicting, and how will it help? Step two, get the data. Step three, clean it, because real data is messy. Step four, build features, the inputs the model will learn from. Step five, train a model. Step six, evaluate it honestly, on data it has never seen. Step seven, deploy it, so real people can use it. And then the most important part. The results send you back to the beginning, because the world keeps changing. It is a loop, not a line.

Where the time goes

Here is something that surprises people. Most of the time in a real project does not go to training. It goes to the early steps, understanding the problem, finding the data, and cleaning it. Training a model might take minutes. Getting the data right can take weeks. And a simple model on great data usually beats a brilliant model on poor data. That is why this course spends real time on data, and not only on algorithms.

Summary