Last month, we looked at how we can take our data and, using machine learning, categorize the data by groupings we’d already decided upon—spam, ham, fraudulent, good, etc. This usage is a great way to start, but you might quickly find you want to take things further. Perhaps this means analyzing the data differently, to come up with new relationships you hadn’t previously considered, or maybe you want to improve the accuracy of your results by avoiding some common pitfalls in machine learning. These are the questions we are addressing today, starting with a different approach to machine learning—unsupervised learning.