A walk through imbalanced classes in machine learning through a visual cheat sheet

There are many detailed articles explaining the problem of imbalanced training samples and how to cope up with it. In this article, I summarize the understanding of the problem into a visual cheat sheet. I often find it useful as it comes handy for me whenever I have to revert back to the basic definitions (or I have an interview lined up). The cheat sheet below starts with the background on why accuracy doesn't always give a correct insight related to your classification algorithm and then moves on to defining other meaningful performance metrics. The cheat sheet then provides an example showing how to calculate those metrics for a three-class classification problem.