Here is a (growing) collection of my thinking on machine learning, data science, and AI, written up as blog posts as I explore the space.
Class Imbalance. In classification problems (where we predict discrete outcomes such as loan default vs non-default, or customer churn vs retention), most ML frameworks perform best when classes are relatively balanced. However, in real-world business settings, this is often not the case. Class imbalance can therefore cause challenges for standard modelling approaches. Here, I briefly explore why this happens, how class imbalance can lead to misleading business outcomes, and how we can address it in practice.