Underfitting and OverfittingConclusion
The journey toward a perfect model is almost always a delicate balancing act. If you build a model that is entirely too simple, it will stubbornly ignore your data and suffer from massive bias. If you build a model that is completely over-complicated, it will rigidly memorize every pointless detail and suffer from crushing variance.
Your job as a data scientist is to carefully test hyperparameter dials, like the value in K-Nearest Neighbors, to find the exact point where these two forces safely balance out!
Tying It All Together
To keep this chapter focused, we intentionally skipped over some incredibly powerful techniques for automatically reigning in wild, complex models—a process formally known as
But fear not! You will master all of those advanced techniques and more in our upcoming chapters. For now, keep moving forward!