Supervised Machine Learning with Python

Strategies for handling high bias

If you determine that you're suffering from a high bias problem, you can try making your model more complex by engineering more informative signal-rich features. For example, here, one thing you could try doing is creating new features that are polynomial combinations of your x1 so, you can create logit function of x1, and that would model our function perfectly. You can also try tuning some of the hyperparameters, for instance, KNNs, even though it's a high variance model, and it can become highly biased very quickly as you increase the k hyperparameter, and vice versa: