An honest diagnosis model

In the early 1990s, researchers at the University of Wisconsin digitised images of fine-needle aspirates of breast masses and measured 30 properties of the cell nuclei: radius, texture, smoothness, concavity and more. Each of the 569 cases is labelled benign or malignant.

You'll build a regularised logistic regression from scratch, but the real subject is the protocol: lock away a test set, scale without leaking, tune with cross-validation, and report one honest number. In this dataset y = 1 means benign; we'll treat malignant as the positive class, because that's the one we must not miss.