Overview
The Breast Cancer Wisconsin (Diagnostic) dataset holds 569 samples, each described by 30 features computed from digitised images of cell nuclei. The task is to tell benign tumours from malignant ones.
Accuracy is not the whole story here. A missed malignancy costs far more than a false alarm, so the evaluation puts recall alongside precision, F1 and the confusion matrix rather than stopping at a single score.
How it works
Load
569 samples, 357 benign and 212 malignant, loaded through scikit-learn.
Split and scale
An 80/20 train-test split, then StandardScaler gives every feature zero mean and unit variance.
Train
A LogisticRegression model is fitted on the training set.
Evaluate
Accuracy, precision, recall, F1 and a confusion matrix, with the plots and feature importance saved to a results folder.