Cancer Classification - Malignant vs Benign

01 - The problem

A diagnostic dataset needed a classifier where a false negative carries far more cost than a false positive.

02 - Approach

Started with an interpretable Decision Tree Classifier as the baseline, then boosted with XGBoost and compared both on precision, recall and confusion-matrix behaviour rather than raw accuracy alone.

03 - Outcome

Boosting delivered a significant lift in prediction accuracy over the baseline tree while keeping the decision logic explainable to non-technical reviewers.