Breast cancer diagnosis using logistic regression on top predictive features

Breast cancer is among the most prevalent and life-threatening female cancers globally. Early and precise diagnosis is essential in enhancing patient survival. This research evaluates logistic regression to classify breast tumours as malignant or benign through a publicly available database of 569 cases. We centred on two sets of features: the top 5 and top 10 most predictive features for tumour size and shape irregularity. The logistic regression classifier attained an accuracy of about 94.7% using the top 5 features and 97.4% using the top 10 features, exhibiting sound performance with a smaller set of features. Visualization methods also attested to unique distribution patterns between benign and malignant cases. Our results demonstrate the promise of feature selection and simple but robust models for accurate breast cancer diagnosis. Ensemble methods and validation using an external dataset will be considered in future work to improve generalizability.