Luminal breast cancers, which have oestrogen receptor (ER) positivity and human epidermal growth factor receptor 2 (HER2) negativity, are divided into luminal A and B subtypes. This study focuses on using Mueller matrix polarimetry to analyze polarised light interactions with complex breast tissues to determine luminal subtype. Polarimetric and clinical parameters from unstained breast core biopsies were analyzed and classified according to luminal subtypes via the BluePrint molecular assay. The results were used to train logistic regression, linear discriminant analysis, support vector machine, random forest, and XGBoost learning models to differentiate between subtypes. The extracted features of the model were ranked by importance using an RF classifier. It was found that LDA yielded the highest specificity of 78% with limited sensitivity of 40%, indicating it could identify luminal A well, but not luminal B. This study demonstrates that biopsies can be used to obtain biomarkers used towards distinguishing A and B subtypes, leading to further improvement in treatment and therapies for each subtype.