The Classification of Breast Cancer: The Effect of Hyperparameter Optimisation Towards the Efficacy of Feature-Based Transfer Learning Pipeline

Breast cancer is the primary cause of death amongst women around the globe. The detection of such cancer at an early stage could allow patients to receive medical attention, which could increase their survival chances. This study attempts to investigate the effect of hyperparameter optimisation towards the classification efficacy of breast cancer. A total of 1080 histopathological images of benign and malignant breast tumours were split into the 70:15:15 ratio for training, testing and validation. The images are evaluated by two feature-based transfer learning pipelines, i.e., MobileNet-VanillaSVM and MobileNet-OptimisedSVM, respectively. It was demonstrated through this study that optimising the hyperparameters of the SVM model does increase the ability of the pipeline to discern the classes of the tumour.