Explainable Multi-Class Classification Based on Integrative Feature Selection for Breast Cancer Subtyping

Breast cancer subtype classification is a multi-class problem that can be handled using computational methods. Three main challenges need to addressed. Consider first the high dimensionality of available datasets relative extremely small number instances. Second, integration different levels data makes even more challenging. The third challenging issue ability explain predictions provided by machine learning model. Recently, several deep models have been proposed for feature extraction and classification. However, due size datasets, they were unable achieve satisfactory results, particularly in Aside from that, explaining impact features on has not addressed previous works. To cope with these problems, we propose multi-stage selection (FS) framework two schemes. Using multi-omics data, four models, namely support vector machines, random forest, extra trees, XGBoost, investigated at each level. SHAP was used how specific influenced Experimental results demonstrated ensemble early stage improved compared baseline experiments state-of-the art Furthermore, explanations regarding implications relevant are provided, which could serve as future biological investigations.