Classification of Wisconsin Breast Cancer Data with Extreme Learning Machine and Osprey Optimization Algorithm

The illness reported for higher mortality in humans is cancer. On top of this, breast cancer is the most deadliest one among global women. Several clinicians and researchers are working towards earlier detection and its appropriate treatment. Thus, the requirement of robust Computer-Aided Diagnosis (CAD) frameworks is always in demand globally for tireless and effective diagnosis of such cancer. The paper intends to propose a CAD framework for the classification of breast tumors promptly. The research begins with the employment of initial data taken from the Wisconsin Breast Cancer database. The data preprocessing and its associated visualization are then performed. Afterward, the predominant features are selected for improving the efficacy of the framework. For this, a recently evolved Osprey Optimization Algorithm (OOA) is adopted in this work. Here, the optimization is based on the effective tactic of ospreys while hunting fishes from the oceans. Next to feature selection, the significant feature vectors are classified using Extreme Learning Machine (ELM). In addition to this, the work utilized K-Nearest Neighbour (KNN) and Mixture Kernel SVM (MiSVM) models for performance analysis. Finally, OOA features with ELM architecture provides superior accuracy of 97.66% for breast cancer classification.