Ensemble Transfer Learning Coupled Osprey Optimization Algorithm for Feature Selection from Skin Cancer Image Datasets

Among different cancer diseases, skin cancer is one of the most common leading diseases that increases the mortality rate due to lack of adequate and timely treatment. The enhancement of technology helps dermatologists to use recent deep learning methods to identify diseases in the early stage to save lives of human beings. Different existing procedures have been performed with higher precision than naked eye inspections performed by dermatologists. Even if deep learning models have already been adopted for an easy and accurate diagnosis of the disease, they face different challenges, such as skin image segmentation, noise involvement while capturing the image, etc. This study presents a ensemble transfer learning coupled with an Osprey optimization algorithm to select significant features. The ensemble transfer learning concept is used for primary feature extraction to achieve better results. Later, the Osprey optimization algorithm(OOA) is implemented to select the optimal features. We access and compared the proposed methodology using a large-scale dataset of skin cancer. Support Vector Machine is used to figure out the performance of the ensemble TL - OOA. The experimental result shows the excellence of performance metrics such as accuracy, precision, specificity, and sensitivity. This ensemble TL - OOA has achieved a diagnostic accuracy of approximately 98.61%, precision of 96. 00%, sensitivity of 100%, and specificity of 97.62%.