CAViaR Secretary Bird Optimization Algorithm Enabled Transfer Learning for Melanoma Classification

Melanoma is one amongst aggressive diseases with some ordinary treatment choices. The fatality rates and their occurrence vary expansively around the world relying upon access to earlier identification and basic medical care. The significant demerits of the existing classification models are a lack of predictive significance and a small relationship with treatment results. Here, CAViaR Secretary Bird Optimization Algorithm enabled Transfer Learning (CSBOA_TL) is introduced for melanoma classification. Firstly, an Adaptive Kalman Filter (AKF) is utilized to pre-process input skin images. After pre-processing, skin lesion segmentation is conducted employing Deep Fuzzy Clustering (DFC). Thereafter, statistical features, Oriented FAST Rotated and BRIEF (ORB) and Haralick features are extracted. Lastly, melanoma classification is conducted by Convolutional Neural Network (CNN) based TL, where CNN is utilized with fetched hyperparameters from EfficientNet. The CNN-based TL is trained utilizing the CAViaR Secretary Bird Optimization Algorithm (CSBOA), which is presented by incorporating Conditional Autoregressive Value at Risk (CAViaR) with Secretary Bird Optimization Algorithm (SBOA). Additionally, CSBOA_TL achieved a maximal accuracy of 91.852%, a True Negative Rate (TNR) of 92.788% and a True Positive Rate (TPR) of 90.715%.