Design an Optimization Based Ensemble CNN Technique to Classify the Various Stages of Skin Cancer

The skin is one of the few places in the body where cancer can begin. The size of the main tumor and the extent of the cancer's dissemination are both described by the cancer's stage. The classification of the various stages is more important for the treatment of skin cancer because it is impossible to take treatment measures without knowing the stage. The current study has proposed a cat swarm optimization with an ensemble CNN model (CSO-ECNN) for the exact prediction of the skin cancer stages. This method updates the cat swarm fitness in the classification layer to improve the prediction level of skin cancer. The six-phased proposed technique is shown in Figure 1. The selection of a dataset for the experiment is the first step in the suggested method. The studies in this study make use of the HAM 10,000 dataset. Before model training, several steps in the preprocessing stage must be finished. The relevance of the features is then assessed using the features extraction approach. Segmentation is employed using the GrabCut algorithm. Following that, prediction and classification are carried out using an ensemble deep learning approach. The developed model used cat swarm fitness in the classification phase for accurate prediction and classification of skin cancer. KDNN and DNN are the two DL classifiers used to determine the stages of skin cancer. Additionally, the proposed technique gained outcomes are validated with other conventional models in term of accuracy, specificity, recall, and precision.