Cancer is a fatal illness with several underlying causes in metabolism and genetics. The third most common cancer-related cause of death worldwide and the sixth most common disease overall is skin cancer. The skin plays a crucial role in the human body. The skin controls cholesterol homeostasis and stores fat-soluble vitamins. Skin cancer may be treated more effectively if detected and classified early on. Brain, breast, skin, and other tumors may be examined using imaging technology. Convolutional neural networks (CNNs) have improved medical imaging for identifying and classifying cancer compared to earlier methods. To help radiologists notice anomalies more quickly, this research presents a classification model for recognizing skin tumors utilizing a convolutional neural network (CNN), an enhanced optimization technique, and transfer learning. The MPA algorithm for detecting marine predators) was the basis for our optimization strategy. Using an improved version of the marine predator’s algorithm (MPA), developed as part of this project, I was able to identify the ideal values for the CNN’s metaparameters. The proposed method combines a version of the ResNet50 CNN model that has already been trained with the MPA algorithm to create the MPA-ResNet50 architecture. This MPA offers superior precision than competing optimization techniques. MPA modifies the primary discovery and exploitation procedures to zero down on the most relevant features for reliable categorization.