Cancer is a fatal condition brought on by the unchecked expansion of bodily cells. Every year, a considerable number of individuals die from cancer, which has been dubbed the most important public health issue. Any part of the human anatomy, which contains trillions of cells is susceptible to the development of cancer. Skin cancer, which starts in the skin’s outermost layer, is one of the most common kinds of cancer. Previously, a range of imaging modalities and protein sequences were utilized in conjunction with machine learning techniques to identify skin cancer. Machine learning techniques have the drawback of requiring human-engineered qualities, which is a laborious and time-consuming operation. By enabling autonomous feature extraction, deep learning partially solved this issue. Several methods have been employed in this study to detect skin cancer using a public dataset. Cancer detection is a sensitive process that might go wrong if it is not carried out accurately and immediately. The capacity of every machine learning model to detect cancer has limitations. Individual learners’ collective decisions should be more accurate than their individual decisions. The ensemble learning approach uses a range of learners to yield better outcomes. The ensemble learning approach uses a range of learners to yield better outcomes. Therefore, pooling student decisions on delicate topics like cancer diagnosis can increase prediction accuracy. In this research, we employ the Mobile Net, Xception, Efficient, Decision Tree, and Voting Classifier (DT + Gradient Boosting) models for recognition and categorization purposes. This work’s experimental results provide strong support for its application in the diagnosis of various illnesses.