Classification of skin abnormalities in dermatology plays an important role in the detection and diagnosis of various skin diseases such as melanoma. However, accurately classifying skin cancer remains difficult due to imbalanced classes, variances within classes, lack of training data, and failure to focus on semantically significant lesion parts. To address these issues, the proposed novel Deep-Convolutional Network (DCN) with a Machine-Learning (ML)-based SVM technique has been used to categorize skin cancer. To prevent overfitting, the DCN-SVM method is used as a classifier. Furthermore, the classification model applies the expert framework for learning based on domain functions. The effectiveness of test results using a UNet-MobileNet with 10,000 training images (HAM10000) dataset in seven distinct human-machine classes. The results demonstrate the effectiveness of this method in classifying dermatological diseases, demonstrating its potential to improve dermatological care and diagnostic accuracy.