A Novel Methodology to Predict Dermatological Disease Based on Image Classification Principle with Deep Learning Association

This study introduces a novel hybrid model, the Efficient Deep Kernel Classifier (EDKC), for classifying dermatological diseases such as Monkeypox, Measles, Chickenpox, Herpes, and Melanoma. The EDKC model combines a lightweight EfficientNet-inspired feature extraction framework with a kernel-based classification layer, addressing challenges like limited datasets and overlapping class features. Trained on the Skin Disease Lightweight Dataset (1840 images) with preprocessing techniques, the model achieved impressive performance with an accuracy of 97.99%, surpassing nine existing models. It also excelled in precision (97.56%), recall (97.34%), and F1 score (97.45%). The results demonstrate EDKC's potential for real-time dermatological disease classification in mobile health apps and clinical systems.