Abstract:
Skin diseases pose a worthy of attention in health concern thus millions of people across various demographics. Timely and error-free diagnosis is critical for effective medical care and control but unfortunately there is limited access to dermatologists in different regions. In response to this concern, the deep learning models have spring up as potential tools for automatically classifying skin diseases. Existing research focused on using single-sourced datasets thus restricting their generalization abilities. This study bestows a panoramic approach for detecting skin diseases by merging two large-scale datasets HAM10000 and Dermnet – leveraging big data techniques for diagnostic accuracy improvement. Integrating big data in dermatology for automated diagnosis offers improved diagnostic accuracy, identifying disease patterns and trends for early detection of diseases. The study employs hybrid deep learning models utilizing VGG19 and DenseNet121 to enhance feature extraction and model performance.The system is evaluated using state-of-the-art performance metrics. The results showed that the hybrid model outperformed individual models, achieving an impressive 99% pre- cision, recall, and F1-score, with 0.99 G-Mean and 0.98 AUC. This research con- tributes to the growing field of AI-assisted dermatology, highlighting the role of big data in automation of machine learning models for real-world medical applications
