Skin Cancer Classification Using a Hybrid Pre-trained CNN with Random Forest Classifier

Kostidjan, Okky Darmawan
Purwanto, Yudhi
Yuniarti, Anny

Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi)
Vol 8 No 4 (2024): August 2024

Skin cancer, a leading cause of global mortality with 10 million deaths annually, is projected to increase rapidly if not diagnosed early. Traditional diagnosis relies on visual evaluation and histopathology, which are subjective and time-consuming. Recent advances in Convolutional Neural Networks (CNN) enable automated, accurate image analysis for early identification. This study explores pre-trained CNN models, including DenseNet-201, InceptionV3, MobileNet, ResNet50, and VGG16, by modifying them to better identify skin lesions as malignant or benign. The proposed models outperformed the state-of-the-art CNN models evaluated on publicity with traditional test data. The proposed models achieved 94.20% accuracy, which is higher than that of traditional CNN models.