CNN-Based for Skin Cancer Classification with Dull Razor Filtering and SMOTE

KZ, Widhia Oktoeberza ((Scopus ID: 56669681900, Universitas Bengkulu))
Prasetio, Wahyu Dwi (Universitas Bengkulu)
Ramadhan, Adam Idham (Universitas Bengkulu)
Putra, Adde Nanda C. (Universitas Bengkulu)
Eliora, Firsti (Universitas Bengkulu)
Mainil, Afdhal Kurniawan (Universitas Bengkulu)
Susanto, Agus (Universitas Bengkulu)

Telematika
Vol 18, No 1: February (2025)

Skin cancer is usually diagnosed by dermatologists through biopsy, which can be a time-consuming process due to limited resources. Early detection of skin cancer can increase the survival rate to over 99%, but if it's detected late, the rate drops to around 14%. This finding highlights the need for a rapid and accurate computing system for early cancer detection, which can prevent severe consequences. The purpose of this study is to classify skin cancer images into benign and malignant classes based on their nature. To facilitate the classification of CNN-based skin cancer, this research employs dull razor filtering. Additionally, the SMOTE method handles the unbalanced dataset. The classification results indicate that the proposed approach has an accuracy of 88.54%, a precision of 88%, and a sensitivity of 88%. These findings suggest that CNN-based methods can aid dermatologists in the diagnosis of skin cancer.