Skin Cancer Classification Using Random Forest Algorithm

Abstract

Skin cancer is an excessive lump of skin tissue that affects the skin, has an irregular structure with cell differentiation at various levels in chromatin, nucleus and cytoplasm, is expansive, infiltrative to damage the surrounding tissue, and metastasizes through blood vessels and lymph vessels. Diagnosis of skin cancer by biopsy process is considered less effective because it costs a lot and can injure human skin as a sample. For that, we need a system for classification of skin cancer types that are effective and accurate. The application of machine learning has been widely used in the health sector. One of the machine learning methods is Random Forest. In this study, the histogram color feature extraction will be carried out, the hue moment shape extraction, and the haralick texture extraction. Furthermore, the image will be classified using the Random Forest algorithm. The best accuracy value obtained from the histogram feature extraction process and classification with Random Forest is 0.850822. The novelty of this research is the use of more diverse feature extraction and better accuracy results than previous studies. Future research is expected to use deep learning algorithms with CNN (Convolutional Neural Network) architecture to get better accuracy results and add application designs for the application of models that have been formed in the study so that they can be directly applied by the medical team.