Machine Learning Models for Prediction of Melanoma Skin Cancer Types

The skin, a crucial organ safeguarding muscles, bones, and other tissues beneath the environment, faces ongoing exposure to harmful elements like ultraviolet (UV) radiation, dust, and pollution. This increases the risk of disorders such as melanoma. Despite melanoma being less common than other skin cancers, it is more aggressive and contributes disproportionately to global skin cancer fatalities. The stage of spreading malignant melanocyte cells throughout the body is vital, making early detection and staging crucial. Identifying skin cancer is a resource-intensive process. This research employs computerized image analysis to enhance melanoma detection by extracting new information from skin lesion photos, reducing misdiagnosis rates. The approach involves using 13 Haralick texture characteristics, k-means clustering for computational efficiency, and the YCbCr color space for skin tone segmentation. Results indicate that the Random Forest (RF) classifier achieves the highest classification accuracy at 90%, surpassing Support Vector Machines (SVM) at 87.05% and Artificial Neural Networks (ANN) at 75%.