Analysis of Skin Lesion Classification using Computational Models

A common and sometimes lethal condition skin cancer requires early identification and precise categorization in order to be effectively treated. This study investigates the use of machine learning techniques, that includes Random Forest, K- Nearest Neighbors (KNN) and Support Vector Machine (SVM) in the classification of different types of skin cancer. The HAM10000 dataset used, which represents the major seven types of skin diseases, is initially processed and balanced with the help of Synthetic Minority Over-Sampling Technique (SMOTE) and SMOTE-ENN approaches. After data sampling, the findings show an important rise in accuracy, with Random Forest and SVM polynomial kernel utilizing the SMOTE-ENN approach, achieving the highest accuracies of 90.65% and 91.43% respectively. Evaluation criteria like as F1-score, precision, and recall provides additional evidence for the effectiveness of these models. The work emphasizes how crucial it is to correct dataset imbalances related to skin cancer and how conventional machine learning methods can be used to reliably and quickly classify skin cancer.