Comparative Study for Predicting Melanoma Skin Cancer Using Linear Discriminant Analysis (LDA) and Classification Algorithms

Early detection and prediction of melanoma can help control large-scale outbreaks and reduce the transmission of epidemics in rapid response to serious public health events. Melanoma is one of the most dangerous types of skin cancer in the world which is responsible for the greatest number of skin cancer-related deaths. As predicted by the World Health Organization (WHO), the number of cases of diagnosed melanoma skin cancer will rise by 18% from 287,723 to 340,271 in 2025, and the number of deaths caused by melanoma will increase by 20%. WHO also predicts there will be a 20% increase in death from 60,712 in 2018 to 72,886 in 2025 and will reach 105,904, a 74% increase by 2040. However, proffering efficient methods for the early prediction can prevent further spread and melanoma-related deaths, the use of classification algorithms is one of such efficient methods for melanoma prediction. Various Machine Learning algorithms have been used in prediction of diseases which include Support Vector Machines (SVM), K-Nearest Neighbour (KNN), Naïve Bayesian, Convolutional Neural Networks (CNN) and other classification algorithms. Therefore, this study used Linear Discriminant Analysis (LDA) for feature extraction, SVM and KNN as classification algorithms for the prediction of melanoma skin cancer. The result obtained shows that both algorithm perform excellently well when optimized with LDA with an accuracy of 89% and 86% respectively. The study proposed that other algorithms/optimization techniques can be introduced to improve the robustness of the prediction.