Novel machine learning method in skin cancer diagnosis based on patient's metadata

Abstract

The skin's fundamental function in human body as a whole-body covering is crucial. Only if it is discovered while it is in its early stages can skin cancer be cured. Skin function plays a big part in the body's overall system and will be significantly impacted by even the slightest modification. The goal of this work was to develop an effective Machine Learning (ML) based technique for identification of skin cancer using patient information. To diagnose skin cancer with lesions image, this research introduces a novel Augmented May Fly optimized with K-Nearest Neighbors (AMFO-KNN) technique. Here, the AMFO approach is used to improve the classification efficiency of KNN. Utilizing the PAD-UFES-20 and Fitzpatrick17k datasets, the efficiency of suggested method is examined. The noisy data are removed from the raw data samples using Adaptive Median Filter (AMF). The properties are taken out of segmented data using Kernel Principal Component Analysis (KPCA). The performance metrics of research show that recommended methodology performs better than traditional approaches in terms of accuracy, precision, f1-score, and recall measures. The encouraging results demonstrate the effectiveness of suggested strategy and show that including the patient's information with lesions image may improve the performance of skin cancer diagnosis.