A COMPARATIVE ANALYSIS OF SUPERVISED MACHINE LEARNING ALGORITHM FOR AUTOMATIC SKIN DISEASE IDENTIFICATION

Abstract

Dermatologist disease comprise a broad spectrum of conditions that impair the integrity and physiological function of skin. Diagnosing skin conditions can be both time consuming and subjective, when conducted manually by medical professionals. This highlights the need for automatic prediction of skin disorder. In our prior research, we focused on classifying skin lesion using KNN model, conducted using the HAM10000 dataset. The KNN model demonstrated accuracy rate of 97%, indicating an efficient approach to support preliminary recognition and classification of skin lesion. Further, extending the earlier work, in this study we conduct a comparative evaluation of multiple supervised learning techniques for skin disorder identification tested on HAM10000 and ISIC datasets. Our analysis focuses on metrics to gauge the effectiveness of each method in classifying skin lesion. Our findings reveal that SVM, achieve better results than conventional machine learning approaches in terms of accuracy and robustness.