The Improvement Method of Skin Cancer Detection by Machine Learning

Abstract

This paper proposed the method for improving skin cancer detection by finding the edges of skin regions with snake algorithm with several machine learning methods for analyze skin cancer disease by design dataset and used a machine learning model using the nearest neighbor (KNN) method, artificial neural networks (ANN), Adaptive Boosting (AdaBoost), Stochastic Gradient Descent (SGD), and Logistic Regression. By this method, the mass binding technique was applied from the value of the assigned weight from the learning data and obtained the score, matrix assessment model, method of the snake algorithm and the set of parameters to find the edges of the skin cancer images based on the basic geometric shapes to solves the problem found that the standard accuracy, recall, F1 score, and area under the curve used to generate weight vectors to find learning guidelines. Learning groups and test results based on a set of skin image data were used for testing of 1,372 images of normal skin, 1,432 images of skin cancer, 254 images of dermatitis skin were used as the learning data set and another group of test image is used as test data set of 600 images are used for increase detection efficiency. From this research, the developed program can analyze the test images with a high speed and yielded the accuracy of the data in the learning base and outside the learning base at 99.74% and 83.3% respectively.