Implementation of Feature Selection Method to Diagnosis of Skin Disease By using Classification Techniques

Abstract

Skin cancer detection develops into a thought-provoking concern in identifying the correct location in the skin. Image processing occupies a significant part in diagnosis. For initial diagnosis of skin cancer, this image analysis feature is applied. Skin cancer appears to be the most mutual disease in recent years. Initial diagnosis decreases the death rate. This is accomplished by combining feature extraction and segmentation methods. The best appropriate feature subclasses of skin lesion like color, texture are removed and accepted for classification. A method is developed to choose features from huge data sets which contain more inappropriate or repeated features. In this paper, a three-step procedure aimed at selecting features is proposed. The system in the Genetic Algorithm is a classical machine learning technique originates its routine from a representation of evolution. Improving the quality of the image for perception is the main task of GA in order to improve the Feature selection, when applied to classification for skin cancer images, signifies the importance of feature selection while decreasing the number of features required to design classifier. For classification to improve, the images of the affected skin are removed from the normal skin. An automatic skin growth classification scheme is proposed with the associations of skin lesions through the ANN training network.