Data Mining Techniques for Prediction of Different Categories of Dermatology Diseases

摘要: INTRODUCTION In medical science, diagnosis of health conditions is a challenging task. Medical history data comprises number tests essential to diagnose particular disease and the are based on experience physician; less physician can problem incorrectly. Hence, it possible for care industry increase advantages through use mining techniques develop decision support system (DSS) which will uniformly intelligently. Therefore, an effective intelligent DSS different types diseases requirement care. Dermatology study skin that extremely complex difficult diagnose, ultimately may be leading cause cancer. The six categories these share similar clinical features erythema (Guvenir & Emeksiz, 2000; Elsayad, 2010b). Classification robust technique in mining. Even though most studies conducted field classification erythemato-squamous diseases, researchers still working find best classifier this kind dataset (Ubeyli, 2008 2009; Several authors (Guvenir, 1998; Guvenir Nanni, 2006; 2010b) have used disease. et al. (1998, 2000) were pioneers area done lots works classifier. their work 2000), they developed graphical user interface (GUI) with all visible information nearest neighbor, naive Bayesian voting intervals-5 assist involved domain. A domain expert while intern-doctors verify knowledge; model has achieved remarkably high accuracy 99.2% set collected own. Other (Bojarczuk, 2001; Ubeyli Guler, 2005; Nani, Polat Gunes, Dogdu, 2010; Barati al., 2011) also various such as tree, neuro-fuzzy, k-means clustering, SVM same purpose between 94.22% 98.3%. Elsayad (2010b) investigated ensemble using multilayer neural network, tree linear discriminant analysis (LDA) got 98.23% accuracy. Recently, Xie Wang (2011) applied vector machine novel hybrid feature selection methods 98.61% Among above authors, Emeksiz (2000) highest 99.2%. study, we presented classifications Support Vector Machine (SVM) Artificial Neural Network (ANN), diseases. dermatology taken from University California at Irvine (UCI) learning (web source http://archive.ics.uci.edu/ml/datasets.html, last accessed Jan 2012) demonstrate techniques. obtained piece research close among other models suggested literature. proposed only artificial network 98.99%. our competitive compared by DATA SET DESCRIPTION Each sample classified into categories: psoriasis, seboreic dermatitis, lichen planus, pityriasis rosea, chronic dermatitis rubra pilaris. …