Classification of Erythemato-Squamous Diseases Using Association Rules and Fuzzy c-Means Clustering

Abstract

Automatic classification of the erythemato-squamous diseases is an important problem in dermatology. Despite very little differences, almost all erythemato-squamous diseases have similar clinical features of erythema and scaling. Thus, a hybrid classification model is proposed for automatic classification of the erythemato-squamous diseases. This hybrid model consists of two sub steps. In the first sub step, a feature selection schema is considered. In this context, the association rules (ARs) method is employed for feature selection purposes. The Apriori algorithm is considered as mining the ARs. The fuzzy c-means (FCM) clustering algorithm is located in the second step of the proposed hybrid model where the membership degrees of the data points are obtained through iterative minimization of a cost function. Several experimentations are conducted for evaluating the performance of the proposed hybrid method on detection of the erythemato-squamous diseases on MATLAB environment. We also compared our proposal with the standard FCM and k-means clustering algorithms. The performance evaluation of the proposed method was realized according to the several criterions such as classification accuracy, sensitivity and specificity values. According to the performance evaluation criterions, our proposal yielded better classification performance than the compared clustering methods. While the proposed AR+FCM obtained 75.96% classification accuracy, FCM and k-means produced 75.14 and 68.85% classification accuracies, respectively. Based on the obtained results, the proposed hybrid scheme improves the correct classification rate of erythemato-squamous diseases.