A new adaptive support vector machine for diagnosis of diseases

Abstract

Automatic disease diagnosis systems have been used to treat diseases for many years. The data used in the construction of these systems require correct classification. Therefore, previous literature has proposed a variety of methods. This paper develops a general-purposed, fast and adaptive automatic disease diagnosis system, using the information generated by the newly-designed classifier, which makes decisions with a simple rule base comprising the rules in if-then' form. This newly-proposed methodology is based on the support vector machine (SVM), a powerful classification algorithm. In the proposed method of this study, we added a feature of adaptivity to an SVM. In order to increase the success rate and decrease the decision-making time, the bias value of the standard SVM is changed in an adaptive structure. This process introduces a new kind of SVM, adaptive SVM', seeking a diagnosis of diseases in a more successful way. During the training and test operations of this newly designed system, we used diabetes and breast cancer datasets, acquired from the medical database of California University. This newly proposed methodology has 100% correct classification rates on both diabetes and breast cancer datasets.