Decision support system for medical diagnosis using a kernel-based approach

This work focuses on the issue of diseases diagnosis based on data classification approaches. We consider mainly the diagnosis of heart diseases, diabetes, hepatitis and fetal risks. To do so, we employ a modified version of the SVDD algorithm, endowed with efficient tools to manage the multi-classification problems. Some other conventional algorithms such as SVM and RBF are, likewise, used to take full advantages of all. The aim is to generate, from a small number of patterns, a classification model on a wider number of unknown patterns. This model can be exploited afterward to draw a useful Medical Decision-Support System (MDSS). The effectiveness of the developed approach is assessed and proved by measuring various performance criteria as Recall rate, Precision and F-measure on real datasets.