An intelligent system for diabetes prediction

With the emerging increase of diabetes, that recently affects around 346 million people, of which more than one-third go undetected in early stage, a strong need for supporting the medical decision-making process is generated. A number of researches have focused either in using one of the algorithms or in the comparisons of the performances of algorithms on a given, usually predefined and static datasets that are accessible through the Internet. This paper focuses on the joint implementation of the support vector machine (SVM) and Naïve Bayes statistical modeling, in the dataset acquired from the medical examinations of 402 patients, in order to improve the computer-supported diagnosis reliability. The dataset contains some attributes that have not been previously used in computer-based evaluations. The results show that the joint implementation of two algorithms improves significantly the overall reliability of the system outcome, which is crucial in the computer-supported diabetes diagnostic process.