e-Doctor: A Web based Support Vector Machine for Automatic Medical Diagnosis

This paper proposes e-doctor; a web-based application that makes automatic diagnoses about health problems. The whole procedure is based on Support Vector Machines (SVMs), which are supervised learning models that analyze data and proceed to decisions, based on their knowledge. System administrators define specific characteristics for each health problem that can be diagnosed, and educate the SVM by entering sample files of statistical data. After that, medical staff can enter exam information about patients, and e-doctor makes an automatic diagnosis/prediction by means of answering if the patient has (or may have in the future) a specific health problem. The application can be used in cases where statistical information plays a vital role on deciding about a patient's condition. A prototype was developed and the system trained and tested for the case of heart symptoms. The results were satisfactory.