31 Comparing Performance of Back Propagation Networks and Support Vector Machines in Detecting Disease Outbreaks
-
Published:2008
Download citation file:
Syndromic Surveillance is used for the early detection of disease outbreaks. Back Propagation has been demonstrated to be an accurate, robust, and scalable detection technique for disease outbreaks in over the counter pharmaceutical sales. The purpose of this study is to determine whether Support Vector Machines are comparable to Back Propagation in performing Syndromic Surveillance. Back Propagation and Support Vector Machines are used to detect outbreaks based on emergency department and Telehealth data. A data simulation methodology has been used to produce sufficient quantities of realistic data to perform this study. The results demonstrated that Support Vector Machines with a polynomial kernel are more successful to Back Propagation for detecting disease outbreaks based on data from emergency department in terms of false detections; however, they are comparable in terms of the detection time. In addition, Support Vector Machines with a polynomial kernel are comparable to Back Propagation for detecting outbreaks based on data from Telehealth in terms of false detections and the detection time.