부산대학교 도서관

Dengue Hemorrhagic Fever (DHF) is an infectious disease caused by dengue virus and transmitted by Ae mosquitoes. Aegypti. The mortality rate due to dengue disease is relatively high due to patient delay in realizing the early symptoms of DHF. Early detection of Dengue Fever is an effort made to determine the possibility of Dengue Hemorrhagic fever in a person's body. Problems that arise from early detection is the data used is worth Yes and No so it can raise the problem outlier in the data because only see from the clinical symptoms of dengue disease as a result of the condition of one hot encoding. This research aims to find the best Backpropagation Algorithm to solve the problem by performing the analysis of early detection of DHF with the addition of optimization on Multi Layer Perceptron (MLP) through five kinds of Back-propagation training algorithms, Gradient Descent (GD), BFGS Quasi-Newton (BQN), Conjugate Gradient Descent — Powel (CGD), Resilient Backpropagation (RB), and Levenberg Marquardt (LM). Variables used in this study are the initial symptoms of DHF patients as many as 10 variables. All research data is taken based on medical record data in RSUP Dr Karyadi Semarang a total of 76 data. The results showed that in the detection of dengue disease the best performance is the Levenberg Marquardt algorithm.