LAUSR

The spread of dengue hemorrhagic fever (DHF) globally with a frequency level that tends to be high in the past 50 years raises a systematic idea of prevention. One of the efforts to prevent DHF is the need for early identification of areas that are potentially epidemic. Early identification is carried out by getting an overview of the incident one step ahead by data forecasting. The focus of the study was the development of area stratification algorithms as an early identification of DHF outbreak areas by using data forecasting methods with surveillance data variables. Surveillance data which became the references for system modeling were DHF case data, rainfall, humidity, air temperature, wind speed and Larva-free Number (ABJ) for the span of 2010-2016 in 17 districts in Sleman Regency, Special Region of Yogyakarta. There were four steps during the study, i.e., 1) Forecasting of DHF case for the period of 12 months, 2) Forecasting of Larva-free Number (ABJ), 3) Determination of DHF case pattern for the last three years and the average of ABJ in the third year and 4) Area classification into stratification class. A method used for data forecasting of DHF case was seasonal autoregressive moving average (SARIMA), and the determination of area class pattern was conducted by using a neural network, meanwhile to obtain area stratification class used rule-based approach referring to guidelines controlling DHF outbreaks of the Ministry of Health of the Republic of Indonesia. Early identification was carried out by dividing into 4 area classes. Area class target included endemic (K1), sporadic (K2), a potential (K3) and free (K4). The testing of accuracy forecasting used relative mean absolute error (RMAE) for 12 months period. The results of the forecasting accuracy test on 17 districts in Sleman Regency showed RMAE average of 1.46 was considered low for it was still below 10%. Furthermore, the results of the early identification of area stratification classes in 2014 and 2015 from 17 districts showed that 3 of the four regions were endemic areas while in 2016 almost all districts were endemic areas and only one area was classified as sporadic.