(IJACSA) International Journal of Advanced Computer Science and Applications, Vol. 8, No. 11, 2017 143 | Page www.ijacsa.thesai.org A Model for Forecasting the Number of Cases and Distribution Pattern of Dengue Hemorrhagic Fever in Indonesia Deni Mahdiana Faculty of Mathematics and Natural Sciences Universitas Gadjah Mada Yogyakarta, Indonesia and Faculty of Information Technology Universitas Budi Luhur Jakarta, Indonesia Ahmad Ashari Faculty of Mathematics and Natural Sciences Universitas Gadjah Mada Yogyakarta, Indonesia Edi Winarko Faculty of Mathematics and Natural Sciences Universitas Gadjah Mada Yogyakarta, Indonesia Hari Kusnanto Faculty of Medicine Universitas Gadjah Mada Yogyakarta, Indonesia Abstract—Dengue Hemorrhagic Fever (DHF) ourbreaks is one of the lethal health problems in Indonesia. Aedes aegypti type of insect prolefiration as the main vector of DHF has affected climate factors, such as temperature, humidity, rainfall, and irradiation time. Therefore, to project the number of DHF cases is a very important assignment for the Ministry of Health to initiate contingencies planning as a prevention step in confronting the increasing number of DHF cases in nearby future. This study aims in developing a forecasting model in anticipating the number of cases and distribution pattern of DHF with multivariate time series using Vector Autoregressive Spatial Autocorrelation (VARSA). VARSA model uses multivariate time series, such as a number of DHF case, minimum temperature, maximum temperature, rainfall, average humidity, irradiation time and population density. This modeling is done in two steps: Vector Autoregressive modeling to predict the number of DHF cases and Local Indicators of Spatial Association (LISA) method to visualize distribution pattern of DHF based on the spatial connectivity of the number of DHF cases among the neighboring districts. This study covers 17 districts in Sleman Yogyakarta, resulting in low errors with Root Means Square Error (RMSE) of 2.10 and Mean Absolute Error (MAE ) of 1.51. This model produces smaller errors than using univariate time series methods, such as Linear regression and Autoregressive Integrated Moving Average (ARIMA). Keywords—Dengue Hemorrhagic Fever (DHF); Vector Autoregressive Spatial Autocorrelation (VARSA); forecasting; multivariate time series; Local Indicators of Spatial Association (LISA) I. INTRODUCTION Dengue Hemorrhagic Fever (DHF) is an acute and endemic disease which periodically causes outbreaks and even epidemics. DHF is widely found in tropical and sub-tropical areas [1]. Data from around the world shows Asia ranks first in the number of DHF patients each year. Meanwhile, from 1968 to 2009, the World Health Organization (WHO) listed Indonesia as the country with the highest dengue fever case in Southeast Asia [2]. DHF is a disease caused by dengue virus that is transmitted from person to person through the bite of aedes aegypti mosquitoes [3]. This disease is caused by the dengue virus of the genus flavivirus, flaviviridae family. Dengue transmitting mosquitoes are present in almost all corners of Indonesia, except in places with an altitude of more than 1000 meters above sea level. Several factors that influence the occurrence of DHF include low immune status and a population density of infectious mosquitoes due to mosquito breeding places that usually occur in the rainy season [4]. The number of DHF cases in 34 provinces in Indonesia based on data from the Ministry of Health in 2012 recorded as many as 90,425 cases and 816 people died. In 2013 recorded as many as 112,511 cases and 871 people died, In the year 2014 recorded as many as 71,668 case, and 641 people died , In the year 2015 recorded as many as 129,650 case, and 1,071 people died [5]. Preventive program namely the eradication of mosquito nest has been widely conducted nationally and regionally. This preventive activity is carried out by draining, closing, burying water reservoir and applying larvicide, keeping larvae fish and using a mosquito net, examination and eradication of larvae periodically no more than three months, and fumigation. Nevertheless, those preventive actions have not been able to reduce the number of DHF patients nationally. The low ability in preventing of dengue fever is due to some factors. The first is the unpredictable time, place and number of events. The second is the unavailability of index and vulnerability maps of the region based on the time of the incident. The third is the unavailability of reliable model for forecasting DHF cases. Forecasting the number of DHF cases is very important for public health service to anticipate