www.iaset.us editor@iaset.us APPLICATION SARIMA MODELS ON TIME SERIES TO FORECAST THE NUMBER OF DEATH IN HOSPITAL Hanaa Elgohari 1 , Mohammed AbdulMajeed 2 & Ahmed Elrefaey 3 1 Department of Applied Statistics, Faculty of Commerce Mansoura University, Egypt 2 Faculty of Administration, Faculty of and Economics Salahaddin University, Iraq 3 Professor, Department of pediatric, Mansoura University, Egypt ABSTRACT This paper aims to predict the number of deaths at Mansoura University Children's Hospital by using SARIMA models. It is necessary to use death data to determine the health requirement for hospital and measure medical efficiency within the hospital. We take the death data in hospital from Jan. 2011 to Dec 2017. We concluded that the model SARIMA (1,1,1) (0,1,1) is the best model which gives us the lowest value for each of RMSE and BIC, approximately lowest value for MAE and the largest value for R2. KEYWORDS: Time Series, SARIMA Models, BIC, RMSE, MAE, MAPE, R2, ACF, PACF Article History Received: 19 Apr 2018 | Revised: 10 May 2018 | Accepted: 22 May 2018 INTRODUCTION The records that are gathered over time refer to Time Series analysis, because of the importance of the time order of data. One differentiating characteristic is that the applications of time series applications are very various and the records are dependent in time series. In addition, data may be gathered hourly, daily, weekly, and monthly and yearly, this depends on various applications. Moreover, notation can be used to symbolize ''T'' for a time series of length and the unit of the time scale implied in these notations such as {Xt} or {Yt} (t =1,···,T). We start to introduce a number of real data that are used to indicate the modeling and forecasting of time series. The term of seasonally refers to a regular model of changes which repeat for S time period, in which S refers to the numbers of timer periods till the pattern repeats again. Surly, seasonality causes the time series to be no stationary, a difference between a value and a value with lag and it refers to a multiple of S is called seasonal distinguishing. The term of time series defined as data series that indexed (listed or graphed) in time order. Generally, a time series refers to the word '' sequence'' that is taken at equally, successive, and spaced points in time. Therefore, it is the sequence of separated time data. In addition, to, time series are frequently plotted through line charts. Also, time series applied in signal processing, statistics, the forecasting of weather, econometrics, the finance of mathematics, transport, earthquake prediction, the forecasting of trajectory, astronomy electroencephalography, communications, and control engineering, and broadly in any field of applied science and engineering that includes temporal measurements. International Journal of Applied Mathematics Statistical Sciences (IJAMSS) ISSN(P): 2319-3972; ISSN(E): 2319-3980 Vol. 7, Issue 4, Jun - Jul 2018; 9-18 © IASET