Page 220 - Contributed Paper Session (CPS) - Volume 2

CPS1834 Gumgum D. et al.

Rainfall data modeling by SARIMA based on
Hijriya calendar
1
2
1,2
2
Gumgum Darmawan , Dedi Rosadi , Budi N Ruhyana , Danardono ,
3
Hermansah
1 Universitas Padjadjaran
2 Universitas Gadjahmada
3 Universitas Riau Kepulauan

Abstract
ARIMA modelling procedure is well known developed by G.E.P.Box and
G.M.Jenkins. Using this procedure, We are able to obtain ARIMA model that
can represent past and future patterns of time series. These patterns are
random, seasonal, trend, cycle, or combination patterns. In standard model,
we use Gregorian Calendar to analysis the data in time domain. Here, in this
paper we consider the analysis of the data using what we call Hijri calendar In
particular ,we apply the method to rainfall data. Based on the empirical
analysis we found that SARIMA based on Hijri calendar has lower MAPE than
SARIMA analysis in Gregorian calendar especially in forecast in three future
observations (short-term forecast). However, when we extend the analysis to
long-term forecast, the forecast based on Gregorian calendar , SARIMA model
has better performance.

Keywords
Hijriya Calendar; Sarima Modeling; Rainfall; Opensource Software

1. Introduction
In the time series analysis known for various methods in forecasting,
several classical forecasting methods such as Moving Average, Exponential
Smoothing, ARIMA, and Classical Decomposition are often used in forecasting
the future. Different methods have advantages and disadvantages of each
which will be adjusted to the characteristics of the data to be predicted.
Moving averages use the average value in the previous period data as a
forecast for future data.
The time series data models above are widely used to forecast rainfall such
as [1], [2] and [3] using singular spectrum analysis. Whereas [3] and [4] use
Ensemble methods and [6] use ARMA method to predict rainfall. [7] build a
software to predict rainfall [8] forecast rainfall with Artificial Neural Network.
[9] and [10] using probabilistic forecasting and [11], [12], [13] and [14] using
other model models such as machine learning to forecast rainfall.
From the research above, rainfall data uses time series data with the
Gregorian calendar. While according to Tsubatsa Kohyama (2016), variations

209 | I S I W S C 2 0 1 9