FORECASTING THE PRICE OF CURLY RED CHILI PEPPERS IN EAST JAVA PROVINCE USING ARIMA MODEL WITH ITERATIVE OUTLIER DETECTION PROCEDURE

Abstract

Curly red chili is one of the vegetables with high economic value because it plays a role in supporting the food industry and meeting domestic needs. Fluctuations in the price of curly red chili peppers can change at any time, requiring forecasting to prevent losses for economic actors. This research aims to get the best model for forecasting and determine the accuracy of forecasting the price of curly red chili. The Autoregressive Integrated Moving Average (ARIMA) model is one method that can be used for forecasting with limitations requiring data that must be stationary. Outliers in the ARIMA model affect the autocorrelation structure of a time series so that the estimated values of the Autocorrelation Function (ACF) and Partial Autocorrelation Function (PACF) become biased so that forecasting with the ARIMA model is less accurate and requires handling outliers in the form of outlier detection, one of which is an iterative procedure. From this study, it was found that the ARIMA(0,2,3) model with outlier detection was the best model for forecasting. Forecasting tends to show a downward trend with an accuracy level of MAPE value of 4.612, which means that the model is very good for forecasting.