Perbandingan Model AR(1), ARMA (1,1), dan ARIMA (1,1,1) pada Prediksi Tinggi Muka Air Sungai Bengawan Solo pada Pos Pemantauan Jurug

A river flow determines the prediction of river flow is difficult, usually the value used as a benchmark is the result of monitoring the water level. In July 2016, the flood of Bengawan Solo river caused flood in East Solo area. This is because the water level at the monitoring station Jurug penetrate level 10. Therefore, the prediction of water level is needed as an early warning effort of flood. The measurement of the water level of the Bengawan Solo River at each monitoring post is done daily. The water level data is the time series data. One method of forecasting time series data is Autoregressive Integrated Moving Average (ARIMA), this model has the assumption of homoscedasticity or fixed error variance. But if the variance of the error varies then the model used is the Generalized Autoregressive Conditional Heteroscedasticity (GARCH) model. This study used 60 data from January to February 2017. The data proved stationary based on ADF value 0.0036, therefore ARIMA model can be used. Based on the corelogram pattern, ACF and PACF are cut off after the first lag, this shows that the river water level of the period can be modeled with AR (1), ARMA (1,1), and ARIMA (1,1,1). Based on the comparison of MAPE values, the three models of the lowest value are ARMA (1.1) model, that is 0.668384 which means the error rate on the prediction of ARMA model (1.1) is 66.8384%. So the thing with the MSE value of the three models, the lowest value in the ARMA model (1.1) that is 0.7729 means having a smaller model variance, able to give more consistent results than AR (1) and ARIMA (1.1, 1) that is 1.060288 and 0.996585.