PENERAPAN MODEL ARIMA-NEURAL NETWORK HYBRID UNTUKPERAMALAN TIME SERIES
Ensiwi Munarsih, H. Subanar · 2011
ARIMA Model and Neural Network are methods that was usually used for forcasting time series data. Both of them have the differences, where ARIMA model better used to predict of linear time series data, while Neural Network better used to predict of nonlinear time series data. But in real-world time series problems not only linear or nonlinear, usually both of them (linear and nonlinear). So that created hybrid model, called ARIMA-Neural Network Hybrid model. This model is applied to data that contain seasonal time series. Based on the test using the Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) shows that the error from ARIMA-Neural Network Hybrid is smaller than a single ARIMA. This indicates that the ARIMA model-Hybrid Neural Network is better used for forecasting than a single ARIMA model.