Peramalan Gabungan Rantai Markov dan Model Deret Waktu Pada Kasus Peramalan Kurs Nilai Mata Uang

Abstrak

This research aims to model forecasting of dollar against rupiah by combining the Markov chains and time series models. Probability transition matrix arranged based on 459 time series data of the exchange rates for Australia Dollar (AUD) from October 20, 2014 until August 31, 2016. There are ten classifications were determined based on the exchange rates from sharply lower to sharply higher. Forecast results based on summation of forecast results with the magnitude of the change based on the state prediction probability. Evaluation of the best models are based on the value of Mean Squared Error (MSE) preliminary models. Then, the best models are based on Mean Absolute Percentage Error (MAPE) and Mean Absolute Deviation (MAD) forecast result. The result, there are three models that are considered the best: MC-SMA18, MC-DES10, and MC-DES10.S. The model chosen for this case is MC-DES10.S with MAPE = 0,352% and MAD = 35,107.