Kajian numerik model hidden markov satu waktu sebelumnya untuk nilai tukar Rupiah terhadap Dolar Amerika

Abstract

Hidden Markov model is a model that consists of a pair stochastic processes, ie: the process of observation and the cause factors of the observation, where the cause factors of the observation are not observed directly and assumed to be a Markov chain. The previous time hidden Markov model is a hidden markov model that the observation process is influenced by current events and one previous time cause factors. The previous time hidden Markov model can be applied to the exchange rate of rupiah to US dollar, where the exchange rate of rupiah to US dollar is the observation and the contributing factor is a Markov chain. Parameters of the model are estimated using the maximum likelihood method and calculated using expectation maximization (EM) iterative algorithm. Using the initial value obtained by trial and error, Santoso (2008) was able to model the exchange rate of rupiah to US dollar with MAPE 14.58%. In this paper, the initial value is structuredly generated on a specific interval so that the accuracy of model increases. Retrieved MAPE is 4.13% with only one iteration.