MODEL HIDDEN MARKOV PADA PREDIKSI HARGA BERAS DAN PERPINDAHAN KONSUMEN BERAS DI KOTA SOLOK PROVINSI SUMATERA BARAT

Abstract

ABSTRACT Hidden Markov models is the development of a Markov chain. Markov chain is a stochastic process that has predictive properties. The predictive nature can be understood as an event in the future is highly dependent on the behavior of the present. Parameters of the model is the transition probability matrix, the initial probability matrix, expectation, and variance of the observation process. The model is applied to predict the price of non-subsidized Solok rice and predict the movement of Solok rice consumers. It is assumed that the causes of events and changes in the price of rice Solok rice consumers displacement is not observed directly and form a Markov chain. So that this situation can be modeled with hidden Markov models. Predicted price of rice SMF using 36 rice price data from 2010, 2011, 2012 per month, while for the data transfer predictions rice consumers using questionnaires with a total of 420 respondents spread across two villages in Solok city. Hidden Markov models and algorithms with discrete time and continuous observations are used to predict the price of rice SMF with implementing it using Delphi software to get excellent results. MAPE values are quite small ranging from 2.17% - 3.74 % ( less than 20 % ) so that the data generated predictions approaching the actual data or actual data. Hidden Markov models and the Viterbi algorithm has been modified and implemented using the Delphi software hidden state sequence that produces Solok rice consumer row most likely based on models and observations of five categories of state, so in get results Caredek rice, Anak Daro rice, Sokan rice, Randah Kuniang rice, and Misc rice. Key Words: Hidden Markov Models, Markov Chain, MAPE, The Viterbi Algorithm