FORECASTING CRUDE OIL PRICE TRENDS USING HIDDEN MARKOV MODELS (HMM)

• Main Author: SYAKINA TIKITA FARNE, POCUT
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/19139
• Institution: Institut Teknologi Bandung

Crude oil is one of the crucial commodities in many countries because it is still used as the main fuel nowadays. The fluctuations of crude oil price causing necessity of proper investment decisions. There are two aspects that need to be considered in this problem, the first is observed aspect which is closing price of crude oil and the second is the trend of crude oil price. The problem of forecasting oil price trends can be modeled with Hidden Markov Models (HMM) because this model contains two aspects stated previously. This final project discusses the application of HMM using daily closing price data of crude oil as observation sequence. The data are coded into some symbols and modeled to be initial parameters of HMM on using the increase and decrease frequency of crude oil price because the trends of crude oil price are assumed to be up or down Parameter re-estimation is done using Baum-Welch Algorithm to get the new parameters suitable with given observation sequence. In forecasting crude oil price process, Viterbi Algorithm is used to get the hidden states sequence (crude oil price trends) suitable with the result of parameter re-estimation process. Hidden states sequence obtained from Viterbi Algorithm are used to see the behavior (trends) of crude oil price for several days ahead.