Applying Statistical Models To International Crude Oil Price Forecasting

This thesis works on the area of crude oil price forecasting in the frame of statistic models from the point of view of time series analysis. It mainly focuses on these subjects: First, we introduce the factors which influence the price changes of the crude oil and the main methods used to make analysis and forecasting of crude oil price and propose the main works in this thesis.On the second part, we utilize the technical analysis which is prevailing in the oil future trading to simulate the trading process and conclude some simple results. Condidering the technical rules lack of solid mathematic theoretic foundations, we use one of the most fundamental and important time series analysis model ARIMA to process the crude oil price time series. Moreover, we combine the result of ARIMA and the powerful non-linear fitting ability of artificial neural network (ANN) to improve the prediction accuracy where we find the suitable constructure of ANN through the brute-force method in a given experiential scale. The result shows that this combined prediction method do a more accurate forecasting in the meaning of RSME than they do separately. At last, we take into account the possible regime switching mechanism of the crude oil price dynamics and utilize the state-arbitrary and two-order auto-regression markov regime switching model which is implemented using the C code to fit the series and forecast.The result comes out to be more accurate than ARIMA but not as good as the combined prediction method . Finally, we conclude our work by proposing some further work which could be carried out in the near future.