Analysis On Forecasting Crude Oil Price Based On Bayesian And Maximum Likelihood Method

Oil is an important strategic material to ensure the smooth operation of various sectors of the national economy.In the past ten years,frequent fluctuations in international crude oil prices have increasingly become an unstable factor restricting the stable operation of economies in various countries.China is the world's largest importer of crude oil,and fluctuations in crude oil prices have disrupted the stable operation of the Chinese economy.In the unpredictable world crude oil market,if you can make a correct prediction of the trend of oil prices,this can minimize the adverse economic impact caused by large fluctuations in oil prices and maximize our own benefits.Therefore,paying close attention to the international crude oil market,exploring the potential causes of international crude oil price changes,and making reasonable predictions about the trend of crude oil prices are of great significance to countries,enterprises and individuals.In view of the various attributes of crude oil,such as commodities,finance,and politics,this article summarizes the rich research results of previous people in the prediction of crude oil prices,in an attempt to achieve a more accurate prediction of crude oil prices.As a research object,a Heston crude oil price prediction model based on Bayesian method and maximum likelihood estimation method was proposed,and Bayesian method and classical estimation method were used as two comparative estimation methods to estimate the model parameters.In this paper,the crude oil price is dynamically predicted by the Heston model based on the Bayesian method maximum likelihood estimation method.In the empirical part,this paper first estimates the model parameters based on previous studies,and obtains the analytical expressions of the five parameters of the Heston model under the Bayesian method and the maximum likelihood method.In the execution of the forecasting process,we use the out-ofsample forecasting method of the rolling period.We try to predict the price of crude oil in the next 100 days.Therefore,we fixed the forecast interval for 100 days and set six data length estimation intervals to test the estimates.The effect of interval data length on the prediction effect of crude oil.Five loss functions of MSE,MAE,RMSPE,RMSE,and MAPE are introduced into the evaluation system of the prediction effect of crude oil.After comparing the prediction results of the two methods,we find that: both estimation methods can effectively estimate the parameters of the Heston model;under a specific loss function,when the estimated sample size is small,the Bayesian method is more effective than the classical estimation method It has advantages,and the classical method is slightly better than the Bayesian method when the estimated sample size increases.At the same time,we also use entropy method to measure the predictability of the original oil price.The results show that the entropy value predicted under the maximum likelihood estimation method can better represent the predictability of the future price for a period of time.