Analysis Of Diesel Production Time Series Based On BADL Model

The autoregressive distributed lag model is a more general form of autoregressive model.Unlike autoregressive model,this model contains lag term of endogenous variables.Its estimation method is usually the ordinary least squares.This paper attempts to estimate the parameters with Bayesian method.With Minnesota conjugate prior distribution as prior,the estimation method of the posterior distribution is Theil mixed estimation.Based on the sample information and prior information,the estimated values of the parameters are obtained.In recent years,the growth rate of diesel production is slowing down.This is related to not only the scale of oil refining,but also the yield structure of refined products.Therefore,forecast and analysis of oil refining industry plays a very good guidance for diesel production.In this paper,we considered the factors that may affect the diesel production and get the data.According to deleting outliers,filling missing values and seasonally adjustment,the data was effectively processed.And the data passed the stationarity test,cointegration test and Granger causality test.Then we tried over different combinations of hyper parameters of Bayesian autoregressive distributed lag model,and fitted the data.We selected five models for forecasting by means of comparing the goodness of fit.Finally,we compared the prediction results of autoregressive model,the autoregressive distributed lag model based on frequency method and the Bayesian autoregressive distributed lag model,and obtained the optimal model.