可下载资源

We respectively changed the names of independent variables such as WTI Price Field Production of Crude Oil (Thousand Barrels) to x1, x2… The last emergency event needs to use dummy variables. Only two dummy variables are needed, we will use them as X49, X50, X51, three parameters and their values ​​”positive influence”, “no influence”, “negative influence” were changed to -1,0,1.

It can be seen that those with less impact have been eliminated.

Garch model predicts volatility

We use the Garch model to predict volatility,

> shapiro.test(rlogdiffdata) 
 
	Shapiro-Wilk normality test
 
data:  rlogdiffdata
W = 0.94315, p-value = 1.458e-05

It can be seen from the QQ graph and the p-value that the data roughly conforms to the normal distribution.

Finally, use the VaR curve to warn of severe fluctuations.

The ones with * on the right represent strong influence points.

We construct the F test through the studentized residuals, and finally get the t test to detect abnormal points. by

stdres<-rstudent(lm.sol)

To get the studentized residual, and then use the formula

To calculate Fj and finally convert it to tj,

t=sqrt((144-51-1)*stdres^2/(144-51-stdres^2))

Finally we can check ifThen it is an abnormal point.

R language execution

res<-t>abs(qt(.025, df=92)) 

 You can directly get a Boolean value greater than the corresponding t value.

 A value of True may be an abnormal point.

prediction

We used HoltWinters to predict our oil price range

The true value is basically within the predicted range, but it is still more difficult to predict the net profit.