01-14-2015 03:50 AM
Do you think this is the right algorithm to approach the problem?
Start with a dummy parameters(omega, alpha, beta)
Calculate the cost function ( J=mod(h(xi)-yi)^2/n y=h(xi)=omega + alpha * xi-1 + beta * ei) is the regression function)
And do a Gradient Descent algorithim with a certain learning rate to find the optimum value of alpha , beta and omega where J is min .
Can this works this way?
Thanks and Regards,
01-14-2015 05:28 AM
Hi , In addition to my previous question,Can someone share the algorithm for ARIMA model, I want to cross check the output with real time data .
01-14-2015 02:58 PM
In PROC ARIMA, specify METHOD=ML in the ESTIMATE statement to get maximum likelihood estimates. The algorithm is Marquardt's method for nonlinear least squares estimation.