First, determine the order p of the ARMA model, and then input it as s(n), which is the quarterly data of copper futures prices for 20 years, with a maximum of 80. Then find the autocorrelation of p S(n), use the Levison-Durbin algorithm to derive the parameters of p ARMA models, and finally use the simplest white noise to find the next four data through the linear system method, which generates 90 Random white noise data, h(n) composed of p ARMA model parameters; perform convolution and the data will come out. The 81st to 84th data are the data you want to predict.

Levison-Durbin algorithm is the number that will be introduced in this introduction to ARMA model spectrum estimation. You can refer to it in the book. It is too complicated to describe here.

You can also search for levison-durbin or Yule-Walker equations on www.answers.com