R entry <2>-time series research

The article above

After writing the essay the day before yesterday, I am very proud to show off it with the big guy @ Ba Gong who brought me into data mining and SAS basics. Then I have gained a bunch of time series materials. Thank you very much!

ARIMA does not need to know the principles of graphics, ACF and PACF, because the software has helped us solve the dynamic equation.

In summary

1) The key to ARIMA is to look at the graphics. Looking at ACF and PACF, the formula doesn't have to be clearly understood, because the software has already helped solve the dynamic equation.

2) It is said that auto. ARIMA is selected based on the AIC statistics, not always providing the most robust results. Therefore, manual debugging is very important (you should check the data)

3) The key point is the stable transformation of the sequence (the difference can be set to Level 2 at most), The Arima coefficient is determined, and the seasonal difference is required when there is a seasonal effect. At the same time, you can consider the holiday effect for retail and so on (not necessarily on the day of the holiday, all those related to the holiday are counted as the holiday effect)

In addition, the teacher mentioned that recent phyton already contains many R statistical functions. Phyton can already Replace r to a certain extent, which makes me stunned by the person who studied R from the beginning.

The above, some makeup records, the specific materials have not been viewed

I took my computer home on the National Day and practiced R well!

R entry <2>-time series research