Coding the SARIMA Model : Time Series Talk

Coding a full SARIMA model using real data!

Understanding why we might use the SARIMA model, using ACF / PACF to understand the order of the model, and making the predictions.

Code used in this video:

[egon-csabaosvath]

Hello. I'd like to thank you for these videos and all the effort you put into creating them. I really appreciate your work.

[JayeshPatil-siwu]

Hi, It's a nice video. Can you make a video in which we need to decide p, d, q and P, D, Q on the same dataset in the same problem? Thank you.

[pocketrocket]

For some reason acf() has changed. For an easy correction use:
acf_vals = acf(first_diff, nlags=20)
So you get the 20 needed data points.

[parthify]

Absolutely amazing. Great job of explaining it. I came here for the code but I am going to watch the whole video series on time series forecasting.

[zollen]

It would really help if you could explain how you decided AR(0) and MA(0) of the ARIMA(0, 1, 0), and the Seasonal AR(1) and Seasonal MA(1) of the seasonal component of (1, 0, 1, 12)

[algorithmo]

Why did you take my_seasonal_order = (1, 0, 1, 12) at 4:34? How did you find AR = 1 and MA = 1?

[saeedseyedhossein]

hey man. your videos are great and inspirational. If will be great if make a video about ARIMAX models. thanks

[jongcheulkim]

Thank you very much. This is very helpful.

[sabyasachidas]

Hi ritvik. Thanks for the awesome videos. Can you create a video on how to decide SARIMA parameters- pdq and PDQ?

[plenilune]

I hope I could have a professor like you in my collage.

[rezabazargan]

please explain in another video the best way to choose P, Q, D & p, q, d

[bhanukadissanayake]

Thanks a lot. Can you explain how did you choose P, D and Q for the seasonal part?

[linustan]

Can you please explain what would the values be if the lags are not 12? would the p and q be different? From what i infer now is that since acf and pacf show 12 and m = 12, therefore P and Q is 1??? what if acf or pacf show lesser/more than 12 and what if acf is diff from pacf?

[nicok]

Awesome thanks a lot. I would really like to see a video like this about ARCH and GARCH. :-)

[ahmedaliAhmed_Ibrahim]

Hello I have a forecasting and Anomaly detection module which uses ARIMA for modeling,
Customer's data are streams of time series data stored in a database and the module first fetches history data to select the best model from a predefined sets of (p, d, q) and train it.
my question is what is the min required history for ARIMA to work fine, and is there an equation to calculate it give the (p, d, q) values.
mainly the data have daily and weekly seasonality.

[iftikhar]

but if we have multiple seasonal pattern in our data then what will be the sasonal order in the sarima model?

[SonuGupta-hktb]

Great explanation! Have a query - the data is sampled daily, the column to be forecasted is cumulative and has monthly seasonality. If I take the value of M=30, it works well for months with 30 days but get confused if the month has 31 or 28 days. How should I select the value of M?

[bhavinmoriya]

Why did you choose m = 12. Is it because you have monthly data and each year pattern repeats?

[bhavinmoriya]

If I have daily data and every year pattern repeats, should I go for m = 365?

[fatinelbouhssini]

I have a daily turnover data (for 6 years) and like yours it has a seasonal process of 1 year so 365 observations for my case. Do you think that I should take 365 as the number of lags?