1
$\begingroup$

I want to build a timeseries forecasting model which uses LSTM layers, I want to know what analysis we have to do prior to fitting the model? My data is recorded every 20 minutes, I performed Dickey fuller test, it proved that my data is stationary.

I have 50k records. My task is to take 24 hours of past data, and predict next 12 hours of data. How can I confirm that the data is suitable or not suitable for timeseries analysis? How can I choose correct number of past records so that I can predict the next few records accurately? I'm planning to plot auto-correlation graphs, trend graphs etc.

Can you give me the list of tests and plots which I have to perform before building the model?

Improve this question
$\endgroup$
3
  • 1
    $\begingroup$ This is a lot of stuff for one paragraph. Several things are repeated twice. A clean up would be helpful... $\endgroup$
    – Richard Hardy
    Commented Feb 19, 2022 at 20:48
  • $\begingroup$ Thank you richard, I cleaned up the paragraph. $\endgroup$
    – MasterBlasterCoder
    Commented Feb 21, 2022 at 8:49
  • 1
    $\begingroup$ Looks better now. You still ask about suitability for time series analysis/forecasting twice, but perhaps you mean these as separate questions. Splitting into several paragraphs might improve readability, too. $\endgroup$
    – Richard Hardy
    Commented Feb 21, 2022 at 14:02

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.