I am trying to use LSTM to predict a time series data as you can see in the following image, the predicted graphs is very noisy:

enter image description here

The original data is looking like this: enter image description here

That I normalized it like this and fed it to the LSTM network:

enter image description here

Is there a way to get better result with less noise?

EDIT: I use keras implementation of LSTM and sklearn.preprocessing.normalize library for normalization.

  • $\begingroup$ I don't think we can answer this as it stands now. What are these plots you are showing us? What are the values? Is this time series data? What software, functions are you using? $\endgroup$ Aug 7, 2019 at 8:57
  • $\begingroup$ @user2974951: I edited my question. Yes the data is a time series that you can see in the second pic. And I used sklearn.preprocessing.normalize for normalization that you can see it's result in the last pic. I also use kears implementation of LSTM. $\endgroup$ Aug 7, 2019 at 9:48
  • $\begingroup$ Then this has to do with model building. We would need to see your code, what you've done with some explanations. $\endgroup$ Aug 7, 2019 at 10:37
  • $\begingroup$ @user2974951 In these situations, we've created canonical threads to serve as duplicate targets. See meta discussion here: stats.meta.stackexchange.com/questions/5273/… $\endgroup$
    – Sycorax
    Aug 7, 2019 at 13:33