I am sorry for the mistakes I will make in this post, I'm new to deep learning ^^'

I am trying to build an LSTM model that can help me predict some unique value according to time series indices and previous time series indices correlated with the value.

To make it simple, I have an indice that varies throughout time and "place", it will be my input, and a resulting value (the output) that changes only once a year.

Bellow is a screen shot of the data: how the data looks like ;)

I thought of making this problem a classification problem, but the way I understood classification in this course, is to put (previously?) some values in an "output interval"

I have also tried to follow this tutorial, but I'm getting some weird results (since maybe I have changed the datetime into numeric data?), and a huge rmse of (above 70%)

Thank you for your time and advices:)

  • $\begingroup$ Could you please elaborate more on your data? What is the input sequence? And what is the nature of your output / dependent variable? Classification and regression problems are very different in nature, and you shoulf make your choice based on the nature of your data and the current task. $\endgroup$ – Leevo Sep 16 '19 at 12:10
  • $\begingroup$ Yes of course, the input sequence is the time serie (the indice according to datetime) and the "value" corresponding: there is a unique "value" per year and it's very depending on the time serie associated and to the previous time series and "values". We can say that the "value" is a yield and that the "indice" is some vegetation indice. I hope that it's less confusing I think that a regression solution will be more appropriate :) $\endgroup$ – Nour Sep 16 '19 at 12:20
  • $\begingroup$ Ok I have few other questions: the input time series is a float variable? And you goal is to classify a series in the right year? (So it's a classification problem?) How many output classes do you have? $\endgroup$ – Leevo Sep 16 '19 at 12:23
  • $\begingroup$ The input series is a float variable (in [0,1]) and my goal is the give "ThaVariable" according to the "last" serie and following the logic of the previous series and "variables". I have one output class: the "variable" $\endgroup$ – Nour Sep 16 '19 at 12:26
  • $\begingroup$ sorry, I'm not sure I understood. You want to predict the next step of the same variable? Is it a univariate regression? $\endgroup$ – Leevo Sep 16 '19 at 12:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.