I am having a time-series classification problem. I tried to using movingaverage and firstdifferences as two models to perform my classification. However, the results I am getting are pretty bad.

Therefore, I am thinking whether there are any other types of time-series data that I can use instead (i.e. instead of movingaverageandfirstdifferences`). If so please suggest me.

I am happy to provide more details if needed.

  • $\begingroup$ I do not think these are classification models. How did you classify time series with them? $\endgroup$ Nov 28 '19 at 8:46
  • $\begingroup$ @RichardHardy I use LSTMs to classify my timeseries :) $\endgroup$
    – EmJ
    Nov 28 '19 at 22:16
  • $\begingroup$ Why do you need moving averages or first differences then? Or are they a part of LSTM and you are wondering about the specification of LSTM? (Probably a stupid question since I am not all that familiar with LSTM.) $\endgroup$ Nov 29 '19 at 6:08

Both of the "models" that you listed are particular types of ARIMA models where one series exists and one wishes to use the history of the series to generate a forecast. Examples of this are here Forecasting/predicting total sum of donations (following GLM with poisson family and log link) and here Help me about using ARIMA forecasting rainfall .

If you wish you can post one of your data sets (must be equally spaced in time) and I will try and help further. The methodology underlying model identification is briefly introduced here. How to interpret PACF? .


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.