0
$\begingroup$

The Problem I'm trying to solve is predicting the performance of the computer system based on the past data, Data is taken by the system every 5 minutes and it contains tons of useful features to predict the performance. But as I told you it is a time-series data taken for the same set of features over a time interval. So how I can effectively model this situation with a learning model which preserves its "time-series" ness of the data?

note: Accuracy is not so important, I want to model which preserves "time" features. Thanks

$\endgroup$
  • $\begingroup$ Build memory models that might include arima structure and identifiable/latent determinstiic structure like hour-of-the-day , day-of-the-week , day-of-the-month etc. $\endgroup$ – IrishStat Jun 25 '18 at 9:49
  • $\begingroup$ But I think arima model alone will not help, it more of prediction problem than forecasting $\endgroup$ – PreeJackie Jun 25 '18 at 9:54
  • 1
    $\begingroup$ your comment confuses me as prediction=forecasting in my book..... I was not suggesting arima alone .. I was suggesting a hybrid model that might include arima and possible deterministic structure that is WTD ... waiting to be discovered . $\endgroup$ – IrishStat Jun 25 '18 at 10:41
  • $\begingroup$ Are you dealing with multivariate time-series? Which kind of prediction are you interested in? If you aim at predicting some outcome (e.g.: computer system performance) from past values of some features, well that does sound like (multivariate) forecasting to me. Take a look at this post. $\endgroup$ – fsamu Jun 25 '18 at 10:54
  • $\begingroup$ @fsamu Yeah I'm dealing with multivariate time-series (sorry for not mentioning). I'm interested in predicting the throughput of a computer system, based on the features like system usage, CPU cycles etc. But I don't know about Long short-term memory neural net. Want to have a look, meanwhile do you know any beginner friendly tutorial or notes for LSTM, please share :) $\endgroup$ – PreeJackie Jun 25 '18 at 12:34

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.