I've got a time series with seasonal and trend components, the interval is 1 hour (the size of the input data is: 24 values per a day * 365 days), the task is to predict 24 values (for the next day). What I've tried so far (with the help of stats.stackexchange, thanks!):

  • Linear regression with multiple features (is_weekday, hour etc)
  • FB Prophet
  • LSTM neural network

The problem is that though FB Prophet and SARIMAX perform the best, I can't integrate them in my Java code (I've found a few libraries on GitHub, but they generate some random forecast in practice).

I probably can hardcode the weights of my LSTM model into my Java code as well as use linear regression (there's a simple library for that).

Are there any other techniques to forecast a time series with seasonal and trend components I've missed?



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.