Skip to content
All library documents

Java Libraries for ARIMA and GARCH Forecasting

Article Quant Q&A · Author: mde
Read the original

Summary

The document asks for a free Java library that can produce forecasts from an ARIMA model whose parameters were already estimated in Matlab, with GARCH used for variance estimation. The user wants to run the forecasting step inside a Java application and also considers calling a C or C++ library through JNI or JNA. Remote access to Matlab or R is described as unsuitable for the application, and writing a forecasting implementation from scratch may be impractical.

The answer points to two Java time-series projects as possible starting points. It does not explain whether either library supports the requested GARCH forecasts, accepts externally estimated parameters, or matches Matlab’s forecasting behavior. No comparisons, benchmarks, or implementation details are given, so the recommendations would need to be checked against the model and integration requirements before use.

Key ideas

  • The request concerns forecasting with pre-estimated ARIMA parameters in a Java application.
  • GARCH variance forecasts are part of the stated use case, but the answer does not confirm library support for them.
  • Two Java time-series projects are suggested as candidates for further evaluation.
  • Compatibility with externally estimated parameters and Matlab behavior remains unverified.

Tags

Statistics
Full text
# Please advice free Java library for classical time series forecasting


# Please advice free Java library for classical time series forecasting












I've got an ARIMA model (with a GARCH model for variance estimation) and parameters estimated in Matlab for my set of data. Now I need to use this model in my Java based application for making forecasts for new data.

Are there any free Java libraries with such functionality (mainly interested in analog of ARIMA/forecast functionality from Matlab without estimating etc.)?

I only found commercial SuanShu (from NumericalMethods) and IMSL from RogueWave. They are cool but cost money.

Of course there is way to do own implementation for forecasting but not sure if I have time for that. Using remote connection to Matlab / R server is too not suitable in my case.

So any help would be appreciated.

PS: Actually any C/C++ libraries would be fine too cause I could use JNI/JNA for calling it.

## Answer by Damianos P. Melidis (score 1)

https://quant.stackexchange.com/a/37868

For Java you may try:

- https://github.com/signaflo/java-timeseries https://github.com/signaflo/java-timeseries/wiki/The-timeseries-package https://github.com/signaflo/java-timeseries/wiki/ARIMA-models

- https://github.com/Workday/timeseries-forecast

Hope this helps!

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.