I was wondering if Granger causality would be an efficient tool for searching for relevant input data for an SVM system. For example if I want to forecast SP 500 returns, I could put in my input data the libor rate, the USD/EUR exchange rate, the FTSE index, the Eurostoxx 50 index... Would it be relevant and meaningful to investigate, before adding those data to the SVM input features, if there is Granger causality beetween the SP index returns and the USD/EUR exchange rate and so on? Is Granger causality a adapted tool for SVM input data preselection?

  • $\begingroup$ Granger causality is about linear relationships between stationary time series. I do not know much about SVM, but if it is about non-linear relationships, granger causality seems redundant. On the other hand as a preselecting it might help. Do a comparison and then decide. $\endgroup$ – mpiktas Aug 24 '12 at 8:48
  • $\begingroup$ Yes, I was talking about Granger Causality as a preselecting tool for SVM input data. $\endgroup$ – marino89 Aug 24 '12 at 9:06
  • $\begingroup$ Basically, I'm looking for a way to preselect relevant input data for SVM in order to avoid spurious correlations. I've thought about Granger Causality and Random Forest. $\endgroup$ – marino89 Aug 24 '12 at 9:42
  • $\begingroup$ what do you think of Generalized additive model? en.wikipedia.org/wiki/Generalized_additive_model. Here is an interesting post I found on R-bloggers :r-bloggers.com/… $\endgroup$ – marino89 Aug 27 '12 at 8:25

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.