I'm looking at a collection of problems where I need to forecast the probability of a continuous variable. My dependent variables are a combination of categorical and continuous. It looks like there are four primary candidate areas I should be reviewing and/or testing. Can anyone here shed any light on the relative strengths and risks associated with each of RNADE, Mixture Density Networks, Bayesian Additive Regression Trees, and standard Bayesian Linear Regression?

  • $\begingroup$ Is your data a time series? The reason I ask is because you use the word 'forecast'. $\endgroup$ – tom Nov 16 '17 at 3:16
  • $\begingroup$ Time is a feature of the dataset but it's only a weak predictor. $\endgroup$ – David Nov 16 '17 at 9:25

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.