search

dkesada / dbnR

Gaussian dynamic Bayesian networks structure learning and inference based on the bnlearn package
GNU General Public License v3.0
44 stars 10 forks source link

Some questions about smoothing #24

Closed 1369959395 closed 1 year ago

1369959395 commented 1 year ago

How to explain why the obtained state of the past time is so different from the real data after smoothing.

dkesada commented 1 year ago

Hi, I'm afraid I cannot give you a clear answer to that question. There can be a lot of reasons why your model does not fit your data when performing smoothing, given that it is not a trivial operation. Maybe your data does not show a clear autoregressive component, or maybe it is insufficient to properly learn the underlying distribution. It could also be that the distribution learned from a Gaussian DBN does not fit your dataset adequately, or that you have to increase the size of the network or perform shorter term smoothing. It's a similar case to when you perform forecasting and it does not return very good results, there are many reasons why this can be the case as with any other machine learning model.