Data preprocessing

[1369959395]

Hello, is it necessary to standardize and normalize the data before dynamic Bayesian network prediction?

[dkesada]

Hi,
it depends on each specific case. In theory, standarizing and normalizing the data should be benefitial to the accuracy of the model. In my experience, it is case specific and it does not always improve the results. I would advise you to try both aproaches and see if the prediction accuracy improves.

Just keep in mind that you have to standarize and normalize the data prior to learning the structure and parameters of the DBN. Then, after forecasting you have to undo this process if you want to see the real predicted values and evaluate the real accuracy.

Best regards.

[1369959395]

OK， Thank you!