Our dataset is relatively small (303 x 14) and so we decided to use synthpop package in R.

The basic idea of synthetic data is to replace some or all of the observed values by sampling from appropriate probability distributions so that the essential statistical features of the original data are preserved.

My question is, is it safe to create synthesized data and use it as the dataset to train a random forest classifier in order to make predictions on the original data? So I would use the synthesized data as train, and original data as test.

The synthesized data resemble the actual data as closely as possible, but they are not the exactly same data (perhaps original data is included in it? I am not really sure.) I wonder if this would be case where original data is not unseen data. Synthesized data is built on it, so it has a same mean standard deviation etc..

Any tips would be appreciated!

  • 1
    $\begingroup$ You should not use the synthetic data for training and original for test. This would just lead to an overly optimistic performance estimation. Instead, split your data first into training and test set. Then create new data points on the training data. $\endgroup$ – Laksan Nathan May 22 at 21:38

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.