Can I interpolate a time-series without changing the meaning of the data?

29 Views Asked by At

I am planning do data augmentation through interpolation to aid the small dataset size available for my neural network to train on. How can I interpolate a time-series without changing its meaning?

1

There are 1 best solutions below

0
On BEST ANSWER

See:

Jäger, Sebastian and Allhorn, Arndt and Bießmann, Felix. "A Benchmark for Data Imputation Methods". Frontiers in Big Data. Vol 4, 2021. DOI: 10.3389/fdata.2021.693674. url (accessed Oct 2022): https://www.frontiersin.org/articles/10.3389/fdata.2021.693674

In general, data imputation is an approximation to fill in missing data. It obviously changes the meaning of the observations. Whether the change in meaning is significant depends on a lot of factors, including how significant the attribute missing values is in impacting your measurement.