-1
$\begingroup$

I have been tasked with comparing the capabilities of different startups offering AI-assisted data preprocessing. Due to legal reasons I cannot offer company data for the benchmarking, not even simulated using the same structures and patterns, which means that I either create a dataset with not-company-related simulated data myself or find one on the internet.

To clarify things, by data I mean timeseries. Some simple examples of what I would expect: centering/aligning the timeseries using e.g. cross-correlation, denoising them, finding out correlations, outlier detection, etc.

Are there any (free) datasets designed to be hard to preprocess the data? I could not find any adequate. If not, what is the best way of constructing an own challenging, simulated dataset for this purpose?

The purpose of the preprocessing step is to clean the timeseries data before it is fed into a neural network in order to be classified and tagged.

$\endgroup$
0
$\begingroup$

You could try homomorphic encryption of your data and then deliver that to the different startups. This is the business model of Numer.ai.

$\endgroup$
0
$\begingroup$

Some useful free real world time series data for testing and benchmarking are:

The sunspot time series.

Hyndman's data library.

And I don't know the technical details, but you can download all sorts of financial times series (trading and stock indexes) from Yahoo and Google.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.