Elliptic Dataset

[joshred83]

The elliptic dataset has two versions available, elliptic and elliptic temporal. The temporal version allows the user to select 1 of the 50 timesteps in the original.

Neither of the two datasets has the timestamp available as a feature, nor do they have the temporal attribute or a time dimension accessor. I'm hoping that someone knows the reasoning behind these decisions, so I don't walk into any landmines by adding a time dimension.

Does anyone know why it isn't available as a temporal dataset?

[MichalTesnar]

As in a related Kaggle dataset page we learn that: 'This anonymized data set is a transaction graph collected from the Bitcoin blockchain.' [...] 'Due to intellectual property issues, we cannot provide an exact description of all the features in the dataset.'

I have also downloaded it myself, and tried to check if some of the node features are in order given one of their features (meaning it is time, or at least could be). But it seems that the authors (or rather elliptic.co) have done their homework and really anynomized the data. So the answer is: we will never know, unless you do some real work for de-anonymizing the data.

[joshred83]

I think the actual answer is that it isn't what you'd normally call a temporal graph. I dug around in the code and the ibm paper and it seems like the data is actually 50 disconnected and unrelated graphs, but taken a 50 points in time.

[MichalTesnar]

Yes, that is true. But still each node has a timestamp included somewhere in the features, as according to the paper.