Custom Dataset format

[arita37]

Hello,

if we wan to use on our own dataset (tabular csv)
format.

what will be the format of the dataset ?

thx

[felixleopoldo]

Hi,

It is CSV files still, where the first line tells the labels of the variables. For multinomial data, the second line should contain the number of states for the variables.

Have a look at this example for continuous data and this example for multinomial data.

Just let me know if you run into trouble or have any questions

[arita37]

Sure, Thx. Is there any pre-processing required for a new dataset ? Any constraints on the format requirements ? Thx

…

On Sep 4, 2022, at 23:34, Felix Rios ***@***.***> wrote:  Hi, It is CSV files still, where the first line tells the labels of the variables. For multinomial data, the second line should contain the number of states for the variables. Have a look at this example for continuous data and this example for multinomial data. Just let me know if you run into trouble or have any questions — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.

[felixleopoldo]

For continuous data, it is recommended to standardize first. For multinomial data, some algorithms may require enumeration of e.g. k states from 0 to k-1. Also, have a look at the Benchpress paper for a further description of the data format.
In general, you might have to transform your data to meet the assumptions of the algorithms you are using. You can do pairwise scatterplots of the data using the ggally_ggpairs module.

I hope this could help you

[arita37]

Sure, thanks
The columns name should be in the json format ?

[felixleopoldo]

No, it should be a CSV file, just have to look into the other data files here and you will see the structure.
Here is a real data JSON config example that you may also look at. If you don't know the true graph, graph_id should be set to null.