Hi,
I'm working on a social network analysis problem.
My goal of the analysis is to predict links in this network.
My network is bipartite and i want to split it into train, validation and test sets ( every set is a network ) in order to check my model efficiency.
what is the best way to split my initial network into train, validation and test networks.
Do you mean the best dataset split ratio? Or are you talking about techniques to do the splitting?
I'm really new in ML, but if it is about the first case, I started reading about K-fold cross-validation, where you should split the dataset into K folds, then use one as a testset and the union of the remaining as the testset. You do it for all the folds, and get the average of your measurements:
https://medium.com/@rabinpoudyal1995/train-test-split-and-cross-validation-in-python-434ecba10909
Gabriela Bosetti !
Thanks for your reply
Actually my question is about techniques to do splitting...
Best regards
Hi Masoud,
You may consider this recent paper : https://arxiv.org/pdf/1811.12159.pdf
Methodology is described in Section 2. In particular, 2.1 may answer to your question.
Regards
I was also looking for the same. Thanks for those wonderful answers
- Badges
- Science topic