I am trying to cluster a set of tags in huge sparse matrix of user-tag in matlab, say:
35000*74000 with 99% sparsity.
I probably have two choices:
1: Making Co-occurrence matrix of tags and apply clustering on Tag-Tag co-occurrence matrix.
2: Considering users as features of tags and use Transpose(user-tag) for tag clustering.
I have no idea which one is better, but both of them have problem with large size of users and tags.
Really appreciate any comment or proposed solution.