I have a dataset with 'tag-like' groupings: Id tags 0 item1 ['friends','family'] 1 item2 ['friends'] 2 item3 [] 3 item4 ['family','holiday'] So a row can belong to several groups. I want to create an object similar to groupby, so that I can use agg etc. df.groupby('tags').count() expected result tags count 0 'friends' 2 1 'family' 2 2 'holiday' 1 But of course it won't work because it treats the whole list as the key, rather than the individual tags. Here's an attempt tagset = set(df.tags.explode()) grpby = { t: df.loc[df.tags.str.contains(t, regex=False)] for t in tagset } From what I understand, groupby objects are structured a bit like this. But how to make it a groupby object? So that I can do things like grpby.year.mean() etc? Continue reading...
