Limiting Decision Tree branching factor
Hi everybody ,
I have a dataset with 5 attributes , one is nominal and it has large number of possible values (~5000 values) , I want to train a decision tree on this dataset but the problem is that when I include this feature , the branching factor for this attribute is very large and so model doesn't in the memory (I use 74 GB of main memory) , my dataset has about 620 K instances (rows) ,
Is it possible to put a limit on branching factor for this attribute ?
Thanks ,
Arian
I have a dataset with 5 attributes , one is nominal and it has large number of possible values (~5000 values) , I want to train a decision tree on this dataset but the problem is that when I include this feature , the branching factor for this attribute is very large and so model doesn't in the memory (I use 74 GB of main memory) , my dataset has about 620 K instances (rows) ,
Is it possible to put a limit on branching factor for this attribute ?
Thanks ,
Arian
Find more posts tagged with
Sort by:
1 - 1 of
11
no, you can't limit the branching factor - for each nominal value a single branch will be created. But probably an attribute with that many features is probably not the best choice anyway. But tell me, are the values a fixed set, or is possible that new data contains different, new values? In that case the example is useless anyways.
Best regards,
Marius