Deal with highly unbalanced data classes

Asked Jan 24 '19 at 16:55

Active Jan 24 '19 at 16:55

Viewed 22 times

I got this horrible(hope you'll find the good in it) dataset, with 15 classes. Any suggestion to deal with it? I was wondering to group at least into 3 classes, however the first one is tremendously populated. Thanks

          CLASSES   tot
            <int> <int>
 1              1  3122
 2              2   716
 3              3   481
 4              4   288
 5              5   216
 6              6   150
 7              7   106
 8              8    86
 9              9    47
10             10    40
11             11    34
12             12    27
13             13    13
14             14    19
15             15    18

asked Jan 24 '19 at 16:55

3nomis

1

It's hard to say anything with no idea of what the classes signify, what other data you have available, or what you're trying to do. Provide context. – Kodiologist Jan 24 '19 at 19:03
You can do the oversampling or undersampling of classes to make it a balanced distribution. More details of the problem would be better. – Harshit Mehta Jan 24 '19 at 21:01
Ok thank you! I thought i needed to do some oversampling. – 3nomis Jan 25 '19 at 10:52

Deal with highly unbalanced data classes

0 Answers0