My work is on multi classification. I am planning to use libsvm initially with RBF kernel.In my dataset there are around 60 features and 5 classes. what should be the proper size of my dataset ? I mean how many records i have to use in training to avoid over fitting problem?
Ekarin Varutbangkul
Hi krupal,
As a rule of thumb, the size of your training set should be at least 10 times of your features to avoid the curse of dimensionality problem. Therefore, you should have at least 600 instances in your training set.
If your data set is small, you may perform dimensionality reduction such as feature selection or feature extraction.
1 votes
0 thanks
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training
More Krupal Parikh's questions See All
Similar questions and discussions