As I observed that if the dataset is small then tree based ML method is good, if dataset is imbalanced then svm is bad, if data is non linear then Deep learning. Could any one suggest me how can I start selection of ML method depending on structure of dataset and pattern initially.
Sayan Putatunda
When dimension is greater than size of instance then one can first try dimension reduction techniques such as PCA and then apply any ML algorithm for supervised learning. With high imbalance, Random forests with undersampling/oversampling works well in most occasions.
0 votes
0 thanks
Erik Cuevas
Machine learnin techniques opérates in all these cases. I recomend you bosting schemes. They conider cases in which the data set is reduced and even with outliers
1 votes
1 thanks
- Badges
- Science topic
More Soumen Roy's questions See All
Similar questions and discussions