Timeline for Decision tree classifier: possible overfitting
Current License: CC BY-SA 3.0
10 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Nov 10, 2017 at 11:36 | history | edited | CezarySzulc | CC BY-SA 3.0 |
Add new model with better prediction.
|
| Nov 3, 2017 at 21:20 | answer | added | Dan Jarratt | timeline score: 1 | |
| S Nov 3, 2017 at 14:58 | history | suggested | Michal_Szulc | CC BY-SA 3.0 |
Fixed spellings and added tag
|
| Nov 3, 2017 at 10:35 | comment | added | enterML |
There are a lot of things going on here. There are too many negatives. Try oversampling. Also, Decision Trees are prone to overfiitng. Try to use ensembles. And as @RicardoCruz said, max_depth is way too much. Typical values of Max_depth should be between 6-14
|
|
| Nov 2, 2017 at 21:21 | comment | added | Ricardo Cruz |
Decision trees are known for overfitting data. They grow until they explain all data. I noticed you have used max_depth=42 to pre-prune your tree and overcome that. But that value is sill too high. Try smaller values. Alternatively, use random forests with 100 or more trees.
|
|
| Nov 2, 2017 at 20:49 | review | Suggested edits | |||
| S Nov 3, 2017 at 14:58 | |||||
| Nov 2, 2017 at 20:00 | comment | added | CezarySzulc | I tried StratifiedShuffleSplit, but result is the same. Recall for positive in test set still is ~0.24 | |
| Nov 2, 2017 at 18:35 | comment | added | Hobbes | It's possible that your test data and train data are presenting a different 'story'. What if your try shuffling all your data and then cross-validating. StratifiedShuffleSplit will preserve the ration of positives. You can run cross_val_score on all your data. | |
| Nov 2, 2017 at 18:35 | review | First posts | |||
| Nov 2, 2017 at 19:41 | |||||
| Nov 2, 2017 at 18:30 | history | asked | CezarySzulc | CC BY-SA 3.0 |