Discover all of the brand-new features and improvements to existing capabilities in the Dataiku 11.3 updateLET'S GO

Instability of results Random Forest

LA
Level 1
Instability of results Random Forest

Hello,

I trained the random forest model for the first time and I had very good results after I retrained the same model with the same parameters on the same dataset and I obtained very different results ( at the beginning I had good results (F1 score = 85%) and  now F1 score = 65% )

Any solutions please.

0 Kudos
2 Replies
AlexT
Dataiker

Hi @LA ,

Can you share what your Sampling & Splitting policy is? Perhaps the underlying dataset changed and the sample used was different train/test datasets which could explain the different results.
Can you try using "Explicit extracts from two datasets" and see if you get consistent results?


Thanks

0 Kudos
LA
Level 1
Author

Hi Alex,

 

Thanks for your answer. 

 

I really didn't touch the data or the model, I just retrained the model without modifying anything.

0 Kudos