I am trying out hyperparameter tuning vs manually selecting best parameter (max_leaf_nodes) on a Decision Tree model with mean absolute error as scoring. In theory both should give me the same mae and max_leaf_nodes. But both are giving me different mae. Also if I change the value of cv in GridSearchCV I get different results. So basically I have two questions: 1.) Why am I getting different Max_leaf_nodes and mae in both the cases? 2.) How do I determine the value of cv in GridsearchCV because I get different results for cv = 3, cv = 5 and cv = 10.
The "for" file attached is where I manually set the max_leaf_node parameter and the "grid" file is where I use GridSearchCV(). Both give me different results!!
Paolo Inglese
I am performing cross validation! The GridSearchCV in addition to searching for best hyperparameters also performs cross validation ( hence the "CV" in GridSearchCV.Ok I got the answer to my second question as it makes sense that selecting different values of cv will give different results since the splits are different.
But why am i getting different mean absolute error values for both cases?
Dear Shrey Jain
These links might be useful, have a look:
1. https://www.dezyre.com/recipes/optimize-hyper-parameters-of-decisiontree-model-using-grid-search-in-python
2. https://stackoverflow.com/questions/59147246/how-can-i-improve-the-accuracy-of-my-prediction-from-a-decision-tree-model-using
Kind Regards
Qamar Ul Islam
Paolo Inglese
Ok I got your point. You are right. So you mean to say that in the first code, the MAE I am getting is from the test data (as I am predicting on the test data) and in the second code where I use GridSearch the MAE is obtained only from the train data (as I am not using the test data). Hence the different MAE values. Please correct me if I'm wrong.- Badges
- Science topic
- Similar topics
- Filing