0
$\begingroup$

I have 41 continuous columns and all are distributed roundly to each pair: enter image description here

I used:

  • SMOTE for resampling data ( my dataset is imbalanced)
  • Test dataset: last month in the data dataset. Train dataset: remaining data set
  • GridSearchCV for tunning parameters for RandomForest and LinearRegression and got ROCs like this:

Linear Regression

enter image description here

Random Forest

enter image description here

I believe this is a bad result. I think I missed an important step in feature extraction and feature selection to achieve good results and I believe that it related to the circular distribution. How can I improve my models (I think we can focus on circular distribution of continuous columns)

Improve this question
$\endgroup$

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.