Skip to content

Latest commit

 

History

History
58 lines (32 loc) · 1.35 KB

Summary_Report.md

File metadata and controls

58 lines (32 loc) · 1.35 KB

Summary Report

Credit Risk Resample:

Random Sampling

Smote Over Sampling

markdown

Under Sampling

Combination Sampling

  1. Which model had the best balanced accuracy score?

Balanced Accuracy Score is almost the same in 3 models, Random & SMOTE Oversampling Models have 84% & Undersampling Model has 82%

  1. Which model had the best recall score?

I beleive SMOTE Oversampling Model for Recall Score, even though Undersampling provides higher Recall score for high-risk applications but it will compromise Recall Score for low-risk applications.

  1. Which model had the best geometric mean score?

The Best Model for Geometric Mean Score is the SMOTE Oversampling Model.


Credit Risk Ensample:

Balanced Random Forest Classifier

Easy Ensemble Classifier

  1. Which model had the best balanced accuracy score? Easy Ensemble Model has a perfect balanced accuracy score with 93%

  2. Which model had the best recall score? Easy Ensemble Model has the best recall score for high and low risk loan applications.

  3. Which model had the best geometric mean score? Easy Ensemble Model

  4. What are the top three features?

    • 0.0787 : 'total_rec_prncp'
    • 0.0588 : 'total_pymnt'
    • 0.0562 : 'total_pymnt_inv'