Skip to content

Predict credit risk using a variety of Resampling Models and algorithms.

Notifications You must be signed in to change notification settings

zobairhas/Credit_Risk_Analysis

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

12 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Credit_Risk_Analysis

Project Overview

Assist the client to predict credit risk using a variety of Resampling Models and algorithms.

Methodology

Tools/Programs/Languages used:

  • imbalanced-learn library
  • scikit-learn library
  • ETL
  • Machine Learning

I oversampled data via RandomOverSample and SMOTE algorithms using credit card credit dataset from LendingClub. ClusterCentroids algorithm was used to undersample the data. SMOTEENN algorithm was used for over-and-undersampling. Then I used two machine learning models to predict credit risk.

Summary of Results

  • Accuracy score of Random Oversampling using RandomOVersampler
  • Accuracy score was 0.657 which means that the model was correct 65.7% of the time.

image

  • Accuracy score of SMOTE Oversampling using SMOTE
  • Accuracy score was 0.662 which means that the model was correct 66.2% of the time.

image

  • Accuracy score of Undersampling using ClusterCentroids
  • Accuracy score was 0.544 which means that the model was correct 54.4% of the time.

image

  • Accuracy score of Combination (Over and Under) Sampling using SMOTEENN
  • Accuracy score was 0.644 which means that the model was correct 64.4% of the time.

image

  • Accuracy score of ML model using BalancedRandomForestClassifier
  • Accuracy score was 0.778 which means that the model was correct 77.8% of the time.

image

  • Accuracy score of ML model using EasyEnsembleClassifier
  • Accuracy score was 0.920 which means that the model was correct 91.1% of the time.

image

Recommendations

  • Overall, the models had pretty average accuracy scores. The clear winner was the ML model using EasyEnsembleClassifier.
  • The Accuracy score was much higher compared to any of the other models.

About

Predict credit risk using a variety of Resampling Models and algorithms.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published