Credit_Risk_Analysis

• The purpose of this analysis was to apply machine learning to solve and predict credit risk. Refer to credit_risk_resampling.ipynb and credit_risk_ensemble.ipynb

Challenge

• Use Resampling Models to Predict Credit Risk (Deliverble1):
  • Naive Random Oversampling results: accuracy is 64%, precision of high_risk is 1%, and the recall is 73%
  • SMOTE oversampling results: accuracy is 66%, precision of high_risk is 1%, and the recall is 63%
• Use the SMOTEENN algorithm to Predict Credit Risk (Deliverble2):
  • Undersampling results: accuracy is 52%, precision of high_risk is 1%, and the recall is 69%
  • Combination results: accuracy is 64%, precision of high_risk is 1%, and the recall is 72%
  • 
• Use Ensemble Classifiers to Predict Credit Risk (Deliverble3):
  • Balanced Random Forest Classifier results: accuracy is 78%, precision of high_risk is 3%, and recall is 69%
  • Easy Ensemble Classifier results: accuracy is 93%, precision of high_risk is 9%, and recall is 92%

Summary

• Recommendation: Base on the results, We should use Easy Ensemble AdaBoost Classifier because it had high accuracy and good balance of precision and recall.
• I would not recommend to use the first four models to predict credit risk.