The purpose of this analysis is to create various machine learning models to see if I can create a model that accurately predicts credit risk.
- Naive Random Oversampling
- Accuracy score: 0.64
- Precision score: 0.01
- Recall/sensitivity score: 0.69
- SMOTE Oversampling
- Accuracy score: 0.66
- Precision score: 0.01
- Recall/sensitivity score: 0.63
- Cluster Centroids
- Accuracy score: 0.54
- Precision score: 0.01
- Recall/sensitivity score: 0.69
- SMOTEENN
- Accuracy score: 0.67
- Precision score: 0.01
- Recall/sensitivity score: 0.76
- Balanced Random Forest
- Accuracy score: 0.79
- Precision score: 0.03
- Recall/sensitivity score: 0.7
- Easy Ensemble AdaBoost
- Accuracy score: 0.93
- Precision score: 0.09
- Recall/sensitivity score: 0.92
AdaBoost had the highest scores with an accuracy of 0.93, precision of 0.09, and sensitivity of 0.92. None of the models have a very good precision score, so I would not recommend them if you would like high precision in detecting credit risk. If you value sensitivity in your model, I would recommend using the AdaBoost as it had the highest sensitivity rating.