Bankruptcy Prediction through Classifiers
Compared classifiers in identifying whether a bank is about to go bankrupt or not, on the basis of 10 internal factors.
Decision Tree using CART algorithm:
| Prediction | Bankrupt | Still operating |
|---|---|---|
| Bankrupt | 60 | 28 |
| Still operating | 50 | 262 |
Accuracy : 80.5 %
95% CI : (0.7627, 0.8427)
No Information Rate : 0.725
P-Value [Acc > NIR] : 0.0001352
Kappa : 0.4786
Mcnemar’s Test P-Value : 0.0174171
Sensitivity : 0.5455
Specificity : 0.9034
Pos Pred Value : 0.6818
Neg Pred Value : 0.8397
Prevalence : 0.2750
Detection Rate : 0.1500
Detection Prevalence : 0.2200
Balanced Accuracy : 0.7245
K-Nearest Neighbor algorithm
| Prediction | Bankrupt | Still operating |
|---|---|---|
| Bankrupt | 59 | 29 |
| Still operating | 51 | 261 |
Optimal value : k = 29
Accuracy : 80 %
95% CI : (0.7574, 0.8381)
No Information Rate : 0.725
P-Value [Acc > NIR] : 0.0003348
Kappa : 0.4652
Mcnemar’s Test P-Value : 0.0188810
Sensitivity : 0.5364
Specificity : 0.9000
Pos Pred Value : 0.6705
Neg Pred Value : 0.8365
Prevalence : 0.2750
Detection Rate : 0.1475
Detection Prevalence : 0.2200
Balanced Accuracy : 0.7182
- Bayesian Generalized Logistic Regression
| Prediction | Bankrupt | Still operating |
|---|---|---|
| Bankrupt | 37 | 8 |
| Still operating | 73 | 262 |
Accuracy : 79.75 %
95% CI : (0.7547, 0.8358)
No Information Rate : 0.725
P-Value [Acc > NIR] : 0.0005153
Kappa : 0.3781
Mcnemar’s Test P-Value : 1.151e-12
Sensitivity : 0.3364
Specificity : 0.9724
Pos Pred Value : 0.8222
Neg Pred Value : 0.7944
Prevalence : 0.2750
Detection Rate : 0.0925
Detection Prevalence : 0.1125
Balanced Accuracy : 0.6544
Result: The KNN model performed the best, at 80.5 %