Credit-Risk-Modelling.md

Credit Risk Modelling

Loading the data

Data can be found Here. We will exclude the phone variable. We will also convert the dependent variable default to a factor

library(tidyverse)
credit_data <- read_csv("https://raw.githubusercontent.com/obaidpervaizgill/CreditRiskModelling/master/credit.csv") %>% select(- 16) %>% 
  mutate(default = factor(default))
head(credit_data)

# A tibble: 6 x 16
  checking_balance months_loan_dur~ credit_history purpose amount
  <chr>                       <dbl> <chr>          <chr>    <dbl>
1 < 0 DM                          6 critical       furnit~   1169
2 1 - 200 DM                     48 good           furnit~   5951
3 unknown                        12 critical       educat~   2096
4 < 0 DM                         42 good           furnit~   7882
5 < 0 DM                         24 poor           car       4870
6 unknown                        36 good           educat~   9055
# ... with 11 more variables: savings_balance <chr>, employment_duration <chr>,
#   percent_of_income <dbl>, years_at_residence <dbl>, age <dbl>,
#   other_credit <chr>, housing <chr>, existing_loans_count <dbl>, job <chr>,
#   dependents <dbl>, default <fct>

# Tabulating the dependent variable
table(credit_data$default)

 no yes 
700 300 

Creating the train and test sets

We will use the CreateDataPartition function of the caret package to ensure proportionality of the response variable in the train and test sets.

library(caret)
indexes <- createDataPartition(credit_data$default, times = 1, p = 0.7)
train <- credit_data[indexes$Resample1, ]
test <- credit_data[-indexes$Resample1, ]

Training the model

We can use logistics regression since the response variable is binary.

library(randomForest)
train_model <- randomForest(default ~., data = train)

# Predicting
test_pred <- predict(train_model, test[, 1:15])

Evaluating the model performance

actual_test <- credit_data[-indexes$Resample1, "default"]
actual_test <- as.factor(actual_test$default)

confusionMatrix(test_pred, actual_test)

Confusion Matrix and Statistics

          Reference
Prediction  no yes
       no  195  57
       yes  15  33
                                          
               Accuracy : 0.76            
                 95% CI : (0.7076, 0.8072)
    No Information Rate : 0.7             
    P-Value [Acc > NIR] : 0.01249         
                                          
                  Kappa : 0.3407          
                                          
 Mcnemar's Test P-Value : 1.352e-06       
                                          
            Sensitivity : 0.9286          
            Specificity : 0.3667          
         Pos Pred Value : 0.7738          
         Neg Pred Value : 0.6875          
             Prevalence : 0.7000          
         Detection Rate : 0.6500          
   Detection Prevalence : 0.8400          
      Balanced Accuracy : 0.6476          
                                          
       'Positive' Class : no              
                                          

Our model is upto 74% accurate in predicting loan defaulting. To improve the model performance, we would need to peforme feature engineering.