Load Data
Dataset is available here.
# Loading the data here
library(haven)
bank_loan_df <- read_sav("P4_bankloan_5000_clients.sav")
bank_loan_df$defaulted_loan<-as.factor(bank_loan_df$defaulted_loan)
bank_loan_df$education_level<-as.factor(bank_loan_df$education_level)
str(bank_loan_df)
## tibble [5,000 x 9] (S3: tbl_df/tbl/data.frame)
## $ age : num [1:5000] 41 30 40 41 57 45 36 39 43 34 ...
## ..- attr(*, "label")= chr "Age in years"
## ..- attr(*, "format.spss")= chr "F4.0"
## ..- attr(*, "display_width")= int 6
## $ education_level : Factor w/ 5 levels "1","2","3","4",..: 3 1 1 1 1 1 1 1 1 3 ...
## $ current_employ_year : num [1:5000] 17 13 15 15 7 0 1 20 12 7 ...
## ..- attr(*, "label")= chr "Years with current employer"
## ..- attr(*, "format.spss")= chr "F4.0"
## $ current_address_year: num [1:5000] 12 8 14 14 37 13 3 9 11 12 ...
## ..- attr(*, "label")= chr "Years at current address"
## ..- attr(*, "format.spss")= chr "F4.0"
## ..- attr(*, "display_width")= int 9
## $ income_household : num [1:5000] 35.9 46.7 61.8 72 25.6 28.1 19.6 80.5 68.7 33.8 ...
## ..- attr(*, "label")= chr "Household income in thousands"
## ..- attr(*, "format.spss")= chr "F8.2"
## ..- attr(*, "display_width")= int 10
## $ debt_income_ratio : num [1:5000] 11.9 17.9 10.6 29.7 15.9 ...
## ..- attr(*, "label")= chr "Debt to income ratio (x100)"
## ..- attr(*, "format.spss")= chr "F8.2"
## ..- attr(*, "display_width")= int 10
## $ credit_card_debt : num [1:5000] 0.504 1.353 3.439 4.166 1.498 ...
## ..- attr(*, "label")= chr "Credit card debt in thousands"
## ..- attr(*, "format.spss")= chr "F8.2"
## ..- attr(*, "display_width")= int 10
## $ other_debts : num [1:5000] 3.77 7 3.14 17.2 2.56 ...
## ..- attr(*, "label")= chr "Other debt in thousands"
## ..- attr(*, "format.spss")= chr "F8.2"
## ..- attr(*, "display_width")= int 10
## $ defaulted_loan : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 2 1 1 1 ...
## - attr(*, "label")= chr "Bank Loan Default -- Binning"
## - attr(*, "notes")= chr [1:7] "DOCUMENT This is a hypothetical data file that concerns a bank's efforts to redu" " ce" "the rate of loan defaults. This file contains financial and demographic" "information on 5000 past customers that the bank will use to create binning rule" ...
This is a hypothetical data file that concerns a bank’s efforts to reduce the rate of loan defaults. This file contains financial and demographic information on 5000 past customers that the bank will use to create binning rule.
Train Test Validation
Is not this step a universal step in ML?
library(caret)
## Warning: package 'caret' was built under R version 4.1.2
## Loading required package: ggplot2
## Loading required package: lattice
Testing the data into training and testing set
set.seed(1234)
ind = sample(2,nrow(bank_loan_df),replace = T, prob = c(0.8, 0.3))
train_data <- bank_loan_df[ind==1,]
test_data <- bank_loan_df[ind==2,]
Logistic Regression
Training the logictic model
logic_model <- train(defaulted_loan~., data = train_data, method = "glm", family= "binomial")
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
summary(logic_model)
##
## Call:
## NULL
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.5859 -0.6588 -0.3438 0.1138 3.3020
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.169417 0.268244 -4.360 1.30e-05 ***
## age 0.004410 0.008174 0.540 0.5895
## education_level2 0.224490 0.108351 2.072 0.0383 *
## education_level3 0.259264 0.153824 1.685 0.0919 .
## education_level4 0.250029 0.185073 1.351 0.1767
## education_level5 0.018646 0.446741 0.042 0.9667
## current_employ_year -0.182293 0.012469 -14.619 < 2e-16 ***
## current_address_year -0.092239 0.010140 -9.096 < 2e-16 ***
## income_household -0.003279 0.003835 -0.855 0.3925
## debt_income_ratio 0.099422 0.012702 7.827 4.98e-15 ***
## credit_card_debt 0.425010 0.043483 9.774 < 2e-16 ***
## other_debts 0.013697 0.030109 0.455 0.6492
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 4124.2 on 3650 degrees of freedom
## Residual deviance: 2946.7 on 3639 degrees of freedom
## AIC: 2970.7
##
## Number of Fisher Scoring iterations: 6
Testing the Logistic model
pred1 <- predict(logic_model, test_data)
Confusion Matrix
confusionMatrix(pred1, test_data$defaulted_loan)
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 944 177
## 1 70 158
##
## Accuracy : 0.8169
## 95% CI : (0.7952, 0.8372)
## No Information Rate : 0.7517
## P-Value [Acc > NIR] : 6.180e-09
##
## Kappa : 0.4508
##
## Mcnemar's Test P-Value : 1.534e-11
##
## Sensitivity : 0.9310
## Specificity : 0.4716
## Pos Pred Value : 0.8421
## Neg Pred Value : 0.6930
## Prevalence : 0.7517
## Detection Rate : 0.6998
## Detection Prevalence : 0.8310
## Balanced Accuracy : 0.7013
##
## 'Positive' Class : 0
##
KNN Model with train/test validation
knn_model<-train(defaulted_loan~.,data = train_data,
method="knn",
preProcess = c("center", "scale"),
tuneLength = 10
)
Obtain the result
knn_model$result
## k Accuracy Kappa AccuracySD KappaSD
## 1 5 0.7454197 0.2900344 0.011038244 0.02396152
## 2 7 0.7580398 0.3100812 0.011263579 0.02591339
## 3 9 0.7653001 0.3192384 0.012598023 0.02663501
## 4 11 0.7699141 0.3216275 0.012504271 0.02653078
## 5 13 0.7726898 0.3244535 0.013151810 0.03094986
## 6 15 0.7762592 0.3283860 0.012201179 0.02642366
## 7 17 0.7777579 0.3261075 0.012592331 0.02979285
## 8 19 0.7800209 0.3287250 0.010386847 0.02551861
## 9 21 0.7818723 0.3300842 0.009801227 0.02492881
## 10 23 0.7831656 0.3305184 0.009523488 0.02541875
Testing the model
pred2 <- predict(knn_model, test_data)
Confusion Matrix
confusionMatrix(pred2,test_data$defaulted_loan)
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 949 218
## 1 65 117
##
## Accuracy : 0.7902
## 95% CI : (0.7675, 0.8117)
## No Information Rate : 0.7517
## P-Value [Acc > NIR] : 0.0004827
##
## Kappa : 0.3366
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Sensitivity : 0.9359
## Specificity : 0.3493
## Pos Pred Value : 0.8132
## Neg Pred Value : 0.6429
## Prevalence : 0.7517
## Detection Rate : 0.7035
## Detection Prevalence : 0.8651
## Balanced Accuracy : 0.6426
##
## 'Positive' Class : 0
##
Fitting Naive Bayes
Training the model
library(e1071)
naive_model <- naiveBayes(defaulted_loan~., train_data)
summary(naive_model)
## Length Class Mode
## apriori 2 table numeric
## tables 8 -none- list
## levels 2 -none- character
## isnumeric 8 -none- logical
## call 4 -none- call
Testing the model
pred3 <- predict(naive_model, test_data)
Confusion Matrix
confusionMatrix(pred3,test_data$defaulted_loan)
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 974 260
## 1 40 75
##
## Accuracy : 0.7776
## 95% CI : (0.7545, 0.7996)
## No Information Rate : 0.7517
## P-Value [Acc > NIR] : 0.01408
##
## Kappa : 0.2364
##
## Mcnemar's Test P-Value : < 2e-16
##
## Sensitivity : 0.9606
## Specificity : 0.2239
## Pos Pred Value : 0.7893
## Neg Pred Value : 0.6522
## Prevalence : 0.7517
## Detection Rate : 0.7220
## Detection Prevalence : 0.9148
## Balanced Accuracy : 0.5922
##
## 'Positive' Class : 0
##
Support Vector Machine (SVM) Model
Training the model
svm_model <- svm(formula= defaulted_loan~., data = train_data, type = "C-classification", kernel= "linear")
summary(svm_model)
##
## Call:
## svm(formula = defaulted_loan ~ ., data = train_data, type = "C-classification",
## kernel = "linear")
##
##
## Parameters:
## SVM-Type: C-classification
## SVM-Kernel: linear
## cost: 1
##
## Number of Support Vectors: 1614
##
## ( 810 804 )
##
##
## Number of Classes: 2
##
## Levels:
## 0 1
Testing the Model
pred4 <- predict(svm_model, test_data)
Confusion Matrix
confusionMatrix(pred4, test_data$defaulted_loan)
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 960 200
## 1 54 135
##
## Accuracy : 0.8117
## 95% CI : (0.7898, 0.8322)
## No Information Rate : 0.7517
## P-Value [Acc > NIR] : 8.805e-08
##
## Kappa : 0.4095
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Sensitivity : 0.9467
## Specificity : 0.4030
## Pos Pred Value : 0.8276
## Neg Pred Value : 0.7143
## Prevalence : 0.7517
## Detection Rate : 0.7116
## Detection Prevalence : 0.8599
## Balanced Accuracy : 0.6749
##
## 'Positive' Class : 0
##
Decision Tree Model
Training the model
decision_tree_model <-train(defaulted_loan~.,
data = train_data,
method="rpart",
parms = list(split = "information"),
tuneLength=10
)
Testing the model
pred5 <- predict(decision_tree_model, test_data)
Confusion Matrix
confusionMatrix(pred5, test_data$defaulted_loan)
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 956 235
## 1 58 100
##
## Accuracy : 0.7828
## 95% CI : (0.7598, 0.8045)
## No Information Rate : 0.7517
## P-Value [Acc > NIR] : 0.004044
##
## Kappa : 0.2932
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Sensitivity : 0.9428
## Specificity : 0.2985
## Pos Pred Value : 0.8027
## Neg Pred Value : 0.6329
## Prevalence : 0.7517
## Detection Rate : 0.7087
## Detection Prevalence : 0.8829
## Balanced Accuracy : 0.6207
##
## 'Positive' Class : 0
##
Artifical Neural Network (ANN) Model
Training the Model
ann_model <- train(defaulted_loan ~ ., data = train_data,
method = "nnet",
preProcess = c("center","scale"),
maxit = 250, # Maximum number of iterations
tuneGrid = data.frame(size = 1, decay = 0),
# tuneGrid = data.frame(size = 0, decay = 0),skip=TRUE, # Technically, this is log-reg
metric = "Accuracy")
## # weights: 14
## initial value 2490.154378
## iter 10 value 1603.415337
## iter 20 value 1512.403963
## iter 30 value 1480.119731
## iter 40 value 1476.081679
## iter 50 value 1469.305229
## iter 60 value 1467.397161
## iter 70 value 1467.376879
## iter 80 value 1467.008906
## iter 90 value 1466.875450
## final value 1466.875090
## converged
## # weights: 14
## initial value 2407.105942
## iter 10 value 1621.907412
## iter 20 value 1558.129266
## iter 30 value 1530.209595
## iter 40 value 1475.296367
## iter 50 value 1455.253543
## iter 60 value 1454.643391
## iter 70 value 1452.315447
## iter 80 value 1452.069503
## iter 90 value 1452.067419
## iter 90 value 1452.067410
## iter 90 value 1452.067409
## final value 1452.067409
## converged
## # weights: 14
## initial value 3316.269093
## iter 10 value 2091.326271
## iter 20 value 1688.424852
## iter 30 value 1627.575732
## iter 40 value 1578.208844
## iter 50 value 1540.242658
## iter 60 value 1536.818068
## iter 70 value 1534.901951
## final value 1534.887159
## converged
## # weights: 14
## initial value 2385.358031
## iter 10 value 1668.690637
## iter 20 value 1583.737340
## iter 30 value 1544.955813
## iter 40 value 1473.578642
## iter 50 value 1451.624024
## iter 60 value 1431.805159
## iter 70 value 1429.709747
## iter 80 value 1427.802753
## iter 90 value 1426.693990
## iter 100 value 1426.660252
## final value 1426.659975
## converged
## # weights: 14
## initial value 3201.590239
## iter 10 value 1735.394117
## iter 20 value 1639.370673
## iter 30 value 1557.667764
## iter 40 value 1520.658633
## iter 50 value 1505.406878
## iter 60 value 1500.146505
## iter 70 value 1500.001251
## iter 80 value 1498.576661
## iter 90 value 1498.031963
## iter 100 value 1498.018686
## final value 1498.017137
## converged
## # weights: 14
## initial value 2699.729482
## iter 10 value 1708.185902
## iter 20 value 1533.672892
## iter 30 value 1485.389639
## iter 40 value 1472.897282
## iter 50 value 1463.951011
## iter 60 value 1461.864517
## iter 70 value 1461.852301
## iter 80 value 1461.295186
## iter 90 value 1461.191318
## iter 100 value 1461.140191
## iter 110 value 1461.060769
## iter 120 value 1461.027597
## iter 120 value 1461.027593
## iter 120 value 1461.027591
## final value 1461.027591
## converged
## # weights: 14
## initial value 2918.658527
## iter 10 value 1747.555356
## iter 20 value 1578.274428
## iter 30 value 1541.761037
## iter 40 value 1512.216219
## iter 50 value 1481.010030
## iter 60 value 1476.352511
## iter 70 value 1476.295136
## iter 80 value 1475.663670
## iter 90 value 1475.556116
## iter 100 value 1475.553767
## iter 110 value 1475.534585
## final value 1475.534359
## converged
## # weights: 14
## initial value 2414.595151
## iter 10 value 1568.757814
## iter 20 value 1502.778270
## iter 30 value 1472.193138
## iter 40 value 1466.293800
## iter 50 value 1462.895500
## iter 60 value 1461.939982
## iter 70 value 1461.885035
## iter 80 value 1461.840147
## final value 1461.834804
## converged
## # weights: 14
## initial value 2930.006588
## iter 10 value 1757.770588
## iter 20 value 1630.089973
## iter 30 value 1571.955426
## iter 40 value 1530.350816
## iter 50 value 1510.110349
## iter 60 value 1504.706374
## iter 70 value 1504.643534
## iter 80 value 1503.452073
## iter 90 value 1503.105631
## iter 100 value 1503.094003
## final value 1503.093402
## converged
## # weights: 14
## initial value 2189.427854
## iter 10 value 1676.144100
## iter 20 value 1599.071383
## iter 30 value 1551.234187
## iter 40 value 1539.680692
## iter 50 value 1536.133313
## final value 1536.132575
## converged
## # weights: 14
## initial value 3830.386414
## iter 10 value 1732.091410
## iter 20 value 1554.539280
## iter 30 value 1476.587233
## iter 40 value 1452.257887
## iter 50 value 1440.977816
## iter 60 value 1436.619308
## iter 70 value 1436.542765
## iter 80 value 1435.267041
## iter 90 value 1434.971108
## final value 1434.964372
## converged
## # weights: 14
## initial value 2393.249806
## iter 10 value 1606.742894
## iter 20 value 1494.230041
## iter 30 value 1470.391210
## iter 40 value 1469.559992
## iter 50 value 1466.959732
## final value 1466.902721
## converged
## # weights: 14
## initial value 2303.745081
## iter 10 value 1704.168147
## iter 20 value 1529.587089
## iter 30 value 1492.297275
## iter 40 value 1464.104086
## iter 50 value 1449.604997
## iter 60 value 1446.084781
## iter 70 value 1445.953737
## iter 80 value 1444.803218
## iter 90 value 1444.548183
## iter 100 value 1444.539852
## iter 110 value 1444.454647
## iter 120 value 1444.440015
## iter 120 value 1444.440011
## final value 1444.439950
## converged
## # weights: 14
## initial value 2803.286864
## iter 10 value 1732.489256
## iter 20 value 1533.201667
## iter 30 value 1447.366296
## iter 40 value 1426.178074
## iter 50 value 1414.650132
## iter 60 value 1410.829602
## iter 70 value 1410.578650
## iter 80 value 1410.384875
## iter 90 value 1410.172837
## iter 100 value 1410.166008
## iter 100 value 1410.166003
## iter 100 value 1410.165995
## final value 1410.165995
## converged
## # weights: 14
## initial value 2211.136287
## iter 10 value 1551.419973
## iter 20 value 1472.810866
## iter 30 value 1444.043116
## iter 40 value 1442.292392
## iter 50 value 1440.049935
## iter 60 value 1439.845773
## final value 1439.845706
## converged
## # weights: 14
## initial value 2758.551553
## iter 10 value 1682.966532
## iter 20 value 1552.372647
## iter 30 value 1487.100391
## iter 40 value 1472.196531
## iter 50 value 1459.148734
## iter 60 value 1456.132849
## iter 70 value 1456.087110
## iter 80 value 1455.471437
## iter 90 value 1455.289969
## iter 100 value 1455.237340
## iter 110 value 1455.134117
## iter 120 value 1455.109959
## iter 120 value 1455.109954
## final value 1455.109915
## converged
## # weights: 14
## initial value 3230.762626
## iter 10 value 2014.197184
## iter 20 value 1616.238525
## iter 30 value 1560.032833
## iter 40 value 1545.733807
## iter 50 value 1492.808071
## iter 60 value 1460.831971
## iter 70 value 1455.416018
## iter 80 value 1448.128086
## iter 90 value 1446.111942
## iter 100 value 1446.097170
## iter 110 value 1445.718095
## iter 120 value 1445.018042
## iter 130 value 1444.847994
## final value 1444.846358
## converged
## # weights: 14
## initial value 2086.395138
## iter 10 value 1651.981162
## iter 20 value 1570.275698
## iter 30 value 1552.487729
## iter 40 value 1533.896054
## iter 50 value 1486.347284
## iter 60 value 1455.570803
## iter 70 value 1452.722282
## iter 80 value 1445.433809
## iter 90 value 1444.580195
## iter 100 value 1444.564862
## iter 110 value 1444.157336
## iter 120 value 1444.042685
## iter 130 value 1444.041039
## final value 1444.040972
## converged
## # weights: 14
## initial value 2167.978779
## iter 10 value 1567.053008
## iter 20 value 1482.319887
## iter 30 value 1478.720905
## iter 40 value 1477.910163
## iter 50 value 1473.937284
## iter 60 value 1472.932923
## iter 70 value 1472.908336
## iter 80 value 1472.501261
## iter 90 value 1472.351506
## final value 1472.350470
## converged
## # weights: 14
## initial value 3352.063487
## iter 10 value 1611.897544
## iter 20 value 1542.164826
## iter 30 value 1480.731131
## iter 40 value 1464.512410
## iter 50 value 1460.819301
## iter 60 value 1456.108430
## iter 70 value 1455.605860
## iter 80 value 1455.521161
## iter 90 value 1455.182616
## iter 100 value 1455.097729
## final value 1455.097357
## converged
## # weights: 14
## initial value 2342.129872
## iter 10 value 1714.695204
## iter 20 value 1451.318364
## iter 30 value 1442.729608
## iter 40 value 1442.590767
## iter 50 value 1441.556518
## iter 60 value 1441.294491
## iter 70 value 1441.291422
## iter 80 value 1441.201001
## iter 90 value 1441.165284
## iter 90 value 1441.165272
## final value 1441.165244
## converged
## # weights: 14
## initial value 2593.837238
## iter 10 value 1863.834407
## iter 20 value 1683.293593
## iter 30 value 1596.141937
## iter 40 value 1527.145467
## iter 50 value 1497.944848
## iter 60 value 1489.105097
## iter 70 value 1479.588621
## iter 80 value 1478.815466
## iter 90 value 1478.588264
## iter 100 value 1478.054665
## iter 110 value 1478.009778
## iter 120 value 1477.910910
## iter 130 value 1477.835726
## iter 140 value 1477.822115
## final value 1477.822076
## converged
## # weights: 14
## initial value 2151.233944
## iter 10 value 1521.277087
## iter 20 value 1500.960138
## iter 30 value 1473.863542
## iter 40 value 1472.983454
## iter 50 value 1471.799061
## iter 60 value 1470.562451
## iter 70 value 1470.496968
## iter 80 value 1470.166746
## iter 90 value 1469.858236
## iter 100 value 1469.814168
## final value 1469.811825
## converged
## # weights: 14
## initial value 2377.280407
## iter 10 value 1722.209249
## iter 20 value 1585.170006
## iter 30 value 1532.813774
## iter 40 value 1519.437463
## iter 50 value 1509.323697
## iter 60 value 1507.368192
## iter 70 value 1507.289372
## iter 80 value 1507.064914
## final value 1507.064734
## converged
## # weights: 14
## initial value 2075.581552
## iter 10 value 1611.779154
## iter 20 value 1553.537167
## iter 30 value 1501.139585
## iter 40 value 1496.648956
## iter 50 value 1486.248966
## iter 60 value 1484.398065
## iter 70 value 1484.371374
## iter 80 value 1483.839765
## iter 90 value 1483.735825
## iter 100 value 1483.698215
## iter 110 value 1483.538856
## iter 120 value 1483.455589
## iter 120 value 1483.455576
## final value 1483.455426
## converged
## # weights: 14
## initial value 2235.760938
## iter 10 value 1605.952447
## iter 20 value 1554.487605
## iter 30 value 1491.921467
## iter 40 value 1482.398128
## iter 50 value 1475.444177
## iter 60 value 1473.703199
## iter 70 value 1473.649950
## iter 80 value 1473.299082
## iter 90 value 1473.176866
## final value 1473.176608
## converged
Testing the model
pred6 <- predict(ann_model, test_data)
Confusion Matrix
confusionMatrix(pred6, test_data$defaulted_loan)
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 941 173
## 1 73 162
##
## Accuracy : 0.8176
## 95% CI : (0.796, 0.8379)
## No Information Rate : 0.7517
## P-Value [Acc > NIR] : 4.148e-09
##
## Kappa : 0.4573
##
## Mcnemar's Test P-Value : 2.754e-10
##
## Sensitivity : 0.9280
## Specificity : 0.4836
## Pos Pred Value : 0.8447
## Neg Pred Value : 0.6894
## Prevalence : 0.7517
## Detection Rate : 0.6976
## Detection Prevalence : 0.8258
## Balanced Accuracy : 0.7058
##
## 'Positive' Class : 0
##
Leave one Out Validation
Read the data
library(haven)
bank_loan_df <- read_sav("P4_bankloan_5000_clients.sav")
bank_loan_df$defaulted_loan<-as.factor(bank_loan_df$defaulted_loan)
bank_loan_df$education_level<-as.factor(bank_loan_df$education_level)
Logistic Regression With LOOCV Validation
Training Logistic Regression Model
set.seed(1234)
library(caret)
ind<-sample(2,nrow(bank_loan_df),replace=T,prob = c(0.7,0.3))
train_data<-bank_loan_df[ind==1,]
test_data<-bank_loan_df[ind==2,]
Setting Up the Train Control
loocv_train_control<-trainControl(method = "LOOCV")
Logistic Regression With LOOCV Validation
Training Logistic Regression Model
#logistic_clf1<-train(defaulted_loan~.,
#data=train_data,
#method="glm",
#family="binomial",
#trControl=loocv_train_control,
#verbose=F
# )
KNN Model with LOOCV validation
Training KNN Model
knn_clf1<-train(defaulted_loan~.,data = train_data,
method="knn",
trControl=loocv_train_control
)
Obtain the result
knn_clf1$result
## k Accuracy Kappa
## 1 5 0.7636879 0.3087625
## 2 7 0.7707801 0.3112221
## 3 9 0.7770213 0.3248772
Confusion Matrix for Model Evaluation
predicted_val_knn1<-predict(knn_clf1,newdata = test_data)
confusionMatrix(predicted_val_knn1,test_data$defaulted_loan)
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 1018 226
## 1 96 135
##
## Accuracy : 0.7817
## 95% CI : (0.7597, 0.8025)
## No Information Rate : 0.7553
## P-Value [Acc > NIR] : 0.009238
##
## Kappa : 0.3277
##
## Mcnemar's Test P-Value : 6.532e-13
##
## Sensitivity : 0.9138
## Specificity : 0.3740
## Pos Pred Value : 0.8183
## Neg Pred Value : 0.5844
## Prevalence : 0.7553
## Detection Rate : 0.6902
## Detection Prevalence : 0.8434
## Balanced Accuracy : 0.6439
##
## 'Positive' Class : 0
##
Naïve Bayes classifier
Training the Model
nb_clf1<-train(defaulted_loan~.,
data=train_data,
method="naive_bayes",
usepoisson = TRUE,
trControl=loocv_train_control
)
Making Prediction on Test Data
predicted_val_nb1<-predict(nb_clf1,newdata = test_data)
Confusion Matrix for Model Evaluation
confusionMatrix(predicted_val_nb1,test_data$defaulted_loan)
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 1094 308
## 1 20 53
##
## Accuracy : 0.7776
## 95% CI : (0.7555, 0.7986)
## No Information Rate : 0.7553
## P-Value [Acc > NIR] : 0.02363
##
## Kappa : 0.1764
##
## Mcnemar's Test P-Value : < 2e-16
##
## Sensitivity : 0.9820
## Specificity : 0.1468
## Pos Pred Value : 0.7803
## Neg Pred Value : 0.7260
## Prevalence : 0.7553
## Detection Rate : 0.7417
## Detection Prevalence : 0.9505
## Balanced Accuracy : 0.5644
##
## 'Positive' Class : 0
##
K-Fold Cross Validation
Reading the File
library(haven)
bank_loan_df <- read_sav("P4_bankloan_5000_clients.sav")
Changing the data type of variables
bank_loan_df$defaulted_loan<-as.factor(bank_loan_df$defaulted_loan)
bank_loan_df$education_level<-as.factor(bank_loan_df$education_level)
Splitting the data into train and test set
set.seed(1234)
library(caret)
ind<-sample(2,nrow(bank_loan_df),replace=T,prob = c(0.7,0.3))
train_data<-bank_loan_df[ind==1,]
test_data<-bank_loan_df[ind==2,]
Setting Up the Train Control
cv_train_control<-trainControl(method = "cv",number = 10)
Logistic Regression With Cross Validation
Training Logistic Regression Model
logistic_clf1<-train(defaulted_loan~.,
data=train_data,
method="glm",
family="binomial",
trControl=cv_train_control
)
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
summary(logistic_clf1)
##
## Call:
## NULL
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.6490 -0.6635 -0.3442 0.1409 3.2833
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.235986 0.272446 -4.537 5.72e-06 ***
## age 0.006492 0.008297 0.782 0.4339
## education_level2 0.227329 0.110244 2.062 0.0392 *
## education_level3 0.260781 0.156468 1.667 0.0956 .
## education_level4 0.285038 0.186776 1.526 0.1270
## education_level5 0.020994 0.447370 0.047 0.9626
## current_employ_year -0.182777 0.012678 -14.416 < 2e-16 ***
## current_address_year -0.094317 0.010300 -9.157 < 2e-16 ***
## income_household -0.002470 0.003879 -0.637 0.5244
## debt_income_ratio 0.099652 0.012885 7.734 1.04e-14 ***
## credit_card_debt 0.425066 0.044558 9.540 < 2e-16 ***
## other_debts 0.006704 0.030495 0.220 0.8260
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 3994.4 on 3524 degrees of freedom
## Residual deviance: 2850.2 on 3513 degrees of freedom
## AIC: 2874.2
##
## Number of Fisher Scoring iterations: 6
Making the Prediction
predicted_val_log1<-predict(logistic_clf1,newdata = test_data)
Confusion Matrix for Evaluation
confusionMatrix(predicted_val_log1,test_data$defaulted_loan)
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 1038 191
## 1 76 170
##
## Accuracy : 0.819
## 95% CI : (0.7984, 0.8383)
## No Information Rate : 0.7553
## P-Value [Acc > NIR] : 2.487e-09
##
## Kappa : 0.4513
##
## Mcnemar's Test P-Value : 3.022e-12
##
## Sensitivity : 0.9318
## Specificity : 0.4709
## Pos Pred Value : 0.8446
## Neg Pred Value : 0.6911
## Prevalence : 0.7553
## Detection Rate : 0.7037
## Detection Prevalence : 0.8332
## Balanced Accuracy : 0.7013
##
## 'Positive' Class : 0
##
KNN Model with Cross validation
Training KNN Model
knn_clf1<-train(defaulted_loan~.,data = train_data,
method="knn",
trControl=cv_train_control
)
Getting the Result of the Model
knn_clf1$result
## k Accuracy Kappa AccuracySD KappaSD
## 1 5 0.7668056 0.3138335 0.01709039 0.03827953
## 2 7 0.7727611 0.3210782 0.01581367 0.03762291
## 3 9 0.7744568 0.3184971 0.01934934 0.05507607
Confusion Matrix for Model Evaluation
predicted_val_knn1<-predict(knn_clf1,newdata = test_data)
confusionMatrix(predicted_val_knn1,test_data$defaulted_loan)
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 1019 226
## 1 95 135
##
## Accuracy : 0.7824
## 95% CI : (0.7604, 0.8032)
## No Information Rate : 0.7553
## P-Value [Acc > NIR] : 0.007801
##
## Kappa : 0.329
##
## Mcnemar's Test P-Value : 3.99e-13
##
## Sensitivity : 0.9147
## Specificity : 0.3740
## Pos Pred Value : 0.8185
## Neg Pred Value : 0.5870
## Prevalence : 0.7553
## Detection Rate : 0.6908
## Detection Prevalence : 0.8441
## Balanced Accuracy : 0.6443
##
## 'Positive' Class : 0
##
Naïve Bayes classifier
Training the Model
library(naivebayes)
## Warning: package 'naivebayes' was built under R version 4.1.2
## naivebayes 0.9.7 loaded
nb_clf1<-train(defaulted_loan~.,
data=train_data,
method="naive_bayes",
usepoisson = TRUE,
trControl=cv_train_control
)
summary(nb_clf1)
##
## ================================== Naive Bayes ==================================
##
## - Call: naive_bayes.default(x = x, y = y, laplace = param$laplace, usekernel = TRUE, usepoisson = TRUE, adjust = param$adjust)
## - Laplace: 0
## - Classes: 2
## - Samples: 3525
## - Features: 11
## - Conditional distributions:
## - KDE: 11
## - Prior probabilities:
## - 0: 0.7461
## - 1: 0.2539
##
## ---------------------------------------------------------------------------------
Making Prediction on Test Data
predicted_val_nb1<-predict(nb_clf1,newdata = test_data)
Confusion Matrix for Model Evaluation
confusionMatrix(predicted_val_nb1,test_data$defaulted_loan)
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 1094 308
## 1 20 53
##
## Accuracy : 0.7776
## 95% CI : (0.7555, 0.7986)
## No Information Rate : 0.7553
## P-Value [Acc > NIR] : 0.02363
##
## Kappa : 0.1764
##
## Mcnemar's Test P-Value : < 2e-16
##
## Sensitivity : 0.9820
## Specificity : 0.1468
## Pos Pred Value : 0.7803
## Neg Pred Value : 0.7260
## Prevalence : 0.7553
## Detection Rate : 0.7417
## Detection Prevalence : 0.9505
## Balanced Accuracy : 0.5644
##
## 'Positive' Class : 0
##
Bagging, Boosting and Random Forest
Reading the File
library(haven)
bank_loan_df <- read_sav("P4_bankloan_5000_clients.sav")
bank_loan_df$defaulted_loan<-as.factor(bank_loan_df$defaulted_loan)
bank_loan_df$education_level<-as.factor(bank_loan_df$education_level)
Splitting the data into train and test set
set.seed(1234)
library(caret)
## Loading required package: ggplot2
## Loading required package: lattice
ind<-sample(2,nrow(bank_loan_df),replace=T,prob = c(0.7,0.3))
train_data<-bank_loan_df[ind==1,]
test_data<-bank_loan_df[ind==2,]
Bagging Model
Training the Model
library("ipred")
## Warning: package 'ipred' was built under R version 4.1.2
bag_dtree_clf<-bagging(defaulted_loan~.,
data = train_data,
coob=T
)
print(bag_dtree_clf)
##
## Bagging classification trees with 25 bootstrap replications
##
## Call: bagging.data.frame(formula = defaulted_loan ~ ., data = train_data,
## coob = T)
##
## Out-of-bag estimate of misclassification error: 0.2295
Making the Prediction
predicted_bag_tree<-predict(bag_dtree_clf,newdata = test_data)
library(caret)
confusionMatrix(predicted_bag_tree,test_data$defaulted_loan)
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 991 191
## 1 123 170
##
## Accuracy : 0.7871
## 95% CI : (0.7653, 0.8078)
## No Information Rate : 0.7553
## P-Value [Acc > NIR] : 0.0021549
##
## Kappa : 0.385
##
## Mcnemar's Test P-Value : 0.0001562
##
## Sensitivity : 0.8896
## Specificity : 0.4709
## Pos Pred Value : 0.8384
## Neg Pred Value : 0.5802
## Prevalence : 0.7553
## Detection Rate : 0.6719
## Detection Prevalence : 0.8014
## Balanced Accuracy : 0.6803
##
## 'Positive' Class : 0
##
Random Forest Model
Training the Model
set.seed(1234)
library(randomForest)
## Warning: package 'randomForest' was built under R version 4.1.2
## randomForest 4.6-14
## Type rfNews() to see new features/changes/bug fixes.
##
## Attaching package: 'randomForest'
## The following object is masked from 'package:ggplot2':
##
## margin
rf_clf<-randomForest(defaulted_loan~.,
data = train_data)
rf_clf
##
## Call:
## randomForest(formula = defaulted_loan ~ ., data = train_data)
## Type of random forest: classification
## Number of trees: 500
## No. of variables tried at each split: 2
##
## OOB estimate of error rate: 20.88%
## Confusion matrix:
## 0 1 class.error
## 0 2420 210 0.07984791
## 1 526 369 0.58770950
Making the Prediction
predicted_rf<-predict(rf_clf,newdata = test_data)
confusionMatrix(predicted_rf,test_data$defaulted_loan)
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 1023 197
## 1 91 164
##
## Accuracy : 0.8047
## 95% CI : (0.7836, 0.8247)
## No Information Rate : 0.7553
## P-Value [Acc > NIR] : 3.459e-06
##
## Kappa : 0.4137
##
## Mcnemar's Test P-Value : 6.125e-10
##
## Sensitivity : 0.9183
## Specificity : 0.4543
## Pos Pred Value : 0.8385
## Neg Pred Value : 0.6431
## Prevalence : 0.7553
## Detection Rate : 0.6936
## Detection Prevalence : 0.8271
## Balanced Accuracy : 0.6863
##
## 'Positive' Class : 0
##
Extreme Gradient Boosting
Training the Model
#xglm_clf<-train(defaulted_loan~.,
#data = train_data,
#method="xgbTree",
#verbose=F
# )
Making the Prediction
#predicted_xgb<-predict(xglm_clf,newdata = test_data)
#confusionMatrix(predicted_xgb,test_data$defaulted_loan)
Comments