# Cargar la biblioteca necesaria
library(readxl)
library(dplyr)##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionlibrary(rpart)
library(randomForest)## randomForest 4.7-1.1## Type rfNews() to see new features/changes/bug fixes.##
## Attaching package: 'randomForest'## The following object is masked from 'package:dplyr':
##
## combine# Especifica la ubicación del archivo
ruta_archivo <- "C:\\Users\\sguerra\\Downloads\\credit_shortclean.xlsx"
# Cargar los datos desde el archivo Excel
fondo <- read_excel(ruta_archivo)
# Revisar la estructura de la base de datos
str(fondo)## tibble [957 × 20] (S3: tbl_df/tbl/data.frame)
## $ loan_status : chr [1:957] "Current" "Charged Off" "Current" "Fully Paid" ...
## $ loan_amnt : num [1:957] 15000 4575 15000 6000 6000 ...
## $ term : chr [1:957] "60 months" "36 months" "60 months" "36 months" ...
## $ int_rate : num [1:957] 11.48 17.27 11.48 13.99 5.32 ...
## $ installment : num [1:957] 330 164 330 205 181 ...
## $ grade : chr [1:957] "B" "D" "B" "C" ...
## $ sub_grade : chr [1:957] "B5" "D3" "B5" "C4" ...
## $ emp_title : chr [1:957] "2nd Pressman" "2nd pressman" "A/B- merchant marine" "A/C Technician" ...
## $ emp_length : chr [1:957] "2 years" "5 years" "7 years" "6 years" ...
## $ home_ownership : chr [1:957] "RENT" "RENT" "OWN" "MORTGAGE" ...
## $ annual_inc : num [1:957] 57000 56000 65000 43700 80000 45000 125000 120000 95000 90000 ...
## $ verification_status: chr [1:957] "Verified" "Not Verified" "Not Verified" "Verified" ...
## $ purpose : chr [1:957] "credit_card" "house" "debt_consolidation" "home_improvement" ...
## $ title : chr [1:957] "Credit card refinancing" "Home buying" "Debt consolidation" "Home improvement" ...
## $ dti : num [1:957] 20.67 10.03 6.63 31.87 13.5 ...
## $ earliest_cr_line : chr [1:957] "Oct-2000" "May-2001" "Dec-2003" "Feb-2005" ...
## $ open_acc : num [1:957] 10 12 17 20 20 16 15 17 14 17 ...
## $ total_acc : num [1:957] 20 17 29 33 39 23 26 26 24 23 ...
## $ initial_list_status: chr [1:957] "w" "w" "w" "f" ...
## $ application_type : chr [1:957] "Individual" "Individual" "Individual" "Individual" ...# Reemplazar los valores faltantes por la media
fondo <- na.omit(fondo)
# Supongamos que "Fully Paid" y "Current" son considerados "Aprobados".
fondo$Clasificación <- ifelse(fondo$loan_status %in% c("Fully Paid", "Current"), "Aprobada", "No Aprobada")
# Estadísticas descriptivas
aprobadas <- subset(fondo, Clasificación == "Aprobada")
no_aprobadas <- subset(fondo, Clasificación == "No Aprobada")
summary(aprobadas)## loan_status loan_amnt term int_rate
## Length:779 Min. : 1000 Length:779 Min. : 5.32
## Class :character 1st Qu.:10000 Class :character 1st Qu.: 8.49
## Mode :character Median :15654 Mode :character Median :11.48
## Mean :15904 Mean :11.58
## 3rd Qu.:20863 3rd Qu.:13.99
## Max. :35000 Max. :27.99
## installment grade sub_grade emp_title
## Min. : 32.97 Length:779 Length:779 Length:779
## 1st Qu.: 263.77 Class :character Class :character Class :character
## Median : 397.52 Mode :character Mode :character Mode :character
## Mean : 456.92
## 3rd Qu.: 617.46
## Max. :1252.56
## emp_length home_ownership annual_inc verification_status
## Length:779 Length:779 Min. : 13000 Length:779
## Class :character Class :character 1st Qu.: 52800 Class :character
## Mode :character Mode :character Median : 75000 Mode :character
## Mean : 83147
## 3rd Qu.:104250
## Max. :450000
## purpose title dti earliest_cr_line
## Length:779 Length:779 Min. : 0.63 Length:779
## Class :character Class :character 1st Qu.:13.29 Class :character
## Mode :character Mode :character Median :18.42 Mode :character
## Mean :18.94
## 3rd Qu.:24.77
## Max. :39.87
## open_acc total_acc initial_list_status application_type
## Min. : 2.00 Min. : 4.00 Length:779 Length:779
## 1st Qu.: 9.00 1st Qu.:18.00 Class :character Class :character
## Median :12.00 Median :24.00 Mode :character Mode :character
## Mean :12.33 Mean :26.41
## 3rd Qu.:15.00 3rd Qu.:33.00
## Max. :35.00 Max. :87.00
## Clasificación
## Length:779
## Class :character
## Mode :character
##
##
## summary(no_aprobadas)## loan_status loan_amnt term int_rate
## Length:144 Min. : 1000 Length:144 Min. : 5.32
## Class :character 1st Qu.:10300 Class :character 1st Qu.:11.48
## Mode :character Median :15150 Mode :character Median :13.99
## Mean :14959 Mean :14.47
## 3rd Qu.:18938 3rd Qu.:17.27
## Max. :35000 Max. :28.49
## installment grade sub_grade emp_title
## Min. : 34.18 Length:144 Length:144 Length:144
## 1st Qu.: 283.17 Class :character Class :character Class :character
## Median : 403.29 Mode :character Mode :character Mode :character
## Mean : 434.45
## 3rd Qu.: 550.25
## Max. :1009.09
## emp_length home_ownership annual_inc verification_status
## Length:144 Length:144 Min. : 18300 Length:144
## Class :character Class :character 1st Qu.: 47000 Class :character
## Mode :character Mode :character Median : 65000 Mode :character
## Mean : 72835
## 3rd Qu.: 92000
## Max. :230000
## purpose title dti earliest_cr_line
## Length:144 Length:144 Min. : 2.96 Length:144
## Class :character Class :character 1st Qu.:15.27 Class :character
## Mode :character Mode :character Median :21.68 Mode :character
## Mean :21.59
## 3rd Qu.:27.46
## Max. :46.71
## open_acc total_acc initial_list_status application_type
## Min. : 5.00 Min. : 6.00 Length:144 Length:144
## 1st Qu.:10.00 1st Qu.:18.00 Class :character Class :character
## Median :12.00 Median :25.00 Mode :character Mode :character
## Mean :13.66 Mean :27.12
## 3rd Qu.:17.00 3rd Qu.:32.25
## Max. :46.00 Max. :89.00
## Clasificación
## Length:144
## Class :character
## Mode :character
##
##
## # Recodificar "Clasificación" como binaria (0 = No Aprobada, 1 = Aprobada)
fondo$Clasificación <- ifelse(fondo$Clasificación == "Aprobada", 1, 0)
# Modelo de regresión logística
modelo <- glm(Clasificación ~ loan_amnt + int_rate + annual_inc + total_acc, data = fondo, family = binomial(link = "logit"))
# Resumen del modelo
summary(modelo)##
## Call:
## glm(formula = Clasificación ~ loan_amnt + int_rate + annual_inc +
## total_acc, family = binomial(link = "logit"), data = fondo)
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 3.167e+00 3.941e-01 8.035 9.34e-16 ***
## loan_amnt 2.308e-05 1.265e-05 1.824 0.0681 .
## int_rate -1.542e-01 2.200e-02 -7.007 2.43e-12 ***
## annual_inc 5.010e-06 2.847e-06 1.760 0.0784 .
## total_acc -8.363e-03 7.617e-03 -1.098 0.2722
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 799.32 on 922 degrees of freedom
## Residual deviance: 737.02 on 918 degrees of freedom
## AIC: 747.02
##
## Number of Fisher Scoring iterations: 5# Dividir los datos en conjuntos de entrenamiento y prueba
set.seed(123) # Establecer una semilla para la reproducibilidad
indice_entrenamiento <- sample(1:nrow(fondo), 0.7 * nrow(fondo))
conjunto_entrenamiento <- fondo[indice_entrenamiento, ]
conjunto_prueba <- fondo[-indice_entrenamiento, ]
# Entrenar el modelo en el conjunto de entrenamiento
modelo_entrenado <- glm(Clasificación ~ loan_amnt + int_rate + annual_inc + total_acc, data = conjunto_entrenamiento, family = binomial(link = "logit"))
# Calibrar el modelo (ajustar los hiperparámetros) si es necesario
# Evaluar el rendimiento del modelo en el conjunto de prueba
predicciones <- predict(modelo_entrenado, newdata = conjunto_prueba, type = "response")
print(predicciones)## 1 2 3 4 5 6 7 8
## 0.8728595 0.8736990 0.7666572 0.8729209 0.8555387 0.7642799 0.7897494 0.8996526
## 9 10 11 12 13 14 15 16
## 0.5088003 0.7721069 0.9110904 0.8713025 0.8154349 0.7707559 0.8158501 0.9544678
## 17 18 19 20 21 22 23 24
## 0.9439561 0.7363522 0.9353888 0.8053333 0.9398285 0.8799401 0.8850186 0.9399287
## 25 26 27 28 29 30 31 32
## 0.5904855 0.8963446 0.8655846 0.8374908 0.9745529 0.7889541 0.9010965 0.8959160
## 33 34 35 36 37 38 39 40
## 0.9167390 0.8845944 0.8124759 0.9770202 0.9800578 0.7050073 0.9179328 0.7478433
## 41 42 43 44 45 46 47 48
## 0.9458151 0.7633633 0.8387085 0.9422930 0.9268202 0.7250763 0.8996102 0.8823639
## 49 50 51 52 53 54 55 56
## 0.7420187 0.7222179 0.8959027 0.9234964 0.9503108 0.9273628 0.8538169 0.8726600
## 57 58 59 60 61 62 63 64
## 0.8339217 0.7907589 0.8430041 0.9677814 0.8048274 0.5915208 0.8392448 0.8873392
## 65 66 67 68 69 70 71 72
## 0.9705960 0.9271537 0.9354912 0.9516411 0.7156451 0.8221838 0.8935333 0.9204717
## 73 74 75 76 77 78 79 80
## 0.9223263 0.8814627 0.7634223 0.8130685 0.6461857 0.9541239 0.8118591 0.9451993
## 81 82 83 84 85 86 87 88
## 0.5375072 0.8730161 0.9549749 0.8541533 0.6247951 0.8967090 0.8853657 0.8633511
## 89 90 91 92 93 94 95 96
## 0.7626375 0.9251620 0.7001810 0.7976293 0.6717222 0.8250717 0.8514157 0.7125103
## 97 98 99 100 101 102 103 104
## 0.9544328 0.9514432 0.6984041 0.9545759 0.9160361 0.9419977 0.9369184 0.7807952
## 105 106 107 108 109 110 111 112
## 0.8714040 0.9274721 0.9569706 0.7492705 0.7801598 0.7230081 0.8309640 0.9247520
## 113 114 115 116 117 118 119 120
## 0.9597383 0.8974746 0.8660418 0.8693998 0.8832033 0.6749534 0.9240798 0.8312964
## 121 122 123 124 125 126 127 128
## 0.8243956 0.7266255 0.8846924 0.8142556 0.9575691 0.8548350 0.9322678 0.9161221
## 129 130 131 132 133 134 135 136
## 0.9073096 0.9351553 0.9492650 0.9289781 0.6353300 0.9135732 0.7410398 0.8082215
## 137 138 139 140 141 142 143 144
## 0.8060336 0.7960229 0.8334201 0.9199717 0.7899506 0.8346327 0.9580142 0.7942083
## 145 146 147 148 149 150 151 152
## 0.7865218 0.9392626 0.9505827 0.9030499 0.9300540 0.7191380 0.9590716 0.5944000
## 153 154 155 156 157 158 159 160
## 0.8962763 0.8998793 0.8376386 0.9573153 0.8782844 0.8512961 0.6497831 0.9280586
## 161 162 163 164 165 166 167 168
## 0.2605232 0.5687515 0.6472477 0.9412335 0.9696478 0.8932471 0.8323443 0.8655830
## 169 170 171 172 173 174 175 176
## 0.8810719 0.8778607 0.9033264 0.9143060 0.8949596 0.8450162 0.6339085 0.9174236
## 177 178 179 180 181 182 183 184
## 0.9101564 0.9456245 0.8984300 0.7160118 0.7924309 0.8935086 0.9053933 0.9480132
## 185 186 187 188 189 190 191 192
## 0.8975425 0.9540021 0.9651597 0.9315435 0.9394075 0.8972451 0.8619886 0.8628466
## 193 194 195 196 197 198 199 200
## 0.8767868 0.9307957 0.8328112 0.8431056 0.9361453 0.8900037 0.9328325 0.9145846
## 201 202 203 204 205 206 207 208
## 0.9697232 0.8472827 0.8336666 0.6314561 0.8567686 0.9105576 0.9359998 0.9567083
## 209 210 211 212 213 214 215 216
## 0.8217165 0.9481791 0.9468998 0.8867472 0.9459730 0.8960596 0.9586198 0.4846268
## 217 218 219 220 221 222 223 224
## 0.9444047 0.9811258 0.8689969 0.9617393 0.8880560 0.7083148 0.8511805 0.9242928
## 225 226 227 228 229 230 231 232
## 0.4740492 0.9499034 0.6155745 0.7856888 0.8951706 0.8086005 0.8695553 0.6775930
## 233 234 235 236 237 238 239 240
## 0.8771469 0.6223923 0.7084090 0.9180886 0.9139725 0.9024338 0.9312299 0.9230080
## 241 242 243 244 245 246 247 248
## 0.5957631 0.9351683 0.9113295 0.8303878 0.8326014 0.7689673 0.8618683 0.8142029
## 249 250 251 252 253 254 255 256
## 0.8410904 0.8958874 0.6505877 0.9264890 0.7664435 0.8657156 0.8606595 0.9278311
## 257 258 259 260 261 262 263 264
## 0.8734387 0.9151947 0.9121932 0.7128763 0.8122951 0.8281929 0.9418868 0.8753699
## 265 266 267 268 269 270 271 272
## 0.9076372 0.8722827 0.8488718 0.6263589 0.8925482 0.9537953 0.8373556 0.8447822
## 273 274 275 276 277
## 0.9545723 0.9256813 0.9753204 0.9759924 0.8476787arbol_decision <- rpart(Clasificación ~ loan_amnt + term + int_rate + installment + grade + sub_grade + emp_title + emp_length + home_ownership + annual_inc + verification_status + purpose + title + dti + earliest_cr_line + open_acc + total_acc + initial_list_status, data = fondo, method = "class")
# Resumen del árbol de decisión
summary(arbol_decision)## Call:
## rpart(formula = Clasificación ~ loan_amnt + term + int_rate +
## installment + grade + sub_grade + emp_title + emp_length +
## home_ownership + annual_inc + verification_status + purpose +
## title + dti + earliest_cr_line + open_acc + total_acc + initial_list_status,
## data = fondo, method = "class")
## n= 923
##
## CP nsplit rel error xerror xstd
## 1 0.65972222 0 1.00000000 1.000000 0.07655730
## 2 0.08333333 1 0.34027778 1.361111 0.08628432
## 3 0.06597222 2 0.25694444 1.354167 0.08612310
## 4 0.02083333 4 0.12500000 1.402778 0.08723283
## 5 0.01736111 5 0.10416667 1.416667 0.08754194
## 6 0.01388889 7 0.06944444 1.416667 0.08754194
## 7 0.01000000 8 0.05555556 1.409722 0.08738782
##
## Variable importance
## emp_title earliest_cr_line sub_grade int_rate
## 60 23 6 2
## verification_status emp_length total_acc dti
## 2 2 1 1
## annual_inc open_acc
## 1 1
##
## Node number 1: 923 observations, complexity param=0.6597222
## predicted class=1 expected loss=0.156013 P(node) =1
## class counts: 144 779
## probabilities: 0.156 0.844
## left son=2 (131 obs) right son=3 (792 obs)
## Primary splits:
## emp_title splits as LRRRRRRLRLRLRLRRLRRRRLRRLRRLRRRRRRRRRRRLLRRRRRRRLRRRRRRLRRRRRLLRRRRLLRRLRRRRRRLRRLRRRRLRLRRRRRRRLRLRRLRLRRRLRRRRRRRRRRLRRRRRRRRRRRRRRRRRRRRRRRRRLRLRRRRRRRRLRRRRRRRRRRRLRLRRRRLRRLRRRRRRRRRRRRRRRRRRRRRRRRRRRRRLLRRLRLRRRRLRRLRRRRRLRRRRRRRLRRRRRRRRLRRRRRRRRRRRRRRRRRRRRRLRRRRRRRLRRLRRRRRRRRRRLLRRRRLRRRRRRRRRRLRRRLRRRRRRRRRLRRRRRRRRRRRRRRRRRRRRRRRRRLRRLRLRRRLRRRRLRRRRLRRRRRLLRRLRRRLRRRRRRRRRRRRRRRLRRRRRRRLRRRRRRLRLRRRRRRRRLRLRRLRRRRRLRRRRRRRLRLRRRRRRRRRRRRRRRRRLRRRRRRRRRRLRRRRRRRRRRRRRRRRRRRRRRLRRRRRLRRRLRRRRRRRRRRRRRLRRRRRRRRRRRRRLRRRLRLRLRLRRLRRRRRRRRRRRRLRRLRRRRRRLRLRRRRRRLRRLRRRRLRRRRLRRLLLRRRLLRLRRLRRRRRRRLRRLLRRRLRRRRRLRRRRRRRRRRRRRRRLRRRRLRRLRRRLRRLRLRRRRRRRLRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRR, improve=152.44160, (0 missing)
## earliest_cr_line splits as RRRRRRRRRRRRRRRRRLLRRLLRRLLRRRRRLRLRRRRLRLRRRRRRRRRRRLRRRLLRRRRRRLRLRRRRRRLLRLRRRRRRLRRRLLLLRLRRLRLRLLRLRRRRRRRRLLLRLRLRRRLLLLLRRLRRRRRRRRRRRRLRLRLRRRRRRLRRLRRRRRLRRRRRRLLLLRRLLRRLLRRRRRRRLRRRRLRRRLLRRRRRLLRRRRRRRRRRRRRRRLLRLLLLLRRLRRLRRRRRRRRLRRRRRRRLRRRRLRLLLRRRRRLRRRRRRRLRRRRRRRRRRLRRLRLRRRRRRLRRRRRRLRRRRRRRRRRRLR, improve= 55.52227, (0 missing)
## sub_grade splits as RRRRRRRLRRRLRLLLLLLLLLLLLRLLLLRL, improve= 12.36119, (0 missing)
## int_rate < 13.715 to the right, improve= 11.71562, (0 missing)
## grade splits as RRLLLLL, improve= 10.34451, (0 missing)
## Surrogate splits:
## earliest_cr_line splits as RRRRRRRRRRRRRRRRRRRRRRRRRRLLRRRRLRRRRRRRRRRRRRRRRRRRRRRRRLLRRRRRRRRRRRRRRRRLRLRRRRRRRRRRRRRRRRRRRRLRRLRRRRRRRRRRRRRRRRRRRRRRRRRRRLRRRRRRRRRRRRRRRRRRRRRRRLRRLRRRRRRRRRRRRRRRRRRRRRRRLRRRRRRRRRRRRRRRRRRRRRRRRLRRRRRRRRRRLRRRRRRRRRRRRRRRRRLRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRLRRRRRRRRRRRRRRRRRRRRRRLRRRRRRRRRRRRRRRRRRRRRRRRRRRR, agree=0.878, adj=0.137, (0 split)
## sub_grade splits as RRRRRRRRRRRRRRRRRRRRRRRRRRRRLLRL, agree=0.861, adj=0.023, (0 split)
## int_rate < 25.15 to the right, agree=0.860, adj=0.015, (0 split)
## grade splits as RRRRRRL, agree=0.859, adj=0.008, (0 split)
##
## Node number 2: 131 observations, complexity param=0.08333333
## predicted class=0 expected loss=0.1374046 P(node) =0.1419285
## class counts: 113 18
## probabilities: 0.863 0.137
## left son=4 (109 obs) right son=5 (22 obs)
## Primary splits:
## earliest_cr_line splits as -----------------LLL-LL--LLR----L------LRL-RL-L------L---LL------L-L-R-LL-LL-L------L---LLLL----L-RLLL-L---------LL-L-L--LLL-L--RL---------R--L-LRL------L--L-----L-----LLLLL--RL---L-------L--R-L-R-LL----R-L----------R----LL-LLLLL--L--L--------L----R--L-L--LLL-L-----L-------L----L-L-L-L--LRL------L------L---LLLRR---L-, improve=21.344510, (0 missing)
## emp_title splits as L------R-L-R-R--R----L--L--L-----------LR-------L------R-----LL----LL--R------L--L----L-L-------L-L--L-L---L----------R-------------------------L-L--------L-----------L-L----L--L-----------------------------LL--R-R----R--L-----L-------L--------L---------------------L-------L--L----------RL----L----------L---L---------L-------------------------L--L-L---L----L----L-----LR--L---L---------------R-------L------L-R--------L-L--R-----L-------L-L-----------------L----------L----------------------L-----L---L-------------L-------------L---L-L-L-L--L------------L--L------L-L------L--L----L----L--LLL---LL-L--L-------L--LL---L-----R---------------L----L--L---L--L-L-------L-------------------------------, improve=13.053440, (0 missing)
## sub_grade splits as R---LLLLLLLLLLLLLLLLLRLLL-L-RL-L, improve= 3.379482, (0 missing)
## total_acc < 31.5 to the right, improve= 1.665680, (0 missing)
## emp_length splits as LLLLLRLLRRR, improve= 1.602108, (0 missing)
## Surrogate splits:
## emp_title splits as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agree=0.87, adj=0.227, (0 split)
##
## Node number 3: 792 observations, complexity param=0.06597222
## predicted class=1 expected loss=0.03914141 P(node) =0.8580715
## class counts: 31 761
## probabilities: 0.039 0.961
## left son=6 (74 obs) right son=7 (718 obs)
## Primary splits:
## earliest_cr_line splits as RRRRRRRRRRRRRRRRRRLRRRLRRR--RRRR-RLRRRRRRLRRRRRRRRRRRRRRR--RRRRRRRRLRRRRRRR-R-RRRRRRRRRRRRRRRLRRRRRRL-RRRRRRRRLRLRLRRRRRRRRRLRLRR-RRRRRRRRRRRRRRLRRRRRRRR-RR-RRRRRRRRRLRRRRRRRRLRRRL-RRRRRRRRRRRRRRRRLRLRRRRL-RRRRRRRRRR-RRRRRRRLRRLLRRRRR-RRRRRRRRRRRRRRRRRRRRRRRRLRRRRRR-RRRRRRRRRRRRRRRRRRRRRR-RRRRRRRRRRRRRRRRRRRRRRRRRRRR, improve=14.565400, (0 missing)
## emp_title splits as -RRRRRR-R-R-R-RR-RRRR-RR-RR-RRRRRRRRRRR--RRRRRRR-RRLRRR-RRRRR--RRRR--RR-RRRRRR-RR-RRRR-R-RRRRRRR-R-RR-R-RRR-RRRRRRRRRR-RRRRRRRRRRRRRLRRRRRRRRRRR-R-RRRRRRRR-RRRRRRRRRRR-R-RRRR-RR-RRRRRRRRRRRRRRRRRRRRRRRRRRLRR--RR-R-LRRR-RR-RRRRL-RRRRRRR-RRLRRRRR-RRRRRRRRRRRRRRRRRRRRR-LRRRRRR-RR-RRRRRRRRRR--RRRR-RRRRRRRRRR-RRL-RRRRRRRRR-RRRRRRRRRRRRRRRRRRRLRRRRR-RR-R-RRR-RLLR-RRRR-RRRRR--RR-RRR-RRRRRRRRRRRRRRR-RRRRRRR-RRRRLR-R-RRRRRRLR-R-RR-RRRRR-RRRRRRR-R-RRRLRRRRRLRRLRRRR-RRRRRRRRLR-RRRRRRRRRRRRRRRRRRRRRR-RRRRR-RRR-RRRRRRRRRRRRL-RRRRRRRRRRRRR-RRR-R-R-R-RR-RRRRRRRRRRRR-RR-RRRRRR-R-RRRRRR-RR-RRRR-RRRR-RR---RRR--R-RR-RRRRRRR-RR--RRR-RRRRR-RRRRLRRRLRRRRRR-RRRR-RR-RRR-RR-R-RRRRRRR-RRRLRRRRRRRRRRRRRRRRRRRRRRRRRRR, improve=10.164850, (0 missing)
## sub_grade splits as RRRRRLRRLRLRRLRLRRRLLLLRLRRLR-R-, improve= 2.074506, (0 missing)
## total_acc < 64.5 to the right, improve= 1.575787, (0 missing)
## int_rate < 18.23 to the right, improve= 1.259924, (0 missing)
## Surrogate splits:
## emp_title splits as -RRLRRR-R-R-R-RR-RLRR-RR-RR-RRRRRRRRRRR--LRRRRRR-RRRRRR-RRRRR--RRRR--RR-RRRRRR-RR-RRRR-R-RRRRRRR-L-RR-R-RRR-RRLRRRRRRR-RRRRRRRRRRLRRRRRRRRRRRRRR-R-RRRRRRRR-RLRRRRRRRRL-R-RRRR-RR-RRRRRRRRRRRRRRRRRRRRRRRRRRRRR--RR-R-RRRR-RR-RRRRL-RRRRRRR-RRRRRRRR-RRLRRRRRLRLRRLRLRRRRL-RRRRRRR-RR-LRRRRRRRRR--RLLR-RLRRRRRRRR-RRR-RRRRRRRLL-RRRRRRRRRRRRRRRRRRRRRRRRR-RR-R-RRR-RRRL-RRRR-RRRRR--RR-RRR-RRRRRRRRRRRRRRR-RRLRRRR-RRRRRR-R-RRRRRRRR-L-RR-RRRRL-LRRRRRR-R-RRRRRRRRRRRRRRRRR-RRRRRRRLRR-RRLRRRRRRRRRRRRRRLRLRR-RRRRR-RRR-RRRRRRRRRRRRR-RRRRRLRRRRRRR-RRR-R-R-R-RR-RLRRRRRRRRRR-RR-RRRRLR-R-RRRRRR-RR-RRLR-LRRR-RR---RRR--R-RR-RRLRRRR-RR--RRL-RRRRR-RRRRRRRRRRRRRRR-RRRR-RR-RRR-RR-R-RRRRRRR-RRRRRRRRRRRRRRRRRRRRRRRRRRRRRRR, agree=0.953, adj=0.500, (0 split)
## sub_grade splits as RRRRRRRRRRRRRRRRRRRRRRRRRRRLR-R-, agree=0.908, adj=0.014, (0 split)
##
## Node number 4: 109 observations
## predicted class=0 expected loss=0.009174312 P(node) =0.1180932
## class counts: 108 1
## probabilities: 0.991 0.009
##
## Node number 5: 22 observations, complexity param=0.02083333
## predicted class=1 expected loss=0.2272727 P(node) =0.02383532
## class counts: 5 17
## probabilities: 0.227 0.773
## left son=10 (7 obs) right son=11 (15 obs)
## Primary splits:
## emp_title splits as -------R---R-R--R-----------------------R--------------R---------------R----------------------------------------------R--------------------------------------------------------------------------------------------R-R----R---------------------------------------------------------------------R---------------------------------------------------------------------------L------R----------------------R----------------L-------------L---------------------------------------------------------------------------------------------------------------------------------------------L----------------------------------------------------------L------------------------------L-----------------------------------------, improve=4.870130, (0 missing)
## earliest_cr_line splits as ---------------------------R------------L--R-------------------------R----------------------------L-----------------------------R----------R-----L-----------------------------R---------------L---R-------R------------R-------------------------------L----------------------------------------R---------------------RR-----, improve=2.727273, (0 missing)
## sub_grade splits as R----R--R-LLRLL-RRRR-R------R---, improve=2.272727, (0 missing)
## emp_length splits as LRLRRRR-RRR, improve=2.272727, (0 missing)
## grade splits as RRLRRR-, improve=1.893939, (0 missing)
## Surrogate splits:
## annual_inc < 69000 to the right, agree=0.818, adj=0.429, (0 split)
## loan_amnt < 22000 to the right, agree=0.773, adj=0.286, (0 split)
## installment < 481.75 to the right, agree=0.773, adj=0.286, (0 split)
## sub_grade splits as R----R--R-RLRRR-LRRR-R------R---, agree=0.773, adj=0.286, (0 split)
## emp_length splits as RRLRRRR-RRR, agree=0.773, adj=0.286, (0 split)
##
## Node number 6: 74 observations, complexity param=0.06597222
## predicted class=1 expected loss=0.3378378 P(node) =0.08017335
## class counts: 25 49
## probabilities: 0.338 0.662
## left son=12 (31 obs) right son=13 (43 obs)
## Primary splits:
## emp_title splits as ---R--------------R----------------------R---------R---------------------------------------------R------------R------------------R--L------------------------R--------R--------------R----------------------L---------------------L-----------L--------R-----R-R--R-R----R-L----------R------------RR---R-----------L--------RR--------------------L----------------LLR--------------------------------------R---------L----------R--R--------R-R------------L-----L--L------------RL----R----R---------R-R-------------R--L---R----L------R---LR-----------------R------------------R----------------R--R---------------------R-----------R-----------L-------------L-------------------------L---------------------------, improve=23.430690, (0 missing)
## sub_grade splits as -RRRRLRRLRRLRLLRR-RLLLL-L--L----, improve= 7.630713, (0 missing)
## earliest_cr_line splits as ------------------R---R-----------L------R-------------------------R-------------------------L------R---------R-L-L---------R-R-----------------R---------------------R--------R---L-----------------R-R----L-------------------R--RR------------------------------R----------------------------------------------------------, improve= 5.176290, (0 missing)
## int_rate < 18.23 to the right, improve= 3.919075, (0 missing)
## verification_status splits as RRL, improve= 3.033108, (0 missing)
## Surrogate splits:
## earliest_cr_line splits as ------------------R---R-----------L------R-------------------------L-------------------------L------R---------R-L-L---------R-R-----------------R---------------------R--------R---L-----------------R-R----L-------------------R--RR------------------------------L----------------------------------------------------------, agree=0.716, adj=0.323, (0 split)
## sub_grade splits as -RRRRRRRLRRRRRLRR-RLLRR-L--R----, agree=0.689, adj=0.258, (0 split)
## verification_status splits as RRL, agree=0.689, adj=0.258, (0 split)
## int_rate < 18.23 to the right, agree=0.649, adj=0.161, (0 split)
## dti < 25.81 to the right, agree=0.649, adj=0.161, (0 split)
##
## Node number 7: 718 observations, complexity param=0.01736111
## predicted class=1 expected loss=0.008356546 P(node) =0.7778982
## class counts: 6 712
## probabilities: 0.008 0.992
## left son=14 (28 obs) right son=15 (690 obs)
## Primary splits:
## emp_title splits as -RR-RRR-R-R-R-RR-R-RR-RR-RR-RRRRRRRRRRR---RRRRRR-RRLRRR-RRRRR--RRRR--RR-RRRRRR-RR-RRRR-R-RRRRRRR---RR-R-RRR-RR-RRRRRRR-RRRRRRRRRR-RRRRRRRRRRRRRR-R-RRRRRRRR-R-RRRRRRRR--R-RRRR-RR-RRRRRRRRRRRRRRRRRRRRRRRRRRRRR--RR-R-LRRR-RR-RRRRR-RRRRRRR-RRLRRRRR-RR-RRRRR-R-RR-R-RRRR--LRRRRRR-RR--RRRRRRRRR--R--R-R-RRRRRRRR-RRR-RRRRRRR---RRRRRRRRRRRRRRRRRRRRRRRRR-RR-R-RRR-RRR--RRRR-RRRRR--RR-RRR-RRRRRRRRRRRRRRR-RR-RRRR-RRRRRR-R-RRRRRRLR---RR-RRRR---RRRRRR-R-RRRRRRRRRRRRRRRRR-RRRRRRR-RR-RR-RRRRRRRRRRRRRR-R-RR-RRRRR-RRR-RRRRRRRRRRRRR-RRRRR-RRRRRRR-RRR-R-R-R-RR-R-RRRRRRRRRR-RR-RRRR-R-R-RRRRRR-RR-RR-R--RRR-RR---RRR--R-RR-RR-RRRR-RR--RR--RRRRR-RRRRRRRRLRRRRRR-RRRR-RR-RRR-RR-R-RRRRRRR-RRRRRRRRRRRRRRRRRRRRRRRRRRRRRRR, improve=2.4711500, (0 missing)
## earliest_cr_line splits as RRRRRRRRRRRRRRRRRR-RRR-RRR--RRRR-R-RRRRRR-RRRRRRRRRRRRRRR--RRRRRRRR-RRLRRRR-R-RRRRRRRRRRRRRRR-RRRRRR--RRRRRRRR-R-R-RRRRRRRRR-R-RR-RRRRRRRRRRRRRR-RRRRRRRR-RR-RRRRRRRRR-RRRRRLRR-RRR--RRRRRRRRRRRRRRRR-R-RRRR--RRRRRRRRRR-RRRRRRR-RR--RRRRR-RRRRRRRRRRRRRLRRRRRRRRRR-RRRRRR-RRRRRRRRRRRRLRRRRRRRRR-RRRRRRRRRRRRRRRRRRRRRLLRRRRR, improve=1.2582120, (0 missing)
## total_acc < 64.5 to the right, improve=0.9447919, (0 missing)
## sub_grade splits as RRRRRRRRRRLRRLRLRRRRRRLRRRRRR-R-, improve=0.5483701, (0 missing)
## emp_length splits as RRRRRRRRRRL, improve=0.4453055, (0 missing)
## Surrogate splits:
## earliest_cr_line splits as RRRRRRRRRRRRRRRRRR-RRR-RRR--RRRR-R-RRRRRR-RRRRRRRRRRRRRRR--RRRRRRRR-RRRRRRR-R-RRRRRRRRRRRRRRR-RRRRRR--LRRRRRRR-R-R-RRRRRRRRR-R-RR-RRRRRRRLRRRRRR-RRRRRRRR-RR-RRRRRRRRR-RRRRRRRR-RRR--RRRRRRRRRRRRRRRR-R-RRRR--RRRRRRRRRR-RRRRRRR-RR--RRRRR-RRLRRRRRRRRRRRRRRRRRRRRR-RRRRRR-RRRRRRRRRRRRRRRRRRRRRR-RRRRRRRRRRRRRRRRRLRRRRRRRRRR, agree=0.967, adj=0.143, (0 split)
##
## Node number 10: 7 observations
## predicted class=0 expected loss=0.2857143 P(node) =0.007583965
## class counts: 5 2
## probabilities: 0.714 0.286
##
## Node number 11: 15 observations
## predicted class=1 expected loss=0 P(node) =0.01625135
## class counts: 0 15
## probabilities: 0.000 1.000
##
## Node number 12: 31 observations, complexity param=0.01388889
## predicted class=0 expected loss=0.1935484 P(node) =0.03358613
## class counts: 25 6
## probabilities: 0.806 0.194
## left son=24 (21 obs) right son=25 (10 obs)
## Primary splits:
## sub_grade splits as ---RLLRLLRRL-LLLR--LLLL-L--L----, improve=4.877419, (0 missing)
## earliest_cr_line splits as ------------------L---L-----------L------L-------------------------R-------------------------R------L---------R-L-L---------L-L-----------------L---------------------L--------L---L-----------------R-L----L-------------------L--LL------------------------------R----------------------------------------------------------, improve=3.677419, (0 missing)
## emp_title splits as ------------------------------------------------------------------------------------------------------------------------------------L-----------------------------------------------------------------------L---------------------R-----------L----------------------------L----------------------------------------L------------------------------L----------------LL-------------------------------------------------L-------------------------------------L-----L--R-------------R--------------------------------------L--------L----------L-------------------------------------------------------------------------------------------------------L-------------R-------------------------L---------------------------, improve=3.215881, (0 missing)
## open_acc < 17 to the left, improve=2.582181, (0 missing)
## total_acc < 26.5 to the left, improve=1.949349, (0 missing)
## Surrogate splits:
## earliest_cr_line splits as ------------------L---L-----------R------L-------------------------R-------------------------R------L---------L-L-L---------L-L-----------------L---------------------L--------L---L-----------------L-L----R-------------------L--LL------------------------------L----------------------------------------------------------, agree=0.871, adj=0.6, (0 split)
## total_acc < 27.5 to the left, agree=0.839, adj=0.5, (0 split)
## emp_title splits as ------------------------------------------------------------------------------------------------------------------------------------R-----------------------------------------------------------------------L---------------------L-----------L----------------------------L----------------------------------------L------------------------------L----------------LL-------------------------------------------------L-------------------------------------L-----L--L-------------R--------------------------------------L--------R----------L-------------------------------------------------------------------------------------------------------L-------------L-------------------------L---------------------------, agree=0.806, adj=0.4, (0 split)
## purpose splits as -LL--L--LR-, agree=0.742, adj=0.2, (0 split)
## title splits as R-LL--L--L-, agree=0.742, adj=0.2, (0 split)
##
## Node number 13: 43 observations
## predicted class=1 expected loss=0 P(node) =0.04658722
## class counts: 0 43
## probabilities: 0.000 1.000
##
## Node number 14: 28 observations, complexity param=0.01736111
## predicted class=1 expected loss=0.2142857 P(node) =0.03033586
## class counts: 6 22
## probabilities: 0.214 0.786
## left son=28 (7 obs) right son=29 (21 obs)
## Primary splits:
## earliest_cr_line splits as --------------R----R----------------------------R----------------R--R-L-R------------R----------------R----------------------------------R---------R------------------------L---------------------------R----------------------R-------------R----------L----R---R---------------------L----------R-------------R-RR--RLL-----, improve=7.714286, (0 missing)
## sub_grade splits as R-R--RR-RRLRRL-LR-R--RL--R------, improve=5.428571, (0 missing)
## emp_length splits as RLRLRLRRRLL, improve=5.428571, (0 missing)
## open_acc < 17.5 to the right, improve=2.380952, (0 missing)
## total_acc < 35.5 to the right, improve=1.828571, (0 missing)
## Surrogate splits:
## sub_grade splits as L-R--RR-RRLRRL-RR-R--RR--R------, agree=0.893, adj=0.571, (0 split)
## emp_length splits as RRRRRRRLRLL, agree=0.893, adj=0.571, (0 split)
## open_acc < 24.5 to the right, agree=0.821, adj=0.286, (0 split)
## total_acc < 53.5 to the right, agree=0.821, adj=0.286, (0 split)
## annual_inc < 43500 to the left, agree=0.786, adj=0.143, (0 split)
##
## Node number 15: 690 observations
## predicted class=1 expected loss=0 P(node) =0.7475623
## class counts: 0 690
## probabilities: 0.000 1.000
##
## Node number 24: 21 observations
## predicted class=0 expected loss=0 P(node) =0.0227519
## class counts: 21 0
## probabilities: 1.000 0.000
##
## Node number 25: 10 observations
## predicted class=1 expected loss=0.4 P(node) =0.01083424
## class counts: 4 6
## probabilities: 0.400 0.600
##
## Node number 28: 7 observations
## predicted class=0 expected loss=0.1428571 P(node) =0.007583965
## class counts: 6 1
## probabilities: 0.857 0.143
##
## Node number 29: 21 observations
## predicted class=1 expected loss=0 P(node) =0.0227519
## class counts: 0 21
## probabilities: 0.000 1.000# Modelo Random Forest (conjunto de árboles)
modelo_random_forest <- randomForest(Clasificación ~ loan_amnt + term + int_rate + installment + grade + sub_grade + emp_title + emp_length + home_ownership + annual_inc + verification_status + purpose + title + dti + earliest_cr_line + open_acc + total_acc + initial_list_status, data = fondo)## Warning in randomForest.default(m, y, ...): The response has five or fewer
## unique values. Are you sure you want to do regression?# Resumen del modelo Random Forest
print(modelo_random_forest)##
## Call:
## randomForest(formula = Clasificación ~ loan_amnt + term + int_rate + installment + grade + sub_grade + emp_title + emp_length + home_ownership + annual_inc + verification_status + purpose + title + dti + earliest_cr_line + open_acc + total_acc + initial_list_status, data = fondo)
## Type of random forest: regression
## Number of trees: 500
## No. of variables tried at each split: 6
##
## Mean of squared residuals: 0.1324804
## % Var explained: -0.61Este informe presenta un análisis de riesgo crediticio realizado en el conjunto de datos contenido en el archivo “credit_shortclean.xlsx”. El análisis tiene como objetivo evaluar la probabilidad de incumplimiento de solicitudes de crédito y clasificarlas como “Aprobadas” o “No Aprobadas” en función de las variables “loan_amnt,” “int_rate,” “annual_inc,” y “total_acc.”
Técnicas Aplicadas:
Preparación de Datos:
Se cargaron los datos desde el archivo “credit_shortclean.xlsx.” Se verificó la estructura de la base de datos y se reemplazaron los valores faltantes por la media de las variables correspondientes. Clasificación de Solicitudes:
Se clasificaron las solicitudes en “Aprobadas” y “No Aprobadas” en función de la variable “loan_status.” Se consideraron “Fully Paid” y “Current” como “Aprobadas” y “Charged Off” como “No Aprobada.” Modelo de Regresión Logística:
Se ajustó un modelo de regresión logística para predecir la probabilidad de incumplimiento (Clasificación binaria “Aprobada” o “No Aprobada”) utilizando las variables “loan_amnt,” “int_rate,” “annual_inc,” y “total_acc.” Calidad de Ajuste del Modelo:
El modelo de regresión logística fue ajustado utilizando los siguientes coeficientes:
Coeficiente para “loan_amnt”: 2.308e-05 Coeficiente para “int_rate”: -1.542e-01 Coeficiente para “annual_inc”: 5.010e-06 Coeficiente para “total_acc”: -8.363e-03 Los coeficientes estimados se consideran estadísticamente significativos para “loan_amnt” y “int_rate” (p < 0.05). Esto indica que estas dos variables tienen un impacto significativo en la clasificación de las solicitudes.
Conclusiones:
El análisis de riesgo crediticio realizado en este informe proporciona una base sólida para evaluar la probabilidad de incumplimiento de solicitudes de crédito. Los resultados indican que son muy importantes las variables y que entre mas uses y mas informacion le dispongas al modelo este sera mas preciso.