Chaos_Matrix_HW

library(caret)

Loading required package: ggplot2

Loading required package: lattice

options(show.signif.stars = FALSE)

# import
df <- read.csv("insurance-testing-data2.csv", stringsAsFactors = FALSE, na.strings = c("", "NA"))

# clean up the money columns, they come in as text like "$18,755"
df$INCOME <- as.numeric(gsub("[$,]", "", df$INCOME))
df$HOME_VAL <- as.numeric(gsub("[$,]", "", df$HOME_VAL))
df$BLUEBOOK <- as.numeric(gsub("[$,]", "", df$BLUEBOOK))
df$OLDCLAIM <- as.numeric(gsub("[$,]", "", df$OLDCLAIM))

df$INDEX <- NULL
df$TARGET_AMT <- NULL

# fill NAs with the median
df$AGE[is.na(df$AGE)] <- median(df$AGE, na.rm = TRUE)
df$YOJ[is.na(df$YOJ)] <- median(df$YOJ, na.rm = TRUE)
df$INCOME[is.na(df$INCOME)] <- median(df$INCOME, na.rm = TRUE)
df$HOME_VAL[is.na(df$HOME_VAL)] <- median(df$HOME_VAL, na.rm = TRUE)
df$BLUEBOOK[is.na(df$BLUEBOOK)] <- median(df$BLUEBOOK, na.rm = TRUE)
df$OLDCLAIM[is.na(df$OLDCLAIM)] <- median(df$OLDCLAIM, na.rm = TRUE)
df$CAR_AGE[is.na(df$CAR_AGE)] <- median(df$CAR_AGE, na.rm = TRUE)

# make the text columns factors
df$PARENT1 <- factor(df$PARENT1)
df$MSTATUS <- factor(df$MSTATUS)
df$SEX <- factor(df$SEX)
df$EDUCATION <- factor(df$EDUCATION)
df$JOB <- factor(df$JOB)
df$CAR_USE <- factor(df$CAR_USE)
df$CAR_TYPE <- factor(df$CAR_TYPE)
df$RED_CAR <- factor(df$RED_CAR)
df$REVOKED <- factor(df$REVOKED)
df$URBANICITY <- factor(df$URBANICITY)
df$TARGET_FLAG <- factor(df$TARGET_FLAG, levels = c(0, 1))

# split into train and test
set.seed(42)
split <- createDataPartition(df$TARGET_FLAG, p = 0.75, list = FALSE)
train <- df[split, ]
test <- df[-split, ]

# build the logistic model on train, using every predictor
model <- glm(TARGET_FLAG ~ ., data = train, family = binomial)
summary(model)

Call:
glm(formula = TARGET_FLAG ~ ., family = binomial, data = train)

Coefficients:
                                  Estimate Std. Error z value Pr(>|z|)
(Intercept)                     -1.712e-01  7.569e-01  -0.226 0.821042
KIDSDRIV                         2.775e-02  1.730e-01   0.160 0.872605
AGE                              5.208e-03  1.121e-02   0.464 0.642318
HOMEKIDS                         1.192e-02  9.925e-02   0.120 0.904426
YOJ                             -5.605e-03  2.403e-02  -0.233 0.815527
INCOME                          -5.387e-06  3.437e-06  -1.567 0.117031
PARENT1Yes                       3.780e-01  3.073e-01   1.230 0.218719
HOME_VAL                        -3.670e-06  1.003e-06  -3.658 0.000254
MSTATUSz_No                      3.311e-01  2.279e-01   1.453 0.146167
SEXz_F                           3.368e-01  3.204e-01   1.051 0.293117
EDUCATIONBachelors              -3.207e-01  3.216e-01  -0.997 0.318623
EDUCATIONMasters                -6.953e-01  5.342e-01  -1.302 0.193053
EDUCATIONPhD                     2.417e-01  6.563e-01   0.368 0.712703
EDUCATIONz_High School          -6.722e-02  2.626e-01  -0.256 0.797933
JOBDoctor                       -4.646e-01  7.617e-01  -0.610 0.541901
JOBHome Maker                   -5.526e-01  4.017e-01  -1.376 0.168976
JOBLawyer                        3.047e-01  5.505e-01   0.554 0.579840
JOBManager                      -1.360e+00  3.940e-01  -3.451 0.000559
JOBProfessional                 -4.247e-01  3.481e-01  -1.220 0.222508
JOBStudent                      -7.930e-01  3.655e-01  -2.170 0.030037
JOBz_Blue Collar                -5.654e-01  2.974e-01  -1.901 0.057339
TRAVTIME                         2.119e-02  5.402e-03   3.923 8.75e-05
CAR_USEPrivate                  -1.092e+00  2.493e-01  -4.378 1.20e-05
BLUEBOOK                        -2.763e-05  1.439e-05  -1.920 0.054871
TIF                             -3.625e-02  1.981e-02  -1.830 0.067214
CAR_TYPEPanel Truck              1.116e+00  4.527e-01   2.465 0.013698
CAR_TYPEPickup                   6.233e-01  2.870e-01   2.172 0.029887
CAR_TYPESports Car               1.214e+00  3.549e-01   3.422 0.000621
CAR_TYPEVan                      8.214e-01  3.496e-01   2.350 0.018789
CAR_TYPEz_SUV                    5.202e-01  3.067e-01   1.696 0.089910
RED_CARyes                       1.041e-02  2.532e-01   0.041 0.967223
OLDCLAIM                        -1.687e-05  1.233e-05  -1.369 0.171065
CLM_FREQ                         2.318e-01  8.188e-02   2.831 0.004646
REVOKEDYes                       7.447e-01  2.753e-01   2.705 0.006830
MVR_PTS                          9.847e-02  3.796e-02   2.594 0.009478
CAR_AGE                         -8.597e-03  2.158e-02  -0.398 0.690284
URBANICITYz_Highly Rural/ Rural -2.544e+00  3.038e-01  -8.372  < 2e-16

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 1317.3  on 1141  degrees of freedom
Residual deviance:  978.3  on 1105  degrees of freedom
  (84 observations deleted due to missingness)
AIC: 1052.3

Number of Fisher Scoring iterations: 5

# predict onto test data
prob <- predict(model, newdata = test, type = "response")
pred <- ifelse(prob > 0.5, 1, 0)

# confusion matrix on test data, actual on rows and predicted on columns
cm <- table(Actual = test$TARGET_FLAG, Predicted = pred)
cm

      Predicted
Actual   0   1
     0 253  28
     1  53  50

accuracy <- sum(diag(cm)) / sum(cm)
sensitivity <- cm["1", "1"] / sum(cm["1", ])     # recall, TP / (TP + FN)
precision <- cm["1", "1"] / sum(cm[, "1"])       # TP / (TP + FP)
f1 <- 2 * (precision * sensitivity) / (precision + sensitivity)

accuracy

[1] 0.7890625

sensitivity

[1] 0.4854369

precision

[1] 0.6410256

f1

[1] 0.5524862