Support Vector Machine with Radial Basis Kernel using Package e1071

Data Mining

Description for Support Vector Machine with Radial Basis Kernel using Package e1071

Yeongeun Jeon , Jung In Seo
2023-04-17

Support Vector Machine의 장점


Support Vector Machine의 단점


실습 자료 : 유니버셜 은행의 고객 2,500명에 대한 자료(출처 : Data Mining for Business Intelligence, Shmueli et al. 2010)이며, 총 13개의 변수를 포함하고 있다. 이 자료에서 TargetPersonal Loan이다.




1. 데이터 불러오기

pacman::p_load("data.table", "dplyr",
               "caret",
               "ggplot2", "GGally",
               "e1071")


UB <- fread("../Universal Bank_Main.csv")                               # 데이터 불러오기

UB %>%
  as_tibble
# A tibble: 2,500 × 14
      ID   Age Experience Income `ZIP Code` Family CCAvg Education
   <int> <int>      <int>  <int>      <int>  <int> <dbl>     <int>
 1     1    25          1     49      91107      4   1.6         1
 2     2    45         19     34      90089      3   1.5         1
 3     3    39         15     11      94720      1   1           1
 4     4    35          9    100      94112      1   2.7         2
 5     5    35          8     45      91330      4   1           2
 6     6    37         13     29      92121      4   0.4         2
 7     7    53         27     72      91711      2   1.5         2
 8     8    50         24     22      93943      1   0.3         3
 9     9    35         10     81      90089      3   0.6         2
10    10    34          9    180      93023      1   8.9         3
# ℹ 2,490 more rows
# ℹ 6 more variables: Mortgage <int>, `Personal Loan` <int>,
#   `Securities Account` <int>, `CD Account` <int>, Online <int>,
#   CreditCard <int>

2. 데이터 전처리 I

UB %<>%
  data.frame() %>%                                                      # Data Frame 형태로 변환 
  mutate(Personal.Loan = ifelse(Personal.Loan == 1, "yes", "no")) %>%   # Target을 문자형 변수로 변환
  select(-1)                                                            # ID 변수 제거

# 1. Convert to Factor
fac.col <- c("Family", "Education", "Securities.Account", 
             "CD.Account", "Online", "CreditCard",
             # Target
             "Personal.Loan")

UB <- UB %>% 
  mutate_at(fac.col, as.factor)                                         # 범주형으로 변환

glimpse(UB)                                                             # 데이터 구조 확인
Rows: 2,500
Columns: 13
$ Age                <int> 25, 45, 39, 35, 35, 37, 53, 50, 35, 34, 6…
$ Experience         <int> 1, 19, 15, 9, 8, 13, 27, 24, 10, 9, 39, 5…
$ Income             <int> 49, 34, 11, 100, 45, 29, 72, 22, 81, 180,…
$ ZIP.Code           <int> 91107, 90089, 94720, 94112, 91330, 92121,…
$ Family             <fct> 4, 3, 1, 1, 4, 4, 2, 1, 3, 1, 4, 3, 2, 4,…
$ CCAvg              <dbl> 1.6, 1.5, 1.0, 2.7, 1.0, 0.4, 1.5, 0.3, 0…
$ Education          <fct> 1, 1, 1, 2, 2, 2, 2, 3, 2, 3, 3, 2, 3, 2,…
$ Mortgage           <int> 0, 0, 0, 0, 0, 155, 0, 0, 104, 0, 0, 0, 0…
$ Personal.Loan      <fct> no, no, no, no, no, no, no, no, no, yes, …
$ Securities.Account <fct> 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0,…
$ CD.Account         <fct> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,…
$ Online             <fct> 0, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 1, 0, 1,…
$ CreditCard         <fct> 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0,…
# 2. Convert One-hot Encoding for 범주형 예측 변수
dummies <- dummyVars(formula = ~ .,                                     # formula : ~ 예측 변수 / "." : data에 포함된 모든 변수를 의미
                     data = UB[,-9],                                    # Dataset including Only 예측 변수 -> Target 제외
                     fullRank = FALSE)                                  # fullRank = TRUE : Dummy Variable, fullRank = FALSE : One-hot Encoding

UB.Var   <- predict(dummies, newdata = UB) %>%                          # 범주형 예측 변수에 대한 One-hot Encoding 변환
  data.frame()                                                          # Data Frame 형태로 변환 

glimpse(UB.Var)                                                         # 데이터 구조 확인
Rows: 2,500
Columns: 21
$ Age                  <dbl> 25, 45, 39, 35, 35, 37, 53, 50, 35, 34,…
$ Experience           <dbl> 1, 19, 15, 9, 8, 13, 27, 24, 10, 9, 39,…
$ Income               <dbl> 49, 34, 11, 100, 45, 29, 72, 22, 81, 18…
$ ZIP.Code             <dbl> 91107, 90089, 94720, 94112, 91330, 9212…
$ Family.1             <dbl> 0, 0, 1, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, …
$ Family.2             <dbl> 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, …
$ Family.3             <dbl> 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, …
$ Family.4             <dbl> 1, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0, 0, …
$ CCAvg                <dbl> 1.6, 1.5, 1.0, 2.7, 1.0, 0.4, 1.5, 0.3,…
$ Education.1          <dbl> 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
$ Education.2          <dbl> 0, 0, 0, 1, 1, 1, 1, 0, 1, 0, 0, 1, 0, …
$ Education.3          <dbl> 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 0, 1, …
$ Mortgage             <dbl> 0, 0, 0, 0, 0, 155, 0, 0, 104, 0, 0, 0,…
$ Securities.Account.0 <dbl> 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, …
$ Securities.Account.1 <dbl> 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, …
$ CD.Account.0         <dbl> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, …
$ CD.Account.1         <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
$ Online.0             <dbl> 1, 1, 1, 1, 1, 0, 0, 1, 0, 1, 1, 0, 1, …
$ Online.1             <dbl> 0, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 1, 0, …
$ CreditCard.0         <dbl> 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, …
$ CreditCard.1         <dbl> 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, …
# 3. Combine Target with 변환된 예측 변수
UB.df <- data.frame(Personal.Loan = UB$Personal.Loan, 
                    UB.Var)

UB.df %>%
  as_tibble
# A tibble: 2,500 × 22
   Personal.Loan   Age Experience Income ZIP.Code Family.1 Family.2
   <fct>         <dbl>      <dbl>  <dbl>    <dbl>    <dbl>    <dbl>
 1 no               25          1     49    91107        0        0
 2 no               45         19     34    90089        0        0
 3 no               39         15     11    94720        1        0
 4 no               35          9    100    94112        1        0
 5 no               35          8     45    91330        0        0
 6 no               37         13     29    92121        0        0
 7 no               53         27     72    91711        0        1
 8 no               50         24     22    93943        1        0
 9 no               35         10     81    90089        0        0
10 yes              34          9    180    93023        1        0
# ℹ 2,490 more rows
# ℹ 15 more variables: Family.3 <dbl>, Family.4 <dbl>, CCAvg <dbl>,
#   Education.1 <dbl>, Education.2 <dbl>, Education.3 <dbl>,
#   Mortgage <dbl>, Securities.Account.0 <dbl>,
#   Securities.Account.1 <dbl>, CD.Account.0 <dbl>,
#   CD.Account.1 <dbl>, Online.0 <dbl>, Online.1 <dbl>,
#   CreditCard.0 <dbl>, CreditCard.1 <dbl>
glimpse(UB.df)                                                          # 데이터 구조 확인
Rows: 2,500
Columns: 22
$ Personal.Loan        <fct> no, no, no, no, no, no, no, no, no, yes…
$ Age                  <dbl> 25, 45, 39, 35, 35, 37, 53, 50, 35, 34,…
$ Experience           <dbl> 1, 19, 15, 9, 8, 13, 27, 24, 10, 9, 39,…
$ Income               <dbl> 49, 34, 11, 100, 45, 29, 72, 22, 81, 18…
$ ZIP.Code             <dbl> 91107, 90089, 94720, 94112, 91330, 9212…
$ Family.1             <dbl> 0, 0, 1, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, …
$ Family.2             <dbl> 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, …
$ Family.3             <dbl> 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, …
$ Family.4             <dbl> 1, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0, 0, …
$ CCAvg                <dbl> 1.6, 1.5, 1.0, 2.7, 1.0, 0.4, 1.5, 0.3,…
$ Education.1          <dbl> 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
$ Education.2          <dbl> 0, 0, 0, 1, 1, 1, 1, 0, 1, 0, 0, 1, 0, …
$ Education.3          <dbl> 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 0, 1, …
$ Mortgage             <dbl> 0, 0, 0, 0, 0, 155, 0, 0, 104, 0, 0, 0,…
$ Securities.Account.0 <dbl> 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, …
$ Securities.Account.1 <dbl> 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, …
$ CD.Account.0         <dbl> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, …
$ CD.Account.1         <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
$ Online.0             <dbl> 1, 1, 1, 1, 1, 0, 0, 1, 0, 1, 1, 0, 1, …
$ Online.1             <dbl> 0, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 1, 0, …
$ CreditCard.0         <dbl> 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, …
$ CreditCard.1         <dbl> 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, …

3. 데이터 탐색

ggpairs(UB,                                           # In 2-1
        columns = c("Age", "Experience", "Income",    # 수치형 예측 변수
                    "ZIP.Code", "CCAvg", "Mortgage"),                            
        aes(colour = Personal.Loan)) +                # Target의 범주에 따라 색깔을 다르게 표현
  theme_bw()
ggpairs(UB,                                           # In 2-1
        columns = c("Age", "Experience", "Income",    # 수치형 예측 변수
                    "ZIP.Code", "CCAvg", "Mortgage"), 
        aes(colour = Personal.Loan)) +                # Target의 범주에 따라 색깔을 다르게 표현
  scale_color_brewer(palette="Purples") +             # 특정 색깔 지정
  scale_fill_brewer(palette="Purples") +              # 특정 색깔 지정
  theme_bw()
ggpairs(UB,                                           # In 2-1
        columns = c("Age", "Income",                  # 수치형 예측 변수
                    "Family", "Education"),           # 범주형 예측 변수
        aes(colour = Personal.Loan, alpha = 0.8)) +   # Target의 범주에 따라 색깔을 다르게 표현
  scale_colour_manual(values = c("purple","cyan4")) + # 특정 색깔 지정
  scale_fill_manual(values = c("purple","cyan4")) +   # 특정 색깔 지정
  theme_bw()


4. 데이터 분할

# Partition (Training Dataset : Test Dataset = 7:3)
y      <- UB.df$Personal.Loan                            # Target
 
set.seed(200)
ind    <- createDataPartition(y, p = 0.7, list = T)      # Index를 이용하여 7:3으로 분할
UB.trd <- UB.df[ind$Resample1,]                          # Training Dataset
UB.ted <- UB.df[-ind$Resample1,]                         # Test Dataset

5. 데이터 전처리 II

# Standardization
preProcValues <- preProcess(UB.trd, 
                            method = c("center", "scale"))  # Standardization 정의 -> Training Dataset에 대한 평균과 표준편차 계산 

UB.trd <- predict(preProcValues, UB.trd)                    # Standardization for Training Dataset
UB.ted <- predict(preProcValues, UB.ted)                    # Standardization for Test Dataset

glimpse(UB.trd)                                             # 데이터 구조 확인
Rows: 1,751
Columns: 22
$ Personal.Loan        <fct> no, no, no, no, no, no, no, yes, no, no…
$ Age                  <dbl> -0.05431273, -0.57446728, -0.92123699, …
$ Experience           <dbl> -0.12175295, -0.46882565, -0.98943471, …
$ Income               <dbl> -0.85867297, -1.35649686, 0.56986515, -…
$ ZIP.Code             <dbl> -1.75250883, 0.88354520, 0.53745994, -1…
$ Family.1             <dbl> -0.6355621, 1.5725118, 1.5725118, -0.63…
$ Family.2             <dbl> -0.5774051, -0.5774051, -0.5774051, -0.…
$ Family.3             <dbl> 2.0037210, -0.4987865, -0.4987865, -0.4…
$ Family.4             <dbl> -0.5967491, -0.5967491, -0.5967491, 1.6…
$ CCAvg                <dbl> -0.25119120, -0.53150921, 0.42157204, -…
$ Education.1          <dbl> 1.1482386, 1.1482386, -0.8704018, -0.87…
$ Education.2          <dbl> -0.6196534, -0.6196534, 1.6128838, 1.61…
$ Education.3          <dbl> -0.6408777, -0.6408777, -0.6408777, -0.…
$ Mortgage             <dbl> -0.5664192, -0.5664192, -0.5664192, -0.…
$ Securities.Account.0 <dbl> -2.7998134, 0.3569627, 0.3569627, 0.356…
$ Securities.Account.1 <dbl> 2.7998134, -0.3569627, -0.3569627, -0.3…
$ CD.Account.0         <dbl> 0.2613337, 0.2613337, 0.2613337, 0.2613…
$ CD.Account.1         <dbl> -0.2613337, -0.2613337, -0.2613337, -0.…
$ Online.0             <dbl> 1.2486195, 1.2486195, 1.2486195, 1.2486…
$ Online.1             <dbl> -1.2486195, -1.2486195, -1.2486195, -1.…
$ CreditCard.0         <dbl> 0.6408777, 0.6408777, 0.6408777, -1.559…
$ CreditCard.1         <dbl> -0.6408777, -0.6408777, -0.6408777, 1.5…
glimpse(UB.ted)                                             # 데이터 구조 확인
Rows: 749
Columns: 22
$ Personal.Loan        <fct> no, no, no, no, no, no, no, no, no, no,…
$ Age                  <dbl> -1.7881612, -0.7478521, 1.2460737, 0.81…
$ Experience           <dbl> -1.68358012, -0.64236200, 0.83269699, 0…
$ Income               <dbl> -0.53400522, -0.96689556, -1.11840718, …
$ ZIP.Code             <dbl> -1.17304370, -0.59585545, 1.07366441, 0…
$ Family.1             <dbl> -0.6355621, -0.6355621, 1.5725118, 1.57…
$ Family.2             <dbl> -0.5774051, -0.5774051, -0.5774051, -0.…
$ Family.3             <dbl> -0.4987865, -0.4987865, -0.4987865, -0.…
$ Family.4             <dbl> 1.6747892, 1.6747892, -0.5967491, -0.59…
$ CCAvg                <dbl> -0.19512759, -0.86789083, -0.25119120, …
$ Education.1          <dbl> 1.1482386, -0.8704018, -0.8704018, -0.8…
$ Education.2          <dbl> -0.6196534, 1.6128838, -0.6196534, 1.61…
$ Education.3          <dbl> -0.6408777, -0.6408777, 1.5594690, -0.6…
$ Mortgage             <dbl> -0.5664192, 0.9609885, -0.5664192, -0.5…
$ Securities.Account.0 <dbl> -2.7998134, 0.3569627, 0.3569627, -2.79…
$ Securities.Account.1 <dbl> 2.7998134, -0.3569627, -0.3569627, 2.79…
$ CD.Account.0         <dbl> 0.2613337, 0.2613337, 0.2613337, 0.2613…
$ CD.Account.1         <dbl> -0.2613337, -0.2613337, -0.2613337, -0.…
$ Online.0             <dbl> 1.2486195, -0.8004271, -0.8004271, 1.24…
$ Online.1             <dbl> -1.2486195, 0.8004271, 0.8004271, -1.24…
$ CreditCard.0         <dbl> 0.6408777, 0.6408777, -1.5594690, -1.55…
$ CreditCard.1         <dbl> -0.6408777, -0.6408777, 1.5594690, 1.55…

6. 모형 훈련

Package "e1071"는 Support Vector Machine을 효율적으로 구현할 수 있는 “libsvm”을 R에서 사용할 수 있도록 만든 Package이며, 함수 svm()을 이용하여 Support Vector Machine을 수행할 수 있다. 함수에서 사용할 수 있는 자세한 옵션은 여기를 참고한다.

svm(formula, data, kernel, cost, gamma, probability, ...)
svm.model.rd <- svm(Personal.Loan ~.,     
                    data = UB.trd,  
                    kernel = "radial", 
                    cost = 1,              
                    gamma = 2,
                    probability = TRUE)

summary(svm.model.rd)

Call:
svm(formula = Personal.Loan ~ ., data = UB.trd, kernel = "radial", 
    cost = 1, gamma = 2, probability = TRUE)


Parameters:
   SVM-Type:  C-classification 
 SVM-Kernel:  radial 
       cost:  1 

Number of Support Vectors:  1729

 ( 1549 180 )


Number of Classes:  2 

Levels: 
 no yes

Result! Number of Support Vectors는 결정경계와 가까이 위치한 case의 수이다. 해당 데이터에서는 총 1729개의 case로, "Personal.Loan = no"에 해당하는 case는 1549개, "Personal.Loan = yes"에 해당하는 case는 180개이다. case의 행 번호는 svm.model.rd$index를 이용하여 확인할 수 있다.

# Support Vector Index
svm.model.rd$index   
   [1]    1    2    3    4    5    6    7    9   10   11   12   13
  [13]   15   17   18   19   20   21   22   23   25   26   27   28
  [25]   29   30   31   33   35   36   37   39   40   41   42   44
  [37]   45   47   48   49   50   51   52   53   54   55   56   57
  [49]   58   59   60   61   62   63   64   65   66   67   68   70
  [61]   71   72   73   74   75   76   77   78   79   80   81   82
  [73]   83   84   85   86   87   88   89   90   91   92   93   94
  [85]   95   97   98   99  101  102  103  104  105  106  107  108
  [97]  109  111  112  113  114  116  117  118  119  120  121  122
 [109]  123  124  125  126  127  129  131  132  133  134  135  136
 [121]  137  138  139  140  141  142  144  145  146  147  148  149
 [133]  150  151  152  153  154  155  156  157  158  159  160  161
 [145]  162  163  164  165  166  167  169  170  171  173  174  175
 [157]  176  177  178  179  180  181  182  183  184  185  186  187
 [169]  188  189  190  191  192  193  194  195  196  197  198  199
 [181]  201  202  203  205  206  208  209  211  212  213  214  215
 [193]  216  217  218  221  222  223  227  228  230  231  232  233
 [205]  234  235  237  238  239  240  241  242  243  244  247  249
 [217]  250  251  252  253  254  255  256  257  258  259  260  261
 [229]  262  263  264  265  266  267  269  270  271  272  273  274
 [241]  275  276  278  279  280  281  282  283  284  285  286  287
 [253]  288  289  290  292  294  295  296  297  298  299  300  301
 [265]  302  303  304  306  307  309  310  311  312  313  314  315
 [277]  316  317  318  319  320  321  322  323  324  325  328  329
 [289]  330  331  332  333  335  337  338  339  340  341  342  343
 [301]  344  345  346  347  348  349  350  351  352  353  354  355
 [313]  356  357  358  360  361  362  363  364  365  366  367  368
 [325]  369  370  371  372  373  374  376  377  378  379  381  382
 [337]  383  384  385  386  387  388  389  390  391  392  393  394
 [349]  395  396  397  399  400  401  403  404  405  406  407  408
 [361]  409  410  411  412  413  414  415  416  417  418  419  420
 [373]  421  422  423  424  425  426  427  428  429  430  431  432
 [385]  433  434  435  437  438  439  440  443  444  445  446  447
 [397]  448  449  450  451  452  453  454  456  457  459  460  461
 [409]  462  464  465  466  467  468  469  470  471  473  474  475
 [421]  476  477  478  479  480  481  482  483  484  485  486  487
 [433]  488  489  490  491  492  494  495  496  497  498  499  500
 [445]  501  502  503  504  505  506  507  508  510  511  512  513
 [457]  514  515  516  519  520  521  522  523  524  525  526  527
 [469]  528  529  530  531  532  533  534  535  536  537  538  540
 [481]  542  544  545  546  548  549  552  553  554  555  556  557
 [493]  558  559  560  561  562  564  565  566  568  569  570  571
 [505]  572  573  574  575  576  577  578  579  580  581  582  584
 [517]  585  586  587  588  589  590  591  592  593  594  595  596
 [529]  597  598  599  600  601  602  603  604  605  606  607  608
 [541]  609  610  611  612  613  614  615  616  617  618  620  621
 [553]  623  624  625  626  627  628  630  631  632  633  634  635
 [565]  636  637  638  639  640  641  643  644  645  646  647  648
 [577]  649  650  651  652  653  654  655  656  657  658  659  660
 [589]  661  662  665  666  667  668  669  671  673  674  675  676
 [601]  677  678  680  681  682  683  685  686  687  688  689  690
 [613]  692  693  694  695  696  697  698  699  700  701  702  703
 [625]  704  705  706  707  709  710  711  713  714  716  717  718
 [637]  719  721  722  723  724  725  726  728  729  730  731  732
 [649]  733  734  735  736  737  738  739  740  743  746  747  748
 [661]  749  750  752  753  754  755  756  757  758  759  760  761
 [673]  762  763  764  765  766  767  768  769  771  772  773  774
 [685]  775  776  777  778  779  780  781  784  785  786  787  788
 [697]  789  791  792  793  795  797  798  799  800  801  802  803
 [709]  804  805  806  807  808  810  811  812  813  814  815  817
 [721]  818  819  820  821  822  823  824  825  826  827  828  830
 [733]  832  833  835  836  837  838  839  840  841  842  843  844
 [745]  845  846  847  848  849  850  851  853  854  855  856  857
 [757]  858  859  860  862  863  864  865  866  867  868  869  870
 [769]  871  872  873  874  875  876  878  879  880  881  882  883
 [781]  884  885  886  888  889  890  891  892  893  894  895  896
 [793]  897  898  899  900  901  902  903  904  905  906  907  908
 [805]  910  911  912  914  915  917  918  919  920  921  922  923
 [817]  925  926  927  928  929  930  931  933  934  935  936  937
 [829]  938  939  940  941  942  943  944  945  946  947  948  949
 [841]  950  953  954  955  956  957  958  959  960  961  962  963
 [853]  964  965  966  967  968  970  971  972  974  975  977  980
 [865]  981  983  984  985  986  987  988  989  990  991  992  993
 [877]  994  995  996  997  998  999 1000 1001 1002 1003 1004 1005
 [889] 1006 1008 1009 1011 1012 1013 1014 1015 1016 1017 1018 1019
 [901] 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031
 [913] 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1043 1044
 [925] 1046 1047 1048 1050 1051 1052 1053 1054 1055 1056 1057 1058
 [937] 1060 1061 1062 1063 1065 1066 1067 1068 1069 1070 1071 1072
 [949] 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085
 [961] 1086 1087 1088 1089 1091 1092 1093 1094 1095 1096 1097 1098
 [973] 1099 1101 1102 1104 1105 1106 1108 1109 1111 1113 1114 1115
 [985] 1116 1117 1118 1119 1120 1121 1123 1124 1125 1126 1127 1128
 [997] 1129 1130 1132 1134 1135 1136 1137 1139 1140 1141 1143 1144
[1009] 1145 1147 1148 1149 1150 1151 1152 1153 1156 1157 1158 1159
[1021] 1160 1162 1163 1164 1165 1167 1168 1169 1171 1173 1174 1176
[1033] 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188
[1045] 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200
[1057] 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212
[1069] 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225
[1081] 1226 1227 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238
[1093] 1239 1240 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251
[1105] 1252 1254 1255 1256 1258 1259 1261 1263 1264 1265 1266 1267
[1117] 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279
[1129] 1280 1282 1285 1286 1287 1288 1289 1291 1292 1293 1294 1295
[1141] 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307
[1153] 1309 1310 1311 1312 1313 1314 1315 1316 1318 1319 1320 1321
[1165] 1322 1323 1324 1325 1327 1328 1329 1330 1331 1332 1333 1334
[1177] 1335 1336 1337 1338 1339 1340 1341 1343 1344 1346 1347 1348
[1189] 1349 1350 1351 1352 1353 1354 1355 1356 1357 1359 1361 1362
[1201] 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374
[1213] 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386
[1225] 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398
[1237] 1399 1400 1401 1402 1404 1405 1406 1407 1408 1409 1410 1411
[1249] 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424
[1261] 1426 1427 1428 1430 1431 1432 1433 1434 1435 1436 1437 1439
[1273] 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1451 1452
[1285] 1453 1454 1456 1457 1459 1460 1461 1462 1463 1465 1466 1467
[1297] 1468 1469 1470 1472 1473 1474 1475 1476 1477 1478 1479 1480
[1309] 1481 1482 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493
[1321] 1496 1497 1498 1499 1500 1501 1502 1503 1506 1507 1508 1509
[1333] 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521
[1345] 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1534
[1357] 1535 1536 1537 1538 1539 1541 1542 1543 1544 1545 1546 1548
[1369] 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561
[1381] 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1574
[1393] 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1587 1588
[1405] 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601
[1417] 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613
[1429] 1614 1615 1616 1617 1618 1619 1620 1621 1623 1624 1625 1626
[1441] 1627 1628 1629 1631 1632 1633 1634 1635 1636 1637 1638 1639
[1453] 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1652
[1465] 1653 1656 1657 1658 1659 1660 1661 1662 1663 1664 1666 1667
[1477] 1668 1669 1670 1671 1672 1673 1675 1676 1677 1678 1679 1681
[1489] 1682 1683 1684 1685 1686 1687 1688 1690 1691 1692 1693 1694
[1501] 1695 1696 1698 1699 1700 1701 1702 1704 1705 1706 1707 1710
[1513] 1711 1712 1714 1715 1716 1717 1718 1721 1722 1723 1724 1725
[1525] 1726 1727 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738
[1537] 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750
[1549] 1751    8   14   16   24   32   34   38   43   46   69   96
[1561]  110  115  128  130  143  168  172  200  207  210  219  220
[1573]  224  225  226  245  246  248  268  277  293  305  308  326
[1585]  327  334  336  359  375  380  398  402  436  442  455  458
[1597]  463  472  493  509  517  539  541  543  547  550  551  563
[1609]  567  583  619  622  629  642  663  664  670  672  679  684
[1621]  691  708  712  715  720  727  742  744  745  751  770  782
[1633]  783  790  794  796  809  816  829  834  861  877  887  909
[1645]  916  924  932  951  952  969  973  976  978  979  982 1007
[1657] 1042 1045 1049 1059 1064 1090 1100 1103 1107 1110 1112 1122
[1669] 1131 1138 1142 1146 1154 1155 1161 1166 1170 1172 1175 1213
[1681] 1228 1241 1253 1257 1260 1262 1281 1283 1284 1290 1308 1317
[1693] 1326 1342 1345 1358 1360 1403 1412 1425 1429 1438 1450 1458
[1705] 1464 1471 1495 1504 1505 1533 1540 1547 1549 1573 1585 1586
[1717] 1622 1630 1651 1654 1655 1665 1674 1680 1689 1709 1713 1719
[1729] 1728

7. 모형 평가

Caution! 모형 평가를 위해 Test Dataset에 대한 예측 class/확률 이 필요하며, 함수 predict()를 이용하여 생성한다.

# 예측 class 생성 
svm.rd.pred <- predict(svm.model.rd,
                       newdata = UB.ted[,-1],        # Test Dataset including Only 예측 변수   
                       type = "class")               # 예측 class 생성       

svm.rd.pred %>%
  as_tibble
# A tibble: 749 × 1
   value
   <fct>
 1 no   
 2 no   
 3 no   
 4 no   
 5 no   
 6 no   
 7 no   
 8 no   
 9 no   
10 no   
# ℹ 739 more rows


7-1. ConfusionMatrix

CM   <- caret::confusionMatrix(svm.rd.pred, UB.ted$Personal.Loan, 
                               positive = "yes")     # confusionMatrix(예측 class, 실제 class, positive="관심 class")
CM
Confusion Matrix and Statistics

          Reference
Prediction  no yes
       no  673  76
       yes   0   0
                                          
               Accuracy : 0.8985          
                 95% CI : (0.8746, 0.9192)
    No Information Rate : 0.8985          
    P-Value [Acc > NIR] : 0.5305          
                                          
                  Kappa : 0               
                                          
 Mcnemar's Test P-Value : <2e-16          
                                          
            Sensitivity : 0.0000          
            Specificity : 1.0000          
         Pos Pred Value :    NaN          
         Neg Pred Value : 0.8985          
             Prevalence : 0.1015          
         Detection Rate : 0.0000          
   Detection Prevalence : 0.0000          
      Balanced Accuracy : 0.5000          
                                          
       'Positive' Class : yes             
                                          


7-2. ROC 곡선

# 예측 확률 생성
test.svm.prob <- predict(svm.model.rd, 
                         newdata = UB.ted[,-1],      # Test Dataset including Only 예측 변수  
                         probability = TRUE)         # 예측 확률 생성       

attr(test.svm.prob, "probabilities") %>%
  as_tibble
# A tibble: 749 × 2
      no     yes
   <dbl>   <dbl>
 1 0.895 0.105  
 2 0.921 0.0786 
 3 0.995 0.00543
 4 0.841 0.159  
 5 0.951 0.0492 
 6 0.995 0.00489
 7 0.997 0.00270
 8 0.878 0.122  
 9 0.835 0.165  
10 0.998 0.00248
# ℹ 739 more rows
test.svm.prob <- attr(test.svm.prob, "probabilities")[,2]   # "Personal.Loan = yes"에 대한 예측 확률

ac  <- UB.ted$Personal.Loan                                 # Test Dataset의 실제 class 
pp  <- as.numeric(test.svm.prob)                            # 예측 확률을 수치형으로 변환

1) Package “pROC”

pacman::p_load("pROC")

svm.roc  <- roc(ac, pp, plot = T, col = "gray")             # roc(실제 class, 예측 확률)
auc      <- round(auc(svm.roc), 3)
legend("bottomright", legend = auc, bty = "n")

Caution! Package "pROC"를 통해 출력한 ROC 곡선은 다양한 함수를 이용해서 그래프를 수정할 수 있다.

# 함수 plot.roc() 이용
plot.roc(svm.roc,   
         col="gray",                                        # Line Color
         print.auc = TRUE,                                  # AUC 출력 여부
         print.auc.col = "red",                             # AUC 글씨 색깔
         print.thres = TRUE,                                # Cutoff Value 출력 여부
         print.thres.pch = 19,                              # Cutoff Value를 표시하는 도형 모양
         print.thres.col = "red",                           # Cutoff Value를 표시하는 도형의 색깔
         auc.polygon = TRUE,                                # 곡선 아래 면적에 대한 여부
         auc.polygon.col = "gray90")                        # 곡선 아래 면적의 색깔

# 함수 ggroc() 이용
ggroc(svm.roc) +
annotate(geom = "text", x = 0.9, y = 1.0,
label = paste("AUC = ", auc),
size = 5,
color="red") +
theme_bw()

2) Package “Epi”

pacman::p_load("Epi")       
# install_version("etm", version = "1.1", repos = "http://cran.us.r-project.org")

ROC(pp, ac, plot = "ROC")                                   # ROC(예측 확률, 실제 class)  

3) Package “ROCR”

pacman::p_load("ROCR")

svm.pred <- prediction(pp, ac)                              # prediction(예측 확률, 실제 class)    

svm.perf <- performance(svm.pred, "tpr", "fpr")             # performance(, "민감도", "1-특이도")                      
plot(svm.perf, col = "gray")                                # ROC Curve

perf.auc   <- performance(svm.pred, "auc")                  # AUC
auc        <- attributes(perf.auc)$y.values 
legend("bottomright", legend = auc, bty = "n")


7-3. 향상 차트

1) Package “ROCR”

svm.perf <- performance(svm.pred, "lift", "rpp")            # Lift Chart
plot(svm.perf, main = "lift curve", 
     colorize = T,                                          # Coloring according to cutoff
     lwd = 2)  

Reuse

Text and figures are licensed under Creative Commons Attribution CC BY 4.0. The figures that have been reused from other sources don't fall under this license and can be recognized by a note in their caption: "Figure from ...".