Python을 이용한 SVM (2)

요구사항

이 과제는 제가 anomaly detection 데이터셋(캐글에 올라와있음)을 드립니다. 이것은 해당 결제가 사기인지 아닌지 판별하는 데이터셋이며, 실습코드를 활용하고, 또 본인이 여태 배운 내용을 활용하여 자유롭게 데이터를 가지고 연습해주시면 됩니다.
다만 이 데이터셋은 굉장히 imbalance한 데이터 셋입니다. 실제 사기를 치는 사례가 많지 않으므로 사기인 경우가 전체에 0.17 프로밖에 되지 않습니다. 따라서, 그냥 데이터를 트레이닝 시키면 무조건 사기가 아니라고 판별해 버릴 가능성이 높습니다. 또한, 그대로 트레이닝을 돌리게 되면 엄청나게 많은 데이터 양 때문에 트레이닝조차 힘들 것입니다. 그런데 실제로 정확도를 높이면서 트레이닝을 할 수 있는 방법이 있으니 고민해보세요.

우수과제 선정이유

confusion matrix를 쓴 이유 등을 구체적으로 적어줬고, weight를 줘보는 방법(원래 의도했던 방법)을 시도해보고 하이퍼 파라미터 튜닝도 exponential하게 함으로써 좋은 parameter를 잘 찾는 등 전반적으로 깔끔하고 군더더기없이 과제를 진행해 우수과제로 선정되었습니다.

creditcard

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.svm import SVC
from sklearn.model_selection import train_test_split
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
import matplotlib.pyplot as plt
from sklearn.svm import SVC
from sklearn.model_selection import train_test_split
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn import metrics
from sklearn.model_selection import GridSearchCV, KFold

In [2]:

# Anomaly detection(사기감지 데이터) 로드
data = pd.read_csv('creditcard.csv')
print(data.columns)

Index(['Time', 'V1', 'V2', 'V3', 'V4', 'V5', 'V6', 'V7', 'V8', 'V9', 'V10',
       'V11', 'V12', 'V13', 'V14', 'V15', 'V16', 'V17', 'V18', 'V19', 'V20',
       'V21', 'V22', 'V23', 'V24', 'V25', 'V26', 'V27', 'V28', 'Amount',
       'Class'],
      dtype='object')

In [3]:

X = np.array(data.loc[:, data.columns != 'Class'].values)
y = np.array(data.Class)

In [4]:

# Train, Test 셋 나누기
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.95, random_state=0)

In [5]:

X_train.shape, X_test.shape, y_train.shape, y_test.shape

Out[5]:

((14240, 30), (270566, 30), (14240,), (270566,))

In [6]:

scaler = StandardScaler()
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)

In [7]:

X_test

Out[7]:

array([[ 0.64698237, -0.17516838,  0.67679542, ...,  0.28359773,
         0.53175665, -0.21321296],
       [ 1.30680833, -0.18907225,  0.59652865, ...,  0.19973923,
         0.57913126, -0.38379471],
       [ 1.20674431, -0.85567843, -1.55048369, ...,  0.75445011,
         1.43893504,  0.03803843],
       ...,
       [ 1.19643125, -0.17022017,  0.67311331, ..., -1.53817237,
        -0.40964519, -0.34175497],
       [ 0.80608919,  1.0705119 , -0.04942977, ..., -0.12699757,
        -0.23548545, -0.38819161],
       [-0.30377277,  0.5088799 , -0.01341662, ...,  0.14715975,
         0.02807013, -0.21074532]])

In [8]:

y_train.sum()

Out[8]:

In [9]:

from sklearn.metrics.pairwise import rbf_kernel

In [10]:

import timeit
start = timeit.default_timer()

svc=SVC(kernel='linear', C = 100)
svc.fit(X_train, y_train)

stop = timeit.default_timer()
print(stop - start)

8.0188527

In [11]:

y_pred = svc.predict(X_test)
metrics.confusion_matrix(y_test, y_pred)

Out[11]:

array([[270037,     63],
       [   113,    353]], dtype=int64)

In [12]:

print('Accuracy Score:')
print(metrics.accuracy_score(y_test,y_pred))
print('Recall Score:')
print(metrics.recall_score(y_test, y_pred))
print('Precision Score:')
print(metrics.precision_score(y_test, y_pred))

Accuracy Score:
0.9993495117642276
Recall Score:
0.7575107296137339
Precision Score:
0.8485576923076923

Weight

svc의 class_weight를 통해 적게 관찰되는 경우에 weight를 준다

In [13]:

svc_weight=SVC(kernel='linear', C = 100, class_weight='balanced')
svc_weight.fit(X_train,y_train)

Out[13]:

SVC(C=100, break_ties=False, cache_size=200, class_weight='balanced', coef0=0.0,
    decision_function_shape='ovr', degree=3, gamma='scale', kernel='linear',
    max_iter=-1, probability=False, random_state=None, shrinking=True,
    tol=0.001, verbose=False)

In [14]:

y_pred_weight = svc_weight.predict(X_test)
metrics.confusion_matrix(y_test, y_pred_weight)

Out[14]:

array([[269766,    334],
       [   129,    337]], dtype=int64)

In [15]:

print('Accuracy Score:')
print(metrics.accuracy_score(y_test,y_pred_weight))
print('Recall Score:')
print(metrics.recall_score(y_test, y_pred_weight))
print('Precision Score:')
print(metrics.precision_score(y_test, y_pred_weight))

Accuracy Score:
0.9982887724252123
Recall Score:
0.723175965665236
Precision Score:
0.5022354694485842

RandomUnderSampler

RandomUnderSampler를 통해 0, 1 사이의 대칭을 맞춰준다.

In [16]:

# Train, Test 셋 나누기
X_train2, X_test2, y_train2, y_test2 = train_test_split(X, y, test_size = 0.05, random_state=0)

In [17]:

from imblearn.under_sampling import RandomUnderSampler

Xresampled, yresampled = RandomUnderSampler(random_state=0).fit_sample(X_train2, y_train2)

In [18]:

Xresampled.shape, X_test2.shape, yresampled.shape, y_test2.shape

Out[18]:

((920, 30), (14241, 30), (920,), (14241,))

In [19]:

Xresampled = scaler.transform(Xresampled)
X_test2 = scaler.transform(X_test2)

In [20]:

svc_rus=SVC(kernel='linear', C = 100)
svc_rus.fit(Xresampled, yresampled)

Out[20]:

SVC(C=100, break_ties=False, cache_size=200, class_weight=None, coef0=0.0,
    decision_function_shape='ovr', degree=3, gamma='scale', kernel='linear',
    max_iter=-1, probability=False, random_state=None, shrinking=True,
    tol=0.001, verbose=False)

In [21]:

y_pred_rus = svc_rus.predict(X_test2)
metrics.confusion_matrix(y_test2, y_pred_rus)

Out[21]:

array([[13582,   627],
       [    2,    30]], dtype=int64)

In [22]:

print('Accuracy Score:')
print(metrics.accuracy_score(y_test2, y_pred_rus))
print('Recall Score:')
print(metrics.recall_score(y_test2, y_pred_rus))
print('Precision Score:')
print(metrics.precision_score(y_test2, y_pred_rus))

Accuracy Score:
0.9558317533881048
Recall Score:
0.9375
Precision Score:
0.045662100456621

In [23]:

svc_rus_weight=SVC(kernel='linear', C = 100, class_weight='balanced')
svc_rus_weight.fit(Xresampled, yresampled)

Out[23]:

SVC(C=100, break_ties=False, cache_size=200, class_weight='balanced', coef0=0.0,
    decision_function_shape='ovr', degree=3, gamma='scale', kernel='linear',
    max_iter=-1, probability=False, random_state=None, shrinking=True,
    tol=0.001, verbose=False)

Undersampling으로 인해 weight의 여부가 결과에 변화를 주지 못했다.In [24]:

y_pred_rus_wieght = svc_rus_weight.predict(X_test2)
metrics.confusion_matrix(y_test2, y_pred_rus_wieght)

Out[24]:

array([[13582,   627],
       [    2,    30]], dtype=int64)

In [25]:

print('Accuracy Score:')
print(metrics.accuracy_score(y_test2, y_pred_rus_wieght))
print('Recall Score:')
print(metrics.recall_score(y_test2, y_pred_rus_wieght))
print('Precision Score:')
print(metrics.precision_score(y_test2, y_pred_rus_wieght))

Accuracy Score:
0.9558317533881048
Recall Score:
0.9375
Precision Score:
0.045662100456621

In [26]:

values = [0.001, 0.01, 0.1, 1, 10, 100]
param_grid = [{'kernel':['rbf'], 'C':values, 'gamma':values},
               {'kernel':['linear'], 'C':values}]
gs = GridSearchCV(SVC(), param_grid, cv= KFold(n_splits=5), scoring='accuracy', verbose = 3)
gs.fit(Xresampled, yresampled)

Fitting 5 folds for each of 42 candidates, totalling 210 fits
[CV] C=0.001, gamma=0.001, kernel=rbf ................................
[CV] .... C=0.001, gamma=0.001, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=0.001, kernel=rbf ................................
[CV] .... C=0.001, gamma=0.001, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=0.001, kernel=rbf ................................
[CV] .... C=0.001, gamma=0.001, kernel=rbf, score=0.755, total=   0.0s
[CV] C=0.001, gamma=0.001, kernel=rbf ................................
[CV] .... C=0.001, gamma=0.001, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=0.001, kernel=rbf ................................
[CV] .... C=0.001, gamma=0.001, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=0.01, kernel=rbf .................................
[CV] ..... C=0.001, gamma=0.01, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=0.01, kernel=rbf .................................
[CV] ..... C=0.001, gamma=0.01, kernel=rbf, score=0.000, total=   0.0s

[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.
[Parallel(n_jobs=1)]: Done   1 out of   1 | elapsed:    0.0s remaining:    0.0s
[Parallel(n_jobs=1)]: Done   2 out of   2 | elapsed:    0.0s remaining:    0.0s

[CV] C=0.001, gamma=0.01, kernel=rbf .................................
[CV] ..... C=0.001, gamma=0.01, kernel=rbf, score=0.913, total=   0.0s
[CV] C=0.001, gamma=0.01, kernel=rbf .................................
[CV] ..... C=0.001, gamma=0.01, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=0.01, kernel=rbf .................................
[CV] ..... C=0.001, gamma=0.01, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=0.1, kernel=rbf ..................................
[CV] ...... C=0.001, gamma=0.1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=0.1, kernel=rbf ..................................
[CV] ...... C=0.001, gamma=0.1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=0.1, kernel=rbf ..................................
[CV] ...... C=0.001, gamma=0.1, kernel=rbf, score=0.641, total=   0.0s
[CV] C=0.001, gamma=0.1, kernel=rbf ..................................
[CV] ...... C=0.001, gamma=0.1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=0.1, kernel=rbf ..................................
[CV] ...... C=0.001, gamma=0.1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=1, kernel=rbf ....................................
[CV] ........ C=0.001, gamma=1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=1, kernel=rbf ....................................
[CV] ........ C=0.001, gamma=1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=1, kernel=rbf ....................................
[CV] ........ C=0.001, gamma=1, kernel=rbf, score=0.500, total=   0.0s
[CV] C=0.001, gamma=1, kernel=rbf ....................................
[CV] ........ C=0.001, gamma=1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=1, kernel=rbf ....................................
[CV] ........ C=0.001, gamma=1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=10, kernel=rbf ...................................
[CV] ....... C=0.001, gamma=10, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.001, gamma=10, kernel=rbf ...................................
[CV] ....... C=0.001, gamma=10, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=10, kernel=rbf ...................................
[CV] ....... C=0.001, gamma=10, kernel=rbf, score=0.500, total=   0.1s
[CV] C=0.001, gamma=10, kernel=rbf ...................................
[CV] ....... C=0.001, gamma=10, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=10, kernel=rbf ...................................
[CV] ....... C=0.001, gamma=10, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.001, gamma=100, kernel=rbf ..................................
[CV] ...... C=0.001, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.001, gamma=100, kernel=rbf ..................................
[CV] ...... C=0.001, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.001, gamma=100, kernel=rbf ..................................
[CV] ...... C=0.001, gamma=100, kernel=rbf, score=0.500, total=   0.1s
[CV] C=0.001, gamma=100, kernel=rbf ..................................
[CV] ...... C=0.001, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.001, gamma=100, kernel=rbf ..................................
[CV] ...... C=0.001, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.01, gamma=0.001, kernel=rbf .................................
[CV] ..... C=0.01, gamma=0.001, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.01, gamma=0.001, kernel=rbf .................................
[CV] ..... C=0.01, gamma=0.001, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.01, gamma=0.001, kernel=rbf .................................
[CV] ..... C=0.01, gamma=0.001, kernel=rbf, score=0.783, total=   0.0s
[CV] C=0.01, gamma=0.001, kernel=rbf .................................
[CV] ..... C=0.01, gamma=0.001, kernel=rbf, score=0.364, total=   0.0s
[CV] C=0.01, gamma=0.001, kernel=rbf .................................
[CV] ..... C=0.01, gamma=0.001, kernel=rbf, score=0.364, total=   0.0s
[CV] C=0.01, gamma=0.01, kernel=rbf ..................................
[CV] ...... C=0.01, gamma=0.01, kernel=rbf, score=0.935, total=   0.0s
[CV] C=0.01, gamma=0.01, kernel=rbf ..................................
[CV] ...... C=0.01, gamma=0.01, kernel=rbf, score=0.913, total=   0.0s
[CV] C=0.01, gamma=0.01, kernel=rbf ..................................
[CV] ...... C=0.01, gamma=0.01, kernel=rbf, score=0.924, total=   0.0s
[CV] C=0.01, gamma=0.01, kernel=rbf ..................................
[CV] ...... C=0.01, gamma=0.01, kernel=rbf, score=0.821, total=   0.0s
[CV] C=0.01, gamma=0.01, kernel=rbf ..................................
[CV] ...... C=0.01, gamma=0.01, kernel=rbf, score=0.821, total=   0.0s
[CV] C=0.01, gamma=0.1, kernel=rbf ...................................
[CV] ....... C=0.01, gamma=0.1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.01, gamma=0.1, kernel=rbf ...................................
[CV] ....... C=0.01, gamma=0.1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.01, gamma=0.1, kernel=rbf ...................................
[CV] ....... C=0.01, gamma=0.1, kernel=rbf, score=0.641, total=   0.0s
[CV] C=0.01, gamma=0.1, kernel=rbf ...................................
[CV] ....... C=0.01, gamma=0.1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.01, gamma=0.1, kernel=rbf ...................................
[CV] ....... C=0.01, gamma=0.1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.01, gamma=1, kernel=rbf .....................................
[CV] ......... C=0.01, gamma=1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.01, gamma=1, kernel=rbf .....................................
[CV] ......... C=0.01, gamma=1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.01, gamma=1, kernel=rbf .....................................
[CV] ......... C=0.01, gamma=1, kernel=rbf, score=0.500, total=   0.0s
[CV] C=0.01, gamma=1, kernel=rbf .....................................
[CV] ......... C=0.01, gamma=1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.01, gamma=1, kernel=rbf .....................................
[CV] ......... C=0.01, gamma=1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.01, gamma=10, kernel=rbf ....................................
[CV] ........ C=0.01, gamma=10, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.01, gamma=10, kernel=rbf ....................................
[CV] ........ C=0.01, gamma=10, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.01, gamma=10, kernel=rbf ....................................
[CV] ........ C=0.01, gamma=10, kernel=rbf, score=0.500, total=   0.1s
[CV] C=0.01, gamma=10, kernel=rbf ....................................
[CV] ........ C=0.01, gamma=10, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.01, gamma=10, kernel=rbf ....................................
[CV] ........ C=0.01, gamma=10, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.01, gamma=100, kernel=rbf ...................................
[CV] ....... C=0.01, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.01, gamma=100, kernel=rbf ...................................
[CV] ....... C=0.01, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.01, gamma=100, kernel=rbf ...................................
[CV] ....... C=0.01, gamma=100, kernel=rbf, score=0.500, total=   0.1s
[CV] C=0.01, gamma=100, kernel=rbf ...................................
[CV] ....... C=0.01, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.01, gamma=100, kernel=rbf ...................................
[CV] ....... C=0.01, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.1, gamma=0.001, kernel=rbf ..................................
[CV] ...... C=0.1, gamma=0.001, kernel=rbf, score=1.000, total=   0.0s
[CV] C=0.1, gamma=0.001, kernel=rbf ..................................
[CV] ...... C=0.1, gamma=0.001, kernel=rbf, score=1.000, total=   0.0s
[CV] C=0.1, gamma=0.001, kernel=rbf ..................................
[CV] ...... C=0.1, gamma=0.001, kernel=rbf, score=0.918, total=   0.0s
[CV] C=0.1, gamma=0.001, kernel=rbf ..................................
[CV] ...... C=0.1, gamma=0.001, kernel=rbf, score=0.788, total=   0.0s
[CV] C=0.1, gamma=0.001, kernel=rbf ..................................
[CV] ...... C=0.1, gamma=0.001, kernel=rbf, score=0.799, total=   0.0s
[CV] C=0.1, gamma=0.01, kernel=rbf ...................................
[CV] ....... C=0.1, gamma=0.01, kernel=rbf, score=0.962, total=   0.0s
[CV] C=0.1, gamma=0.01, kernel=rbf ...................................
[CV] ....... C=0.1, gamma=0.01, kernel=rbf, score=0.978, total=   0.0s
[CV] C=0.1, gamma=0.01, kernel=rbf ...................................
[CV] ....... C=0.1, gamma=0.01, kernel=rbf, score=0.929, total=   0.0s
[CV] C=0.1, gamma=0.01, kernel=rbf ...................................
[CV] ....... C=0.1, gamma=0.01, kernel=rbf, score=0.864, total=   0.0s
[CV] C=0.1, gamma=0.01, kernel=rbf ...................................
[CV] ....... C=0.1, gamma=0.01, kernel=rbf, score=0.870, total=   0.0s
[CV] C=0.1, gamma=0.1, kernel=rbf ....................................
[CV] ........ C=0.1, gamma=0.1, kernel=rbf, score=0.451, total=   0.0s
[CV] C=0.1, gamma=0.1, kernel=rbf ....................................
[CV] ........ C=0.1, gamma=0.1, kernel=rbf, score=0.424, total=   0.0s
[CV] C=0.1, gamma=0.1, kernel=rbf ....................................
[CV] ........ C=0.1, gamma=0.1, kernel=rbf, score=0.728, total=   0.0s
[CV] C=0.1, gamma=0.1, kernel=rbf ....................................
[CV] ........ C=0.1, gamma=0.1, kernel=rbf, score=0.016, total=   0.0s
[CV] C=0.1, gamma=0.1, kernel=rbf ....................................
[CV] ........ C=0.1, gamma=0.1, kernel=rbf, score=0.016, total=   0.0s
[CV] C=0.1, gamma=1, kernel=rbf ......................................
[CV] .......... C=0.1, gamma=1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.1, gamma=1, kernel=rbf ......................................
[CV] .......... C=0.1, gamma=1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.1, gamma=1, kernel=rbf ......................................
[CV] .......... C=0.1, gamma=1, kernel=rbf, score=0.500, total=   0.0s
[CV] C=0.1, gamma=1, kernel=rbf ......................................
[CV] .......... C=0.1, gamma=1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.1, gamma=1, kernel=rbf ......................................
[CV] .......... C=0.1, gamma=1, kernel=rbf, score=0.000, total=   0.0s
[CV] C=0.1, gamma=10, kernel=rbf .....................................
[CV] ......... C=0.1, gamma=10, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.1, gamma=10, kernel=rbf .....................................
[CV] ......... C=0.1, gamma=10, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.1, gamma=10, kernel=rbf .....................................
[CV] ......... C=0.1, gamma=10, kernel=rbf, score=0.500, total=   0.1s
[CV] C=0.1, gamma=10, kernel=rbf .....................................
[CV] ......... C=0.1, gamma=10, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.1, gamma=10, kernel=rbf .....................................
[CV] ......... C=0.1, gamma=10, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.1, gamma=100, kernel=rbf ....................................
[CV] ........ C=0.1, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.1, gamma=100, kernel=rbf ....................................
[CV] ........ C=0.1, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.1, gamma=100, kernel=rbf ....................................
[CV] ........ C=0.1, gamma=100, kernel=rbf, score=0.500, total=   0.1s
[CV] C=0.1, gamma=100, kernel=rbf ....................................
[CV] ........ C=0.1, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=0.1, gamma=100, kernel=rbf ....................................
[CV] ........ C=0.1, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=1, gamma=0.001, kernel=rbf ....................................
[CV] ........ C=1, gamma=0.001, kernel=rbf, score=0.989, total=   0.0s
[CV] C=1, gamma=0.001, kernel=rbf ....................................
[CV] ........ C=1, gamma=0.001, kernel=rbf, score=0.995, total=   0.0s
[CV] C=1, gamma=0.001, kernel=rbf ....................................
[CV] ........ C=1, gamma=0.001, kernel=rbf, score=0.929, total=   0.0s
[CV] C=1, gamma=0.001, kernel=rbf ....................................
[CV] ........ C=1, gamma=0.001, kernel=rbf, score=0.848, total=   0.0s
[CV] C=1, gamma=0.001, kernel=rbf ....................................
[CV] ........ C=1, gamma=0.001, kernel=rbf, score=0.848, total=   0.0s
[CV] C=1, gamma=0.01, kernel=rbf .....................................
[CV] ......... C=1, gamma=0.01, kernel=rbf, score=0.967, total=   0.0s
[CV] C=1, gamma=0.01, kernel=rbf .....................................
[CV] ......... C=1, gamma=0.01, kernel=rbf, score=0.984, total=   0.0s
[CV] C=1, gamma=0.01, kernel=rbf .....................................
[CV] ......... C=1, gamma=0.01, kernel=rbf, score=0.935, total=   0.0s
[CV] C=1, gamma=0.01, kernel=rbf .....................................
[CV] ......... C=1, gamma=0.01, kernel=rbf, score=0.891, total=   0.0s
[CV] C=1, gamma=0.01, kernel=rbf .....................................
[CV] ......... C=1, gamma=0.01, kernel=rbf, score=0.897, total=   0.0s
[CV] C=1, gamma=0.1, kernel=rbf ......................................
[CV] .......... C=1, gamma=0.1, kernel=rbf, score=0.783, total=   0.0s
[CV] C=1, gamma=0.1, kernel=rbf ......................................
[CV] .......... C=1, gamma=0.1, kernel=rbf, score=0.810, total=   0.0s
[CV] C=1, gamma=0.1, kernel=rbf ......................................
[CV] .......... C=1, gamma=0.1, kernel=rbf, score=0.913, total=   0.0s
[CV] C=1, gamma=0.1, kernel=rbf ......................................
[CV] .......... C=1, gamma=0.1, kernel=rbf, score=0.940, total=   0.0s
[CV] C=1, gamma=0.1, kernel=rbf ......................................
[CV] .......... C=1, gamma=0.1, kernel=rbf, score=0.935, total=   0.0s
[CV] C=1, gamma=1, kernel=rbf ........................................
[CV] ............ C=1, gamma=1, kernel=rbf, score=0.049, total=   0.0s
[CV] C=1, gamma=1, kernel=rbf ........................................
[CV] ............ C=1, gamma=1, kernel=rbf, score=0.038, total=   0.0s
[CV] C=1, gamma=1, kernel=rbf ........................................
[CV] ............ C=1, gamma=1, kernel=rbf, score=0.598, total=   0.0s
[CV] C=1, gamma=1, kernel=rbf ........................................
[CV] ............ C=1, gamma=1, kernel=rbf, score=0.098, total=   0.0s
[CV] C=1, gamma=1, kernel=rbf ........................................
[CV] ............ C=1, gamma=1, kernel=rbf, score=0.103, total=   0.0s
[CV] C=1, gamma=10, kernel=rbf .......................................
[CV] ........... C=1, gamma=10, kernel=rbf, score=0.005, total=   0.1s
[CV] C=1, gamma=10, kernel=rbf .......................................
[CV] ........... C=1, gamma=10, kernel=rbf, score=0.000, total=   0.1s
[CV] C=1, gamma=10, kernel=rbf .......................................
[CV] ........... C=1, gamma=10, kernel=rbf, score=0.527, total=   0.1s
[CV] C=1, gamma=10, kernel=rbf .......................................
[CV] ........... C=1, gamma=10, kernel=rbf, score=0.054, total=   0.1s
[CV] C=1, gamma=10, kernel=rbf .......................................
[CV] ........... C=1, gamma=10, kernel=rbf, score=0.038, total=   0.1s
[CV] C=1, gamma=100, kernel=rbf ......................................
[CV] .......... C=1, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=1, gamma=100, kernel=rbf ......................................
[CV] .......... C=1, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=1, gamma=100, kernel=rbf ......................................
[CV] .......... C=1, gamma=100, kernel=rbf, score=0.516, total=   0.1s
[CV] C=1, gamma=100, kernel=rbf ......................................
[CV] .......... C=1, gamma=100, kernel=rbf, score=0.054, total=   0.1s
[CV] C=1, gamma=100, kernel=rbf ......................................
[CV] .......... C=1, gamma=100, kernel=rbf, score=0.038, total=   0.1s
[CV] C=10, gamma=0.001, kernel=rbf ...................................
[CV] ....... C=10, gamma=0.001, kernel=rbf, score=0.978, total=   0.0s
[CV] C=10, gamma=0.001, kernel=rbf ...................................
[CV] ....... C=10, gamma=0.001, kernel=rbf, score=1.000, total=   0.0s
[CV] C=10, gamma=0.001, kernel=rbf ...................................
[CV] ....... C=10, gamma=0.001, kernel=rbf, score=0.951, total=   0.0s
[CV] C=10, gamma=0.001, kernel=rbf ...................................
[CV] ....... C=10, gamma=0.001, kernel=rbf, score=0.891, total=   0.0s
[CV] C=10, gamma=0.001, kernel=rbf ...................................
[CV] ....... C=10, gamma=0.001, kernel=rbf, score=0.886, total=   0.0s
[CV] C=10, gamma=0.01, kernel=rbf ....................................
[CV] ........ C=10, gamma=0.01, kernel=rbf, score=0.951, total=   0.0s
[CV] C=10, gamma=0.01, kernel=rbf ....................................
[CV] ........ C=10, gamma=0.01, kernel=rbf, score=0.940, total=   0.0s
[CV] C=10, gamma=0.01, kernel=rbf ....................................
[CV] ........ C=10, gamma=0.01, kernel=rbf, score=0.951, total=   0.0s
[CV] C=10, gamma=0.01, kernel=rbf ....................................
[CV] ........ C=10, gamma=0.01, kernel=rbf, score=0.902, total=   0.0s
[CV] C=10, gamma=0.01, kernel=rbf ....................................
[CV] ........ C=10, gamma=0.01, kernel=rbf, score=0.908, total=   0.0s
[CV] C=10, gamma=0.1, kernel=rbf .....................................
[CV] ......... C=10, gamma=0.1, kernel=rbf, score=0.750, total=   0.0s
[CV] C=10, gamma=0.1, kernel=rbf .....................................
[CV] ......... C=10, gamma=0.1, kernel=rbf, score=0.783, total=   0.0s
[CV] C=10, gamma=0.1, kernel=rbf .....................................
[CV] ......... C=10, gamma=0.1, kernel=rbf, score=0.918, total=   0.0s
[CV] C=10, gamma=0.1, kernel=rbf .....................................
[CV] ......... C=10, gamma=0.1, kernel=rbf, score=0.940, total=   0.0s
[CV] C=10, gamma=0.1, kernel=rbf .....................................
[CV] ......... C=10, gamma=0.1, kernel=rbf, score=0.940, total=   0.0s
[CV] C=10, gamma=1, kernel=rbf .......................................
[CV] ........... C=10, gamma=1, kernel=rbf, score=0.060, total=   0.0s
[CV] C=10, gamma=1, kernel=rbf .......................................
[CV] ........... C=10, gamma=1, kernel=rbf, score=0.054, total=   0.0s
[CV] C=10, gamma=1, kernel=rbf .......................................
[CV] ........... C=10, gamma=1, kernel=rbf, score=0.603, total=   0.0s
[CV] C=10, gamma=1, kernel=rbf .......................................
[CV] ........... C=10, gamma=1, kernel=rbf, score=0.114, total=   0.0s
[CV] C=10, gamma=1, kernel=rbf .......................................
[CV] ........... C=10, gamma=1, kernel=rbf, score=0.125, total=   0.0s
[CV] C=10, gamma=10, kernel=rbf ......................................
[CV] .......... C=10, gamma=10, kernel=rbf, score=0.005, total=   0.1s
[CV] C=10, gamma=10, kernel=rbf ......................................
[CV] .......... C=10, gamma=10, kernel=rbf, score=0.000, total=   0.1s
[CV] C=10, gamma=10, kernel=rbf ......................................
[CV] .......... C=10, gamma=10, kernel=rbf, score=0.527, total=   0.1s
[CV] C=10, gamma=10, kernel=rbf ......................................
[CV] .......... C=10, gamma=10, kernel=rbf, score=0.060, total=   0.1s
[CV] C=10, gamma=10, kernel=rbf ......................................
[CV] .......... C=10, gamma=10, kernel=rbf, score=0.038, total=   0.1s
[CV] C=10, gamma=100, kernel=rbf .....................................
[CV] ......... C=10, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=10, gamma=100, kernel=rbf .....................................
[CV] ......... C=10, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=10, gamma=100, kernel=rbf .....................................
[CV] ......... C=10, gamma=100, kernel=rbf, score=0.516, total=   0.1s
[CV] C=10, gamma=100, kernel=rbf .....................................
[CV] ......... C=10, gamma=100, kernel=rbf, score=0.054, total=   0.1s
[CV] C=10, gamma=100, kernel=rbf .....................................
[CV] ......... C=10, gamma=100, kernel=rbf, score=0.038, total=   0.1s
[CV] C=100, gamma=0.001, kernel=rbf ..................................
[CV] ...... C=100, gamma=0.001, kernel=rbf, score=0.962, total=   0.0s
[CV] C=100, gamma=0.001, kernel=rbf ..................................
[CV] ...... C=100, gamma=0.001, kernel=rbf, score=0.973, total=   0.0s
[CV] C=100, gamma=0.001, kernel=rbf ..................................
[CV] ...... C=100, gamma=0.001, kernel=rbf, score=0.957, total=   0.0s
[CV] C=100, gamma=0.001, kernel=rbf ..................................
[CV] ...... C=100, gamma=0.001, kernel=rbf, score=0.897, total=   0.0s
[CV] C=100, gamma=0.001, kernel=rbf ..................................
[CV] ...... C=100, gamma=0.001, kernel=rbf, score=0.913, total=   0.0s
[CV] C=100, gamma=0.01, kernel=rbf ...................................
[CV] ....... C=100, gamma=0.01, kernel=rbf, score=0.908, total=   0.0s
[CV] C=100, gamma=0.01, kernel=rbf ...................................
[CV] ....... C=100, gamma=0.01, kernel=rbf, score=0.886, total=   0.0s
[CV] C=100, gamma=0.01, kernel=rbf ...................................
[CV] ....... C=100, gamma=0.01, kernel=rbf, score=0.946, total=   0.0s
[CV] C=100, gamma=0.01, kernel=rbf ...................................
[CV] ....... C=100, gamma=0.01, kernel=rbf, score=0.935, total=   0.0s
[CV] C=100, gamma=0.01, kernel=rbf ...................................
[CV] ....... C=100, gamma=0.01, kernel=rbf, score=0.913, total=   0.0s
[CV] C=100, gamma=0.1, kernel=rbf ....................................
[CV] ........ C=100, gamma=0.1, kernel=rbf, score=0.750, total=   0.0s
[CV] C=100, gamma=0.1, kernel=rbf ....................................
[CV] ........ C=100, gamma=0.1, kernel=rbf, score=0.783, total=   0.0s
[CV] C=100, gamma=0.1, kernel=rbf ....................................
[CV] ........ C=100, gamma=0.1, kernel=rbf, score=0.924, total=   0.0s
[CV] C=100, gamma=0.1, kernel=rbf ....................................
[CV] ........ C=100, gamma=0.1, kernel=rbf, score=0.940, total=   0.0s
[CV] C=100, gamma=0.1, kernel=rbf ....................................
[CV] ........ C=100, gamma=0.1, kernel=rbf, score=0.940, total=   0.0s
[CV] C=100, gamma=1, kernel=rbf ......................................
[CV] .......... C=100, gamma=1, kernel=rbf, score=0.060, total=   0.0s
[CV] C=100, gamma=1, kernel=rbf ......................................
[CV] .......... C=100, gamma=1, kernel=rbf, score=0.054, total=   0.0s
[CV] C=100, gamma=1, kernel=rbf ......................................
[CV] .......... C=100, gamma=1, kernel=rbf, score=0.603, total=   0.0s
[CV] C=100, gamma=1, kernel=rbf ......................................
[CV] .......... C=100, gamma=1, kernel=rbf, score=0.114, total=   0.0s
[CV] C=100, gamma=1, kernel=rbf ......................................
[CV] .......... C=100, gamma=1, kernel=rbf, score=0.125, total=   0.0s
[CV] C=100, gamma=10, kernel=rbf .....................................
[CV] ......... C=100, gamma=10, kernel=rbf, score=0.005, total=   0.1s
[CV] C=100, gamma=10, kernel=rbf .....................................
[CV] ......... C=100, gamma=10, kernel=rbf, score=0.000, total=   0.1s
[CV] C=100, gamma=10, kernel=rbf .....................................
[CV] ......... C=100, gamma=10, kernel=rbf, score=0.527, total=   0.1s
[CV] C=100, gamma=10, kernel=rbf .....................................
[CV] ......... C=100, gamma=10, kernel=rbf, score=0.060, total=   0.1s
[CV] C=100, gamma=10, kernel=rbf .....................................
[CV] ......... C=100, gamma=10, kernel=rbf, score=0.038, total=   0.1s
[CV] C=100, gamma=100, kernel=rbf ....................................
[CV] ........ C=100, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=100, gamma=100, kernel=rbf ....................................
[CV] ........ C=100, gamma=100, kernel=rbf, score=0.000, total=   0.1s
[CV] C=100, gamma=100, kernel=rbf ....................................
[CV] ........ C=100, gamma=100, kernel=rbf, score=0.516, total=   0.1s
[CV] C=100, gamma=100, kernel=rbf ....................................
[CV] ........ C=100, gamma=100, kernel=rbf, score=0.054, total=   0.1s
[CV] C=100, gamma=100, kernel=rbf ....................................
[CV] ........ C=100, gamma=100, kernel=rbf, score=0.038, total=   0.1s
[CV] C=0.001, kernel=linear ..........................................
[CV] .............. C=0.001, kernel=linear, score=0.995, total=   0.0s
[CV] C=0.001, kernel=linear ..........................................
[CV] .............. C=0.001, kernel=linear, score=1.000, total=   0.0s
[CV] C=0.001, kernel=linear ..........................................
[CV] .............. C=0.001, kernel=linear, score=0.929, total=   0.0s
[CV] C=0.001, kernel=linear ..........................................
[CV] .............. C=0.001, kernel=linear, score=0.815, total=   0.0s
[CV] C=0.001, kernel=linear ..........................................
[CV] .............. C=0.001, kernel=linear, score=0.826, total=   0.0s
[CV] C=0.01, kernel=linear ...........................................
[CV] ............... C=0.01, kernel=linear, score=0.978, total=   0.0s
[CV] C=0.01, kernel=linear ...........................................
[CV] ............... C=0.01, kernel=linear, score=1.000, total=   0.0s
[CV] C=0.01, kernel=linear ...........................................
[CV] ............... C=0.01, kernel=linear, score=0.946, total=   0.0s
[CV] C=0.01, kernel=linear ...........................................
[CV] ............... C=0.01, kernel=linear, score=0.875, total=   0.0s
[CV] C=0.01, kernel=linear ...........................................
[CV] ............... C=0.01, kernel=linear, score=0.875, total=   0.0s
[CV] C=0.1, kernel=linear ............................................
[CV] ................ C=0.1, kernel=linear, score=0.978, total=   0.0s
[CV] C=0.1, kernel=linear ............................................
[CV] ................ C=0.1, kernel=linear, score=0.962, total=   0.0s
[CV] C=0.1, kernel=linear ............................................
[CV] ................ C=0.1, kernel=linear, score=0.957, total=   0.0s
[CV] C=0.1, kernel=linear ............................................
[CV] ................ C=0.1, kernel=linear, score=0.897, total=   0.0s
[CV] C=0.1, kernel=linear ............................................
[CV] ................ C=0.1, kernel=linear, score=0.897, total=   0.0s
[CV] C=1, kernel=linear ..............................................
[CV] .................. C=1, kernel=linear, score=0.957, total=   0.0s
[CV] C=1, kernel=linear ..............................................
[CV] .................. C=1, kernel=linear, score=0.951, total=   0.0s
[CV] C=1, kernel=linear ..............................................
[CV] .................. C=1, kernel=linear, score=0.957, total=   0.0s
[CV] C=1, kernel=linear ..............................................
[CV] .................. C=1, kernel=linear, score=0.897, total=   0.0s
[CV] C=1, kernel=linear ..............................................
[CV] .................. C=1, kernel=linear, score=0.886, total=   0.0s
[CV] C=10, kernel=linear .............................................
[CV] ................. C=10, kernel=linear, score=0.962, total=   0.1s
[CV] C=10, kernel=linear .............................................
[CV] ................. C=10, kernel=linear, score=0.929, total=   0.1s
[CV] C=10, kernel=linear .............................................
[CV] ................. C=10, kernel=linear, score=0.957, total=   0.1s
[CV] C=10, kernel=linear .............................................
[CV] ................. C=10, kernel=linear, score=0.902, total=   0.1s
[CV] C=10, kernel=linear .............................................
[CV] ................. C=10, kernel=linear, score=0.886, total=   0.1s
[CV] C=100, kernel=linear ............................................
[CV] ................ C=100, kernel=linear, score=0.967, total=   0.5s
[CV] C=100, kernel=linear ............................................
[CV] ................ C=100, kernel=linear, score=0.935, total=   0.7s
[CV] C=100, kernel=linear ............................................
[CV] ................ C=100, kernel=linear, score=0.951, total=   0.6s
[CV] C=100, kernel=linear ............................................
[CV] ................ C=100, kernel=linear, score=0.902, total=   1.0s
[CV] C=100, kernel=linear ............................................
[CV] ................ C=100, kernel=linear, score=0.902, total=   0.6s

[Parallel(n_jobs=1)]: Done 210 out of 210 | elapsed:   10.5s finished

Out[26]:

GridSearchCV(cv=KFold(n_splits=5, random_state=None, shuffle=False),
             error_score=nan,
             estimator=SVC(C=1.0, break_ties=False, cache_size=200,
                           class_weight=None, coef0=0.0,
                           decision_function_shape='ovr', degree=3,
                           gamma='scale', kernel='rbf', max_iter=-1,
                           probability=False, random_state=None, shrinking=True,
                           tol=0.001, verbose=False),
             iid='deprecated', n_jobs=None,
             param_grid=[{'C': [0.001, 0.01, 0.1, 1, 10, 100],
                          'gamma': [0.001, 0.01, 0.1, 1, 10, 100],
                          'kernel': ['rbf']},
                         {'C': [0.001, 0.01, 0.1, 1, 10, 100],
                          'kernel': ['linear']}],
             pre_dispatch='2*n_jobs', refit=True, return_train_score=False,
             scoring='accuracy', verbose=3)

In [27]:

print('optimal parameter ==> {}'.format(gs.best_params_))
print('optimal parameter의 점수 ==> {:.3f}'.format(gs.best_score_))
print('optimal parameter로 일반화 점수 ==> {:.3f}'.format(gs.score(X_test, y_test)))

optimal parameter ==> {'C': 10, 'gamma': 0.001, 'kernel': 'rbf'}
optimal parameter의 점수 ==> 0.941
optimal parameter로 일반화 점수 ==> 0.973

In [28]:

svc_optimal=SVC(kernel='rbf', C = 10, gamma = 0.001)
svc_optimal.fit(Xresampled, yresampled)

Out[28]:

SVC(C=10, break_ties=False, cache_size=200, class_weight=None, coef0=0.0,
    decision_function_shape='ovr', degree=3, gamma=0.001, kernel='rbf',
    max_iter=-1, probability=False, random_state=None, shrinking=True,
    tol=0.001, verbose=False)

In [29]:

y_pred_optimal = svc_optimal.predict(X_test2)
metrics.confusion_matrix(y_test2, y_pred_optimal)

Out[29]:

array([[13834,   375],
       [    2,    30]], dtype=int64)

In [30]:

print('Accuracy Score:')
print(metrics.accuracy_score(y_test2, y_pred_optimal))
print('Recall Score:')
print(metrics.recall_score(y_test2, y_pred_optimal))
print('Precision Score:')
print(metrics.precision_score(y_test2, y_pred_optimal))

Accuracy Score:
0.9735271399480374
Recall Score:
0.9375
Precision Score:
0.07407407407407407

결론 모델 중 Recall Score가 가장 높으면서 다른 Score가 가장 높은 모델은 svc_optimal 였다.

Recall Score가 중요한 이유 기계가 실제 사기인 사건을 사기가 아니라고 판단했을 경우가 더 치명적이기 때문이다.

PreviousPython을 이용한 SVM (1)Next차원 축소

Last updated 4 years ago