Re: 문제13.(오늘의 마지막 문제) 방금 수행했던 독버섯 분류 나이브 베이즈 모델의 라플라스 값을 알아내시오

[12기_유혜영]

# ■ mushrooms 독버섯데이터로 나이브베이즈

import pandas as pd  # 데이터 전처리를 위해서 
import seaborn as sns # 시각화를 위해서 

df = pd.read_csv('c:\\data\\mushrooms.csv') 
df = pd.get_dummies(df)

# X = 전체 행, 마지막 열 제외한 모든 열 데이터 -> n차원 공간의 포인트
X = df.iloc[:, 2:].to_numpy() 
y = df.iloc[:,1].to_numpy()   
  
from sklearn.model_selection import train_test_split 
                                                                
# 훈련 데이터 75, 테스트 데이터 25으로 나눈다. 
X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.25, random_state = 10)

import  numpy  as np
from sklearn.naive_bayes import GaussianNB

errors = []

for i in np.arange(0.001,0.01,0.001):   # 0.001, 0.002 , ... , 0.010
    nb = GaussianNB(var_smoothing=i)
    nb.fit(X_train, y_train)
    pred_i = nb.predict(X_test)
    errors.append(np.mean(pred_i != y_test))
print(errors)

for k, i  in  zip(np.arange(0.001,0.01,0.001), errors):
    print (round(k, 3), '--->', i)  # 0.004 ---> 0.008370260955194485

결과:

0.001 ---> 0.009354997538158542
0.002 ---> 0.008862629246676515
0.003 ---> 0.008862629246676515
0.004 ---> 0.008370260955194485
0.005 ---> 0.008370260955194485
0.006 ---> 0.008370260955194485
0.007 ---> 0.008370260955194485
0.008 ---> 0.008370260955194485
0.009 ---> 0.008370260955194485

import matplotlib.pyplot as plt

plt.plot(np.arange(0.001,0.01,0.001), errors, marker='o')
plt.title('Mean error with K-Value')
plt.xlabel('laplace value')
plt.ylabel('mean error')
plt.show()

#%%

# 학습/예측(Training/Pradiction)
from sklearn.naive_bayes import GaussianNB

# 나이브베이즈 분류기를 생성
classifier = GaussianNB(var_smoothing=0.004) 

# 분류기 학습
classifier.fit(X_train, y_train)

# 예측
y_pred= classifier.predict(X_test)

# 작은 이원교차표
from sklearn.metrics import confusion_matrix
conf_matrix= confusion_matrix(y_test, y_pred)
print(conf_matrix)   

결과: 

[[1055   17]
 [   0  959]]

# 정밀도 , 재현율, f1 score 확인 
from sklearn.metrics import classification_report
report = classification_report(y_test, y_pred)
print(report)

결과:

              precision    recall  f1-score   support

           0       1.00      0.98      0.99      1072
           1       0.98      1.00      0.99       959

    accuracy                           0.99      2031
   macro avg       0.99      0.99      0.99      2031
weighted avg       0.99      0.99      0.99      2031

# 정확도 확인하는 코드 
from sklearn.metrics import accuracy_score
accuracy = accuracy_score( y_test, y_pred)
print(accuracy) 

정확도 : 0.9916297390448056 (Laplace : 0.004)