[문서 제목 자동 분류 작업 | Classification] 2. 모델 별 실행 결과 | 인턴

0. Data

train : 80%
test : 20%

 

 

---

 

1. K-NN

 

neighbor == 3

 

import pandas as pd
from konlpy.tag import Kkma
import string
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier
from sklearn.metrics import accuracy_score, confusion_matrix


csv_file_path = '../../data/spv/spv_data.csv'

try:
    df = pd.read_csv(csv_file_path, encoding='utf-8')
except UnicodeDecodeError:
    print('utf-8으로 디코딩하는 중 오류 발생. 다른 인코딩 시도 필요.')


remove_punct_dict = dict((ord(punct), None) for punct in string.punctuation)
kkma_kor = Kkma()


def KkmaLemNormalize_kor(text):
    if '.' in text:
        text = text.rsplit('.', 1)[0] # 확장자 제거 후 군집화
    tokens = kkma_kor.morphs(text)
    return ','.join(tokens)

tfidf_vect_kor = TfidfVectorizer(tokenizer=KkmaLemNormalize_kor, ngram_range=(1,2))

ftr_vector_kor = tfidf_vect_kor.fit_transform(df['CONTENT_TITLE'])


# 레이블 확인
target_column = 'CATEGORY'  # 실제 레이블 컬럼 이름에 따라 수정해야 함
if target_column not in df.columns:
    print(f"'{target_column}' 컬럼이 데이터프레임에 없습니다.")
else:
    y = df[target_column]


X_train, X_test, y_train, y_test = train_test_split(ftr_vector_kor, y, test_size=0.2, random_state=42)



# RandomForestClassifier 또는 다른 분류 모델을 선택
model = KNeighborsClassifier(n_neighbors=3)

# 모델 훈련
model.fit(X_train, y_train)



# 예측
y_pred = model.predict(X_test)

# df['pred_category'] = y_pred

#
# 정확도 평가
accuracy = accuracy_score(y_test, y_pred)
print(f'Accuracy: {accuracy}')

print('Confusion Matrix:')
cm = confusion_matrix(y_test, y_pred)
print(cm)


import seaborn as sns
import matplotlib.pyplot as plt


# 시각화
plt.figure(figsize=(8, 6))
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues', xticklabels=['Class 0', 'Class 1', 'Class 2'], yticklabels=['Class 0', 'Class 1', 'Class 2'])
plt.title('KNN_Confusion Matrix')
plt.xlabel('Predicted Label')
plt.ylabel('True Label')
plt.show()

 

 

Accuracy: 0.9503

Confusion Matrix:

 

 

 

---

 

 

2. NB

 

import pandas as pd
from konlpy.tag import Kkma
import string
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.model_selection import train_test_split
from sklearn.naive_bayes import MultinomialNB
from sklearn.metrics import accuracy_score, confusion_matrix


csv_file_path = '../../data/spv/spv_data.csv'

try:
    df = pd.read_csv(csv_file_path, encoding='utf-8')
except UnicodeDecodeError:
    print('utf-8으로 디코딩하는 중 오류 발생. 다른 인코딩 시도 필요.')


remove_punct_dict = dict((ord(punct), None) for punct in string.punctuation)
kkma_kor = Kkma()


def KkmaLemNormalize_kor(text):
    if '.' in text:
        text = text.rsplit('.', 1)[0] # 확장자 제거 후 군집화
    tokens = kkma_kor.morphs(text)
    return ','.join(tokens)

tfidf_vect_kor = TfidfVectorizer(tokenizer=KkmaLemNormalize_kor, ngram_range=(1,2))

ftr_vector_kor = tfidf_vect_kor.fit_transform(df['CONTENT_TITLE'])


# 레이블 확인
target_column = 'CATEGORY'  # 실제 레이블 컬럼 이름에 따라 수정해야 함
if target_column not in df.columns:
    print(f"'{target_column}' 컬럼이 데이터프레임에 없습니다.")
else:
    y = df[target_column]


X_train, X_test, y_train, y_test = train_test_split(ftr_vector_kor, y, test_size=0.2, random_state=42)



# RandomForestClassifier 또는 다른 분류 모델을 선택
model = MultinomialNB()

# 모델 훈련
model.fit(X_train, y_train)



# 예측
y_pred = model.predict(X_test)

# df['pred_category'] = y_pred

#
# 정확도 평가
accuracy = accuracy_score(y_test, y_pred)
print(f'Accuracy: {accuracy}')

print('Confusion Matrix:')
cm = confusion_matrix(y_test, y_pred)
print(cm)


import seaborn as sns
import matplotlib.pyplot as plt


# 시각화
plt.figure(figsize=(8, 6))
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues', xticklabels=['Class 0', 'Class 1', 'Class 2'], yticklabels=['Class 0', 'Class 1', 'Class 2'])
plt.title('NB_Confusion Matrix')
plt.xlabel('Predicted Label')
plt.ylabel('True Label')
plt.show()

 

Accuracy: 0.9063

Confusion Matrix:

 

 

 

---

 

3. RF

 

import pandas as pd
from konlpy.tag import Kkma
import string
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score, classification_report, confusion_matrix


csv_file_path = '../../data/spv/spv_data.csv'

try:
    df = pd.read_csv(csv_file_path, encoding='utf-8')
except UnicodeDecodeError:
    print('utf-8으로 디코딩하는 중 오류 발생. 다른 인코딩 시도 필요.')


remove_punct_dict = dict((ord(punct), None) for punct in string.punctuation)
kkma_kor = Kkma()


def KkmaLemNormalize_kor(text):
    if '.' in text:
        text = text.rsplit('.', 1)[0] # 확장자 제거 후 군집화
    tokens = kkma_kor.morphs(text)
    return ','.join(tokens)

tfidf_vect_kor = TfidfVectorizer(tokenizer=KkmaLemNormalize_kor, ngram_range=(1,2))

ftr_vector_kor = tfidf_vect_kor.fit_transform(df['CONTENT_TITLE'])


# 레이블 확인
target_column = 'CATEGORY'  # 실제 레이블 컬럼 이름에 따라 수정해야 함
if target_column not in df.columns:
    print(f"'{target_column}' 컬럼이 데이터프레임에 없습니다.")
else:
    y = df[target_column]


X_train, X_test, y_train, y_test = train_test_split(ftr_vector_kor, y, test_size=0.2, random_state=42)



# RandomForestClassifier 또는 다른 분류 모델을 선택
model = RandomForestClassifier()

# 모델 훈련
model.fit(X_train, y_train)



# 예측
y_pred = model.predict(X_test)

# df['pred_category'] = y_pred

#
# 정확도 평가
accuracy = accuracy_score(y_test, y_pred)
print(f'Accuracy: {accuracy}')

print('Confusion Matrix:')
cm = confusion_matrix(y_test, y_pred)
print(cm)


import seaborn as sns
import matplotlib.pyplot as plt


# 시각화
plt.figure(figsize=(8, 6))
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues', xticklabels=['Class 0', 'Class 1', 'Class 2'], yticklabels=['Class 0', 'Class 1', 'Class 2'])
plt.title('Confusion Matrix')
plt.xlabel('Predicted Label')
plt.ylabel('True Label')
plt.show()

 

Accuracy: 0.9848

Confusion Matrix:

 

 

---

 

4. SVM

 

import pandas as pd
from konlpy.tag import Kkma
import string
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.model_selection import train_test_split
from sklearn.svm import SVC
from sklearn.metrics import accuracy_score, confusion_matrix


csv_file_path = '../../data/spv/spv_data.csv'

try:
    df = pd.read_csv(csv_file_path, encoding='utf-8')
except UnicodeDecodeError:
    print('utf-8으로 디코딩하는 중 오류 발생. 다른 인코딩 시도 필요.')


remove_punct_dict = dict((ord(punct), None) for punct in string.punctuation)
kkma_kor = Kkma()


def KkmaLemNormalize_kor(text):
    if '.' in text:
        text = text.rsplit('.', 1)[0] # 확장자 제거 후 군집화
    tokens = kkma_kor.morphs(text)
    return ','.join(tokens)

tfidf_vect_kor = TfidfVectorizer(tokenizer=KkmaLemNormalize_kor, ngram_range=(1,2))

ftr_vector_kor = tfidf_vect_kor.fit_transform(df['CONTENT_TITLE'])


# 레이블 확인
target_column = 'CATEGORY'  # 실제 레이블 컬럼 이름에 따라 수정해야 함
if target_column not in df.columns:
    print(f"'{target_column}' 컬럼이 데이터프레임에 없습니다.")
else:
    y = df[target_column]


X_train, X_test, y_train, y_test = train_test_split(ftr_vector_kor, y, test_size=0.2, random_state=42)



# RandomForestClassifier 또는 다른 분류 모델을 선택
model = SVC(kernel='linear')

# 모델 훈련
model.fit(X_train, y_train)



# 예측
y_pred = model.predict(X_test)

# df['pred_category'] = y_pred

#
# 정확도 평가
accuracy = accuracy_score(y_test, y_pred)
print(f'Accuracy: {accuracy}')

print('Confusion Matrix:')
cm = confusion_matrix(y_test, y_pred)
print(cm)


import seaborn as sns
import matplotlib.pyplot as plt


# 시각화
plt.figure(figsize=(8, 6))
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues', xticklabels=['Class 0', 'Class 1', 'Class 2'], yticklabels=['Class 0', 'Class 1', 'Class 2'])
plt.title('SVM_Confusion Matrix')
plt.xlabel('Predicted Label')
plt.ylabel('True Label')
plt.show()

 

 

Accuracy: 0.9900

Confusion Matrix:

 

 

---

 

5. 결과 비교

 

Total data : 9560

Train data : 7648개

X_train : (7648, 4313)

Y_train : (7648,)

Test data : 1912 개

X_test : (1912, 4313)

Y_test:(1912,)

 

 

	정확도	confusion matrix
KNN	0.9503	[[ 11 1 7] [ 0 282 35] [ 2 50 1524]]
			NB	0.9063	[[ 0 3 16] [ 0 180 137] [ 2 21 1553]]
						RF	0.9848	[[ 9 1 9] [ 0 302 15] [ 0 0 1576]]
SVM	0.9900	[[ 12 1 6] [ 0 308 9] [ 0 3 1573]]

 

 

 

 

정확도가 99인게 오버피팅 관련해서 마음에 걸리긴 하지만, 데이터가 얼마 없는 것도 감수해야 하기에

그래도 정확도가 높은 모델인 SVM으로 최종 테스트를 진행하기로 결정