자연어 처리 - RNN 스팸 메일 분류
스팸 메일 분류 - RNN
전처리
In [1]:
import pandas as pd
In [2]:
data = pd.read_csv('datasets/spam.csv', encoding='latin1')
data = data[['v1', 'v2']]
data['v1'] = data['v1'].replace(['ham', 'spam'], [0, 1])
# data['v1'] = data['v1'].replace({'ham':0, 'spam':1})
data
Out [2]:
| v1 | v2 | |
|---|---|---|
| 0 | 0 | Go until jurong point, crazy.. Available only ... |
| 1 | 0 | Ok lar... Joking wif u oni... |
| 2 | 1 | Free entry in 2 a wkly comp to win FA Cup fina... |
| 3 | 0 | U dun say so early hor... U c already then say... |
| 4 | 0 | Nah I don't think he goes to usf, he lives aro... |
| ... | ... | ... |
| 5567 | 1 | This is the 2nd time we have tried 2 contact u... |
| 5568 | 0 | Will Ì_ b going to esplanade fr home? |
| 5569 | 0 | Pity, * was in mood for that. So...any other s... |
| 5570 | 0 | The guy did some bitching but I acted like i'd... |
| 5571 | 0 | Rofl. Its true to its name |
5572 rows × 2 columns
- 데이터 중복 확인
In [3]:
data.info()
Out [3]:
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 5572 entries, 0 to 5571
Data columns (total 2 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 v1 5572 non-null int64
1 v2 5572 non-null object
dtypes: int64(1), object(1)
memory usage: 87.2+ KB
In [4]:
data['v2'].nunique()
Out [4]:
5169
In [5]:
data = data.drop_duplicates(subset=['v2'])
data.info()
Out [5]:
<class 'pandas.core.frame.DataFrame'>
Int64Index: 5169 entries, 0 to 5571
Data columns (total 2 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 v1 5169 non-null int64
1 v2 5169 non-null object
dtypes: int64(1), object(1)
memory usage: 121.1+ KB
In [6]:
data['v2'].nunique()
Out [6]:
5169
In [7]:
data['v1'].value_counts()
Out [7]:
0 4516
1 653
Name: v1, dtype: int64
In [8]:
X_data = data['v2']
y_data = data['v1']
- train, test 나누기
In [9]:
from sklearn.model_selection import train_test_split
In [10]:
X_train, X_test, y_train, y_test = train_test_split(X_data, y_data, test_size=0.2, random_state=0, stratify=y_data)
In [11]:
X_train.shape, X_test.shape, y_train.shape, y_test.shape
Out [11]:
((4135,), (1034,), (4135,), (1034,))
- 토큰화
In [12]:
from keras.preprocessing.text import Tokenizer
In [13]:
tok = Tokenizer()
tok.fit_on_texts(X_train)
X_train_encoded = tok.texts_to_sequences(X_train)
X_train_encoded[:2]
Out [13]:
[[102, 1, 210, 230, 3, 17, 39], [1, 59, 8, 427, 17, 5, 137, 2, 2326]]
In [14]:
print(tok.word_index)
Out [14]:
{'i': 1, 'to': 2, 'you': 3, 'a': 4, 'the': 5, 'u': 6, 'and': 7, 'in': 8, 'is': 9, 'me': 10, 'my': 11, 'for': 12, 'your': 13, 'it': 14, 'of': 15, 'have': 16, 'on': 17, 'call': 18, 'that': 19, 'are': 20, '2': 21, 'now': 22, 'so': 23, 'but': 24, 'not': 25, 'can': 26, 'or': 27, "i'm": 28, 'get': 29, 'at': 30, 'do': 31, 'if': 32, 'be': 33, 'will': 34, 'just': 35, 'with': 36, 'we': 37, 'no': 38, 'this': 39, 'ur': 40,
...
'khelate': 7804, 'kintu': 7805, 'opponenter': 7806, 'dhorte': 7807, 'lage': 7808, "tm'ing": 7809, 'laughs': 7810, 'adding': 7811, 'zeros': 7812, 'savings': 7813, 'hos': 7814, 'heroes': 7815, 'tips': 7816, 'genes': 7817, 'begun': 7818, 'registration': 7819, 'permanent': 7820, 'residency': 7821}
In [15]:
total_cnt = len(tok.word_index)
vocab_size = len(tok.word_index) + 1 # zero padding
vocab_size
Out [15]:
7822
- 메일 제목 길이 분포 확인
In [16]:
import matplotlib.pyplot as plt
In [17]:
plt.hist([len(sample) for sample in X_data], bins=100)
plt.show()
Out [17]:
In [18]:
from keras.utils import pad_sequences
In [19]:
max_len = 200 # 너무 긴 제목은 200 단어가 넘어가면 잘라냄, 모자르면 0으로 채움
X_train_padded = pad_sequences(X_train_encoded, maxlen=max_len)
X_train_padded.shape
Out [19]:
(4135, 200)
딥러닝
In [20]:
from keras.models import Sequential
from keras.layers import Embedding, SimpleRNN, Dense
In [21]:
model = Sequential([
Embedding(vocab_size, 100),
SimpleRNN(32),
Dense(1, activation='sigmoid')
])
model.summary()
Out [21]:
Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
embedding (Embedding) (None, None, 100) 782200
simple_rnn (SimpleRNN) (None, 32) 4256
dense (Dense) (None, 1) 33
=================================================================
Total params: 786,489
Trainable params: 786,489
Non-trainable params: 0
_________________________________________________________________
In [22]:
model.compile(optimizer='rmsprop', loss='binary_crossentropy', metrics=['acc'])
history = model.fit(X_train_padded, y_train, epochs=5, batch_size=64, validation_split=0.2)
Out [22]:
Epoch 1/5
C:\Users\user\AppData\Roaming\Python\Python38\site-packages\keras\engine\data_adapter.py:1508: FutureWarning: The behavior of `series[i:j]` with an integer-dtype index is deprecated. In a future version, this will be treated as *label-based* indexing, consistent with e.g. `series[i]` lookups. To retain the old behavior, use `series.iloc[i:j]`. To get the future behavior, use `series.loc[i:j]`.
return t[start:end]
52/52 [==============================] - 3s 36ms/step - loss: 0.2688 - acc: 0.9144 - val_loss: 0.1273 - val_acc: 0.9613
Epoch 2/5
52/52 [==============================] - 2s 31ms/step - loss: 0.0851 - acc: 0.9794 - val_loss: 0.0962 - val_acc: 0.9698
Epoch 3/5
52/52 [==============================] - 2s 32ms/step - loss: 0.0425 - acc: 0.9885 - val_loss: 0.0716 - val_acc: 0.9794
Epoch 4/5
52/52 [==============================] - 2s 31ms/step - loss: 0.0244 - acc: 0.9927 - val_loss: 0.0623 - val_acc: 0.9843
Epoch 5/5
52/52 [==============================] - 2s 33ms/step - loss: 0.0135 - acc: 0.9958 - val_loss: 0.0615 - val_acc: 0.9831
In [23]:
X_test_encoded = tok.texts_to_sequences(X_test)
X_test_padded = pad_sequences(X_test_encoded, maxlen=max_len)
model.evaluate(X_test_padded, y_test)
Out [23]:
33/33 [==============================] - 0s 7ms/step - loss: 0.0889 - acc: 0.9768
[0.08886431902647018, 0.9767891764640808]
딥러닝에서는 일반 머신러닝과 다르게 제목의 문자 조합 즉, 문맥으로 구분한다.
하지만 진짜 의미를 아는게(XAI) 아니다.
댓글남기기