词嵌入+神经网络进行邮件分类

1 问题描述

问题：邮件分类问题（Email classification）

任务：将邮件分为两类(spam or ham)

数据集：https://www.kaggle.com/uciml/sms-spam-collection-dataset#spam.csv

2 数据处理

import pandas as pd
from nltk.corpus import stopwords
from nltk.stem import PorterStemmer
from textblob import Word
import re
from sklearn.model_selection import train_test_split

读取数据

# 读取数据
data = pd.read_csv('spam.csv', encoding = "ISO-8859-1")

data.columns

Index(['v1', 'v2', 'Unnamed: 2', 'Unnamed: 3', 'Unnamed: 4'], dtype='object')

# 查看前5行数据
data.head()

	v1	v2	Unnamed: 2	Unnamed: 3	Unnamed: 4
0	ham	Go until jurong point, crazy.. Available only ...	NaN	NaN	NaN
1	ham	Ok lar... Joking wif u oni...	NaN	NaN	NaN
2	spam	Free entry in 2 a wkly comp to win FA Cup fina...	NaN	NaN	NaN
3	ham	U dun say so early hor... U c already then say...	NaN	NaN	NaN
4	ham	Nah I don't think he goes to usf, he lives aro...	NaN	NaN	NaN

去除无用数据

# 去除无用数据，后3列是无用数据
data = data[['v1', 'v2']]
data.head()

	v1	v2
0	ham	Go until jurong point, crazy.. Available only ...
1	ham	Ok lar... Joking wif u oni...
2	spam	Free entry in 2 a wkly comp to win FA Cup fina...
3	ham	U dun say so early hor... U c already then say...
4	ham	Nah I don't think he goes to usf, he lives aro...

修改表头信息

# 修改表头信息
data = data.rename(columns={"v1":"label","v2":"text"})
data.head()

	label	text
0	ham	Go until jurong point, crazy.. Available only ...
1	ham	Ok lar... Joking wif u oni...
2	spam	Free entry in 2 a wkly comp to win FA Cup fina...
3	ham	U dun say so early hor... U c already then say...
4	ham	Nah I don't think he goes to usf, he lives aro...

去除标点符号及多余的空格

# 去除标点符号及两个以上的空格
data['text'] = data['text'].apply(lambda x:re.sub('[!@#$:).;,?&]', ' ', x.lower()))
data['text'] = data['text'].apply(lambda x:re.sub(' ', ' ', x))
data['text'][0]

'go until jurong point  crazy   available only in bugis n great world la e buffet    cine there got amore wat   '

单词转换为小写

# 单词转换为小写
data['text'] = data['text'].apply(lambda x:" ".join(x.lower() for x in x.split()))
# 或者 
#data['text'] = data['text'].apply(lambda x:x.lower())
data['text'][0]

'go until jurong point crazy available only in bugis n great world la e buffet cine there got amore wat'

去除停止词

# 去除停止词 ，如a、an、the、高频介词、连词、代词等
stop = stopwords.words('english')
data['text'] = data['text'].apply(lambda x: " ".join(x for x in x.split() if x not in stop))
data['text'][0]

'go jurong point crazy available bugis n great world la e buffet cine got amore wat'

分词处理

# 分词处理，希望能够实现还原英文单词原型
st = PorterStemmer()
data['text'] = data['text'].apply(lambda x: " ".join([st.stem(word) for word in x.split()]))
data['text'] = data['text'].apply(lambda x: " ".join([Word(word).lemmatize() for word in x.split()]))
data['text'][0]

'go jurong point crazi avail bugi n great world la e buffet cine got amor wat'

data.head()

	label	text
0	ham	go jurong point crazi avail bugi n great world...
1	ham	ok lar joke wif u oni
2	spam	free entri 2 wkli comp win fa cup final tkt 21...
3	ham	u dun say earli hor u c alreadi say
4	ham	nah think goe usf live around though

3 特征提取

from keras.preprocessing.text import Tokenizer
from keras.preprocessing.sequence import pad_sequences

Using TensorFlow backend.

分出训练集和测试集

#以 8:2 的比例分出训练集和测试集
train, test = train_test_split(data, test_size=0.2)

设置参数

# 每个序列的最大长度，多了截断，少了补0
max_sequence_length = 300

#只保留频率最高的前20000个词
num_words = 20000

# 嵌入的维度
embedding_dim = 100

构建分词器

# 找出经常出现的单词，分词器
tokenizer = Tokenizer(num_words=num_words)
tokenizer.fit_on_texts(train.text)
train_sequences = tokenizer.texts_to_sequences(train.text)
test_sequences = tokenizer.texts_to_sequences(test.text)

# dictionary containing words and their index
word_index = tokenizer.word_index


# print(tokenizer.word_index)
# total words in the corpus
print('Found %s unique tokens.' % len(word_index))
# get only the top frequent words on train

train_x = pad_sequences(train_sequences, maxlen=max_sequence_length)
# get only the top frequent words on test
test_x = pad_sequences(test_sequences, maxlen=max_sequence_length)

print(train_x.shape)
print(test_x.shape)

Found 6702 unique tokens.
(4457, 300)
(1115, 300)

标签向量化

# 标签向量化
# [0,1]: ham;[1,0]:spam
import numpy as np

def lable_vectorize(labels):
    label_vec = np.zeros([len(labels),2])
    for i, label in enumerate(labels):
        if str(label)=='ham':
            label_vec[i][0] = 1
        else:
            label_vec[i][1] = 1
    return label_vec
            
train_y = lable_vectorize(train['label'])            
test_y = lable_vectorize(test['label'])


# 或者
from sklearn.preprocessing import LabelEncoder
from keras.utils import to_categorical

# converts the character array to numeric array. Assigns levels to unique labels.
train_labels = train['label']
test_labels = test['label']

le = LabelEncoder()
le.fit(train_labels)
train_labels = le.transform(train_labels)
test_labels = le.transform(test_labels)

# changing data types
labels_train = to_categorical(np.asarray(train_labels))
labels_test = to_categorical(np.asarray(test_labels))

4 构建模型并训练

# Import Libraries
import sys, os, re, csv, codecs, numpy as np, pandas as pd
from keras.preprocessing.text import Tokenizer
from keras.preprocessing.sequence import pad_sequences
from keras.utils import to_categorical
from keras.layers import Dense, Input, LSTM, Embedding,Dropout, Activation
from keras.layers import Bidirectional, GlobalMaxPool1D,Conv1D, SimpleRNN
from keras.models import Model
from keras.models import Sequential
from keras import initializers, regularizers, constraints,optimizers, layers
from keras.layers import Dense, Input, Flatten, Dropout,BatchNormalization
from keras.layers import Conv1D, MaxPooling1D, Embedding
from keras.models import Sequential

model = Sequential()
model.add(Embedding(num_words,
                    embedding_dim,
                    input_length=max_sequence_length))
model.add(Dropout(0.5))
model.add(Conv1D(128, 5, activation='relu'))
model.add(MaxPooling1D(5))
model.add(Dropout(0.5))

model.add(BatchNormalization())
model.add(Conv1D(128, 5, activation='relu'))
model.add(MaxPooling1D(5))
model.add(Dropout(0.5))

model.add(BatchNormalization())
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dense(2, activation='softmax'))
model.compile(loss='categorical_crossentropy',
              optimizer='rmsprop',
              metrics=['acc'])

model.fit(train_x, train_y,
            batch_size=64,
            epochs=5,
            validation_split=0.2)

Train on 3565 samples, validate on 892 samples
Epoch 1/5
3565/3565 [==============================] - 25s 7ms/step - loss: 0.3923 - acc: 0.8480 - val_loss: 0.1514 - val_acc: 0.9451
Epoch 2/5
3565/3565 [==============================] - 23s 7ms/step - loss: 0.1729 - acc: 0.9372 - val_loss: 0.0789 - val_acc: 0.9753
Epoch 3/5
3565/3565 [==============================] - 25s 7ms/step - loss: 0.0940 - acc: 0.9731 - val_loss: 0.2079 - val_acc: 0.9787
Epoch 4/5
3565/3565 [==============================] - 23s 7ms/step - loss: 0.0590 - acc: 0.9857 - val_loss: 0.3246 - val_acc: 0.9843
Epoch 5/5
3565/3565 [==============================] - 23s 7ms/step - loss: 0.0493 - acc: 0.9882 - val_loss: 0.3150 - val_acc: 0.9877





<keras.callbacks.History at 0x1cac6187940>

5 模型评估

# [0.07058866604882806, 0.9874439467229116]
model.evaluate(test_x, test_y)

1115/1115 [==============================] - 2s 2ms/step





[0.32723046118903054, 0.97847533632287]

# prediction on test data
predicted=model.predict(test_x)
predicted

array([[0.71038646, 0.28961352],
       [0.71285075, 0.28714925],
       [0.7101978 , 0.28980213],
       ...,
       [0.7092874 , 0.29071262],
       [0.70976096, 0.290239  ],
       [0.70463425, 0.29536578]], dtype=float32)

#模型评估
import sklearn
from sklearn.metrics import precision_recall_fscore_support as score
precision, recall, fscore, support = score(test_y,predicted.round())
print('precision: {}'.format(precision))
print('recall: {}'.format(recall))
print('fscore: {}'.format(fscore))
print('support: {}'.format(support))
print("############################")
print(sklearn.metrics.classification_report(test_y,predicted.round()))

precision: [0.97961264 0.97014925]
recall: [0.99585492 0.86666667]
fscore: [0.98766701 0.91549296]
support: [965 150]
############################
             precision    recall  f1-score   support

          0       0.98      1.00      0.99       965
          1       0.97      0.87      0.92       150

avg / total       0.98      0.98      0.98      1115

文章来源: https://foochane.cn/article/2019052202.html