垃圾邮件识别¶

1. 数据集介绍¶

email数据集(email.csv)整理自某邮箱账户2012年前3个月收到的所有邮件，邮件按是否为垃圾邮件进行标注，邮件内容也已进行预处理与特征提取。数据字段及具体含义如下：

  • spam: Indicator for whether the email was spam.
  • to_multiple: Indicator for whether the email was addressed to more than one recipient.
  • from: Whether the message was listed as from anyone (this is usually set by default for regular outgoing email).
  • cc: Number of people cc'ed.
  • sent_email: Indicator for whether the sender had been sent an email in the last 30 days.
  • time: Time at which email was sent.
  • image: The number of images attached.
  • attach: The number of attached files.
  • dollar: The number of times a dollar sign or the word "dollar" appeared in the email.
  • winner: Indicates whether "winner" appeared in the email.
  • inherit: The number of times “inherit” (or an extension, such as “inheritance”) appeared in the email.
  • viagra: The number of times “viagra” appeared in the email.
  • password: The number of times “password” appeared in the email.
  • num_char: The number of characters in the email, in thousands.
  • line_breaks: The number of line breaks in the email (does not count text wrapping).
  • format: Indicates whether the email was written using HTML (e.g. may have included bolding or active links).
  • re_subj: Whether the subject started with "Re:", "RE:", "re:", or "rE:"
  • exclaim_subj: Whether there was an exclamation point in the subject.
  • urgent_subj: Whether the word “urgent” was in the email subject.
  • exclaim_mess: The number of exclamation points in the email message.
  • number: Factor variable saying whether there was no number, a small number (under 1 million), or a big number.

2. 数据探索分析与预处理¶

In [1]:
## load required libraries 
import matplotlib.pyplot as plt 
import pandas as pd 
import numpy as np
import warnings
warnings.simplefilter("ignore")

2.1. 导入数据¶

In [2]:
data = pd.read_csv("./email.csv")
data.head()
Out[2]:
spam to_multiple from cc sent_email image attach dollar winner inherit password num_char line_breaks format re_subj exclaim_subj urgent_subj exclaim_mess number
0 0 no yes no no no no no no no no 11.370 202 HTML no no no 0 big
1 0 no yes no no no no no no no no 10.504 202 HTML no no no 1 small
2 0 no yes no no no no yes no yes no 7.773 192 HTML no no no 6 small
3 0 no yes no no no no no no no no 13.256 255 HTML no no no 48 small
4 0 no yes no no no no no no no yes 1.231 29 Plain no no no 1 none
In [3]:
data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3921 entries, 0 to 3920
Data columns (total 19 columns):
 #   Column        Non-Null Count  Dtype  
---  ------        --------------  -----  
 0   spam          3921 non-null   int64  
 1   to_multiple   3921 non-null   object 
 2   from          3921 non-null   object 
 3   cc            3921 non-null   object 
 4   sent_email    3921 non-null   object 
 5   image         3921 non-null   object 
 6   attach        3921 non-null   object 
 7   dollar        3921 non-null   object 
 8   winner        3921 non-null   object 
 9   inherit       3921 non-null   object 
 10  password      3921 non-null   object 
 11  num_char      3921 non-null   float64
 12  line_breaks   3921 non-null   int64  
 13  format        3921 non-null   object 
 14  re_subj       3921 non-null   object 
 15  exclaim_subj  3921 non-null   object 
 16  urgent_subj   3921 non-null   object 
 17  exclaim_mess  3921 non-null   int64  
 18  number        3921 non-null   object 
dtypes: float64(1), int64(3), object(15)
memory usage: 582.2+ KB

2.2. 数据预处理¶

类别型变量 => 数值型变量¶

比如将二分类变量映射为1/0，例如yes/no

In [4]:
yes_no_map = {
    'yes': 1,
    'no': 0
}
data['to_multiple'] = data['to_multiple'].map(yes_no_map)
data['from'] = data['from'].map(yes_no_map)
data['cc'] = data['cc'].map(yes_no_map)
data['sent_email'] = data['sent_email'].map(yes_no_map)
data['image'] = data['image'].map(yes_no_map)
data['attach'] = data['attach'].map(yes_no_map)
data['dollar'] = data['dollar'].map(yes_no_map)
data['winner'] = data['winner'].map(yes_no_map)
data['inherit'] = data['inherit'].map(yes_no_map)
data['password'] = data['password'].map(yes_no_map)
data['re_subj'] = data['re_subj'].map(yes_no_map)
data['exclaim_subj'] = data['exclaim_subj'].map(yes_no_map)
data['urgent_subj'] = data['urgent_subj'].map(yes_no_map)
In [5]:
# format
data['format'].value_counts()
Out[5]:
format
HTML     2726
Plain    1195
Name: count, dtype: int64
In [6]:
format_map = {
    'HTML': 1,
    'Plain': 0
}
data['format'] = data['format'].map(format_map)
In [7]:
# number
data['number'].value_counts()
Out[7]:
number
small    2827
none      549
big       545
Name: count, dtype: int64

使用哑变量表示多类别变量

In [8]:
data = data.join(pd.get_dummies(data['number'], prefix = 'number'))
data = data.drop(['number'], axis = 1)
data.head()
Out[8]:
spam to_multiple from cc sent_email image attach dollar winner inherit ... num_char line_breaks format re_subj exclaim_subj urgent_subj exclaim_mess number_big number_none number_small
0 0 0 1 0 0 0 0 0 0 0 ... 11.370 202 1 0 0 0 0 True False False
1 0 0 1 0 0 0 0 0 0 0 ... 10.504 202 1 0 0 0 1 False False True
2 0 0 1 0 0 0 0 1 0 1 ... 7.773 192 1 0 0 0 6 False False True
3 0 0 1 0 0 0 0 0 0 0 ... 13.256 255 1 0 0 0 48 False False True
4 0 0 1 0 0 0 0 0 0 0 ... 1.231 29 0 0 0 0 1 False True False

5 rows × 21 columns

2.3. 数据探索性分析¶

因变量: 分布情况¶

In [9]:
print(data['spam'].unique())
data_label_count = data.groupby('spam').count()['cc'].sort_values(ascending = False)
print(data_label_count)
data_label_count.plot.bar()
plt.show()

[0 1]
spam
0    3554
1     367
Name: cc, dtype: int64
No description has been provided for this image

自变量: 分布情况 & 相关性分析¶

In [10]:
import seaborn as sns
graph_by_variables = data.columns
plt.figure(figsize = (15, 18))
for i in range(0, 21):
    plt.subplot(7, 3, i+1)
    sns.distplot(data[graph_by_variables[i]])
    plt.title(graph_by_variables[i])
plt.tight_layout()
No description has been provided for this image
In [11]:
f, ax = plt.subplots(figsize = (15, 15))
sns.heatmap(data.corr(), annot = True, linewidths = 0.5, fmt = '.1f', ax = ax)
Out[11]:
<Axes: >
No description has been provided for this image

2.4. 数据集划分¶

In [12]:
from sklearn.model_selection import train_test_split
# Splitting into train and test sets
X = data.drop(['spam'], axis = 1)
y = data['spam']
# test_size = 0.2意为训练集占80%，测试集占20%，即将数据集按4:1的比例划分为训练集和测试集
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, random_state = 628)
print(len(X_train), len(X_test))

3136 785

3. 垃圾邮件识别¶

3.1. 逻辑回归¶

模型构建¶

In [13]:
from sklearn.linear_model import LogisticRegression
# Fitting a logistic regression model with default parameters
logreg = LogisticRegression()
logreg.fit(X_train, y_train)
Out[13]:
LogisticRegression()
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
LogisticRegression()

效果评价¶

In [14]:
from sklearn.metrics import precision_score
from sklearn.metrics import recall_score
from sklearn.metrics import f1_score
from sklearn.metrics import roc_auc_score
In [15]:
# Prediction & Evaluation
y_hat_test = logreg.predict(X_test)
# Logistic Regression score
print("Logistic regression score for test set:")
print("Predicion: {:.3f}".format(precision_score(y_test, y_hat_test)), "Recall: {:.3f}".format(recall_score(y_test, y_hat_test)))
print("F1 score: {:.3f}".format(f1_score(y_test, y_hat_test)))
print("AUC score: {:.3f}".format(roc_auc_score(y_test, y_hat_test)))

Logistic regression score for test set:
Predicion: 0.667 Recall: 0.116
F1 score: 0.198
AUC score: 0.555

3.2. 朴素贝叶斯¶

模型构建¶

In [16]:
from sklearn.naive_bayes import BernoulliNB
# Fitting a Naive Bayes model with default parameters
clf = BernoulliNB()
clf.fit(X_train, y_train)
Out[16]:
BernoulliNB()
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
BernoulliNB()

效果评价¶

In [17]:
# Prediction & Evaluation
y_hat_test = clf.predict(X_test)
# Naive Bayes score
print("Naive Bayes score for test set:")
print("Predicion: {:.3f}".format(precision_score(y_test, y_hat_test)), "Recall: {:.3f}".format(recall_score(y_test, y_hat_test)))
print("F1 score: {:.3f}".format(f1_score(y_test, y_hat_test)))
print("AUC score: {:.3f}".format(roc_auc_score(y_test, y_hat_test)))

Naive Bayes score for test set:
Predicion: 0.500 Recall: 0.377
F1 score: 0.430
AUC score: 0.670

3.3. 决策树模型¶

模型构建¶

In [18]:
from sklearn.tree import DecisionTreeClassifier
# Fitting a decision tree model with default parameters
dt = DecisionTreeClassifier()
dt.fit(X_train, y_train)
Out[18]:
DecisionTreeClassifier()
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
DecisionTreeClassifier()

效果评价¶

In [19]:
# Prediction & Evaluation
y_hat_test = dt.predict(X_test)
# Decision Tree score
print("Decision tree score for test set:")
print("Predicion: {:.3f}".format(precision_score(y_test, y_hat_test)), "Recall: {:.3f}".format(recall_score(y_test, y_hat_test)))
print("F1 score: {:.3f}".format(f1_score(y_test, y_hat_test)))
print("AUC score: {:.3f}".format(roc_auc_score(y_test, y_hat_test)))

Decision tree score for test set:
Predicion: 0.595 Recall: 0.725
F1 score: 0.654
AUC score: 0.839

决策树可视化¶

In [20]:
# use graphviz
# from sklearn import tree
# import graphviz
# dot_data = tree.export_graphviz(dt, feature_names = X.columns, filled = True, class_names = True,out_file=None)  
# graph = graphviz.Source(dot_data)
# graph


# use matplotlib
import matplotlib.pyplot as plt
from sklearn import tree
plt.figure(figsize=(12,8))
tree.plot_tree(dt, feature_names=X.columns, class_names=[str(c) for c in dt.classes_], filled=True)
plt.show()
No description has been provided for this image

防止模型过度增长¶

In [21]:
from sklearn.tree import DecisionTreeClassifier
# Fitting a decision tree model with default parameters
dt = DecisionTreeClassifier(max_depth = 4)
dt.fit(X_train, y_train)
Out[21]:
DecisionTreeClassifier(max_depth=4)
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
DecisionTreeClassifier(max_depth=4)
In [22]:
# Prediction & Evaluation
y_hat_test = dt.predict(X_test)
# Decision Tree score
print("Decision tree score for test set:")
print("Predicion: {:.3f}".format(precision_score(y_test, y_hat_test)), "Recall: {:.3f}".format(recall_score(y_test, y_hat_test)))
print("F1 score: {:.3f}".format(f1_score(y_test, y_hat_test)))
print("AUC score: {:.3f}".format(roc_auc_score(y_test, y_hat_test)))

Decision tree score for test set:
Predicion: 0.649 Recall: 0.348
F1 score: 0.453
AUC score: 0.665
In [23]:
# use graphviz
# from sklearn import tree
# import graphviz
# dot_data = tree.export_graphviz(dt, feature_names = X.columns, filled = True, class_names = True,out_file=None)  
# graph = graphviz.Source(dot_data)
# graph

# use matplotlib
import matplotlib.pyplot as plt
plt.figure(figsize=(12,8))
tree.plot_tree(dt, feature_names=X.columns, class_names=[str(c) for c in dt.classes_], filled=True)
plt.show()
No description has been provided for this image
In [ ]: