Hand on Machine Learning第三章课后作业(1)：垃圾邮件分类

import os
import email
import email.policy

1. 读取邮件数据

SPAM_PATH = os.path.join(
    "E:\\3.Study\\机器学习\\Hand on Machine Learning\\第三章：分类\\spam_data")
spam_path = os.path.join(SPAM_PATH, "spam")
ham_path = os.path.join(SPAM_PATH, "easy_ham")
spam_list = [name for name in os.listdir(spam_path) if len(name) > 20]
ham_list = [name for name in os.listdir(ham_path) if len(name) > 20]

def load_email(is_spam, filename, spam_path=SPAM_PATH):
    directory = "spam" if is_spam else "easy_ham"
    with open(os.path.join(spam_path, directory, filename), "rb") as f:
        return email.parser.BytesParser(policy=email.policy.default).parse(f)

        # email.message.EmailMessage'类型,没法用list.append接收
        # return email.parser.BytesParser(policy=email.policy.default).parse(f)
# 这里有类型问题，应该记住这种加载email文件的形式。尝试list.append添加数据，加入的数据是generator类型，非email类型
ham_emails = [load_email(is_spam=False, filename=name) for name in ham_list]
spam_emails = [load_email(is_spam=True, filename=name) for name in spam_list]
print(spam_emails[0].get_content().strip())

2.分析邮件结构

def get_email_structure(email):
    # isinstance 函数：判断一个对象是否是已知类型。第一个参数为对象，第二个参数为类型名或者是类型名的列表。返回True/False
    if isinstance(email, str):
        return email
    # get_pyload()函数：返回当前邮件的正文。
    # 如果正文含有多个部分的话(is_multipart=True),返回一个message对象的list列表；
    # 如果is_multipart=False，即正文没有多部份的话，返回一个string类型。
    payload = email.get_payload()
    if isinstance(payload, list):
        return "multipart({})".format(", ".join([
            get_email_structure(sub_email)
            for sub_email in payload
        ]))
    else:
        return email.get_content_type()

from collections import Counter

def structures_counter(emails):
    # Counter类的目的是用来跟踪值出现的次数
    structures = Counter()
    for email in emails:
        structure = get_email_structure(email)
        structures[structure] += 1
    return structures

array = structures_counter(ham_emails).most_common()
array2 = structures_counter(spam_emails).most_common()
print(array)
print(array2)

3.分析邮件头部

for head, value in spam_emails[0].items():
    print(head, ":", value)
print(spam_emails[0]["Subject"])

4. 划分训练集，测试集

import numpy as np
X = np.array(ham_emails+spam_emails)  # ham_emails和span_emails是list类型
Y = np.array([0]*len(ham_emails)+[1]*len(spam_emails))

from sklearn.model_selection import train_test_split
x_train, x_test, y_train, y_test = train_test_split(
    X, Y, test_size=0.33, random_state=42)

5. 邮件文本预处理（转换HTML）

import re                          # regular expressions(正则）
from html import unescape

def html_to_plain_text(html):
    # sub->substitute(替换)
    # 参数1：pattern 正则;
    # 参数2：repl:replacement,被替换的字符串/函数
    # 参数3: string:需要被处理的内容
    # 参数4: count: 匹配的数目 如果正则表达式在string中有多个匹配结果，count控制匹配的数目
    # 参数5: flag : 匹配模式
    #       re.I 匹配对大小写不敏感
    #       re.M 多行匹配（以行为单位匹配）
    #       re.S 使 . 匹配包括换行在内的所有字符
    # ->用pattern模式将string里面count个的字符换成repl
    text = re.sub('<head.*?>.*?</head>', '', html, flags=re.M | re.S | re.I)
    text = re.sub('<a\s.*?>', 'HYPERLINK', text, flags=re.M | re.S | re.I)
    text = re.sub('<.*?>', '', text, flags=re.M | re.S)
    text = re.sub(r'(\s*\n)+', '\n', text, flags=re.M | re.S)
    return unescape(text)

html_spam_emails = [email for email in x_train[y_train == 1]
                    if get_email_structure(email) == "text/html"]
sample_html_spam = html_spam_emails[2]
# 输出html辣鸡邮件的前1000个字符,strip()->去除首尾空格
print(sample_html_spam.get_content().strip()[:1000], "...")
print("-"*30)
print(html_to_plain_text(sample_html_spam.get_content())[:1000], "...")

6.转换所有邮件为文本

def email_to_text(email):
    html = None
    # email->part->part.part 以树的结构存储，walk()用来循环遍历各个树及其子树
    for part in email.walk():
        ctype = part.get_content_type()
        if not ctype in ("text/plain", "text/html"):
            continue   # 跳过不是以上两种类型的部分
        try:
            content = part.get_content()
        except:
            content = str(part.get_payload())
        if ctype == "text/plain":
            return content
        else:
            html = content
    if html:
        return html_to_plain_text(html)

print(email_to_text(sample_html_spam)[:100], "...")

7. 自然语言处理

try:
    import nltk
    stemmer = nltk.PorterStemmer()  # 建立一个波特词干算法(分析单词的词干）
    for word in ("Conputations", "Computation", "Computing", "Computed", "Compulsive"):
        print(word, "=>", stemmer.stem(word))
except ImportError:
    print("Error: stemming requires the NLTK module.")
    stemmer = None

8.URL识别

try:
    import urlextract
    url_extracror = urlextract.URLExtract()
    print(url_extracror.find_urls("will it detect gitub.com and https://www.google.com/search?ei=nqXjXL2VM5vqwQPks4rQAw&q=python+nltk&oq=python+nltk&gs_l=psy-ab.3..0j0i203l2j0j0i203l6.1867661.1868738..1869035...0.0..0.311.750.0j3j0j1......0....1..gws-wiz.......0i71j0i67.eLLHBxPtulQ"))
except ImportError:
    print("Error:url_extracror requires the urlextract module.")
    urlextract = None

9. 对邮件内的所有单词进行计数

from sklearn.base import BaseEstimator, TransformerMixin

class EmailToWordCounterTransformer(BaseEstimator, TransformerMixin):
    def __init__(self, strip_headers=True, low_case=True, remove_punctuation=True,
                 repalce_urls=True, replace_numbers=True, stemming=True):
        self.strip_headers = strip_headers
        self.low_case = low_case
        self.remove_punctuation = remove_punctuation
        self.replace_urls = repalce_urls
        self.replace_numbers = replace_numbers
        self.stemming = stemming

    def fit(self, X, Y=None):
        return self

    def transform(self, X, Y=None):
        X_transform = []
        for email in X:
            text = email_to_text(email) or ""
            if self.low_case:
                text = text.lower()
            if self.replace_urls and url_extracror is not None:
                # list(set()) 创建一个不重复的元素集
                urls = list(set(url_extracror.find_urls(text)))
                urls.sort(key=lambda url: len(url),
                          reverse=True)  # 根据url的长度对url进行排序
                for url in urls:
                    text = text.replace(url, "URL")  # 用“URL”换所有真实的url
            if self.replace_numbers:               # 将所有数字转换为NUMBER字符
                text = re.sub(r'\d+(?:\.\d*(?:[eE]\d+))?', 'NUMBER', text)
            if self.remove_punctuation:            # 删除所有标点符号
                text = re.sub(r'\W+', ' ', text, flags=re.M)  # \W 匹配任何非单词字符
            # Counter()返回一个特殊的字典，包含单词种类和单词数量。eg:{"a":3,"b""2}
            word_count = Counter(text.split())
            if self.stemming and stemmer is not None:
                stemmed_word_counts = Counter()
                for word, count in word_count.items():       # 分析单词的词干，统计词干的数量
                    stemmed_word = stemmer.stem(word)
                    stemmed_word_counts[stemmed_word] += count
                word_count = stemmed_word_counts
            X_transform.append(word_count)                  # 将每个邮件的字符字典存到list中
        return np.array(X_transform)

X_few = x_train[:3]
X_few_wordcounts = EmailToWordCounterTransformer().fit_transform(X_few)
print(X_few_wordcounts)

from scipy.sparse import csr_matrix                         # 压缩稀疏行矩阵

class WordCounterToVectorTransformer(BaseEstimator, TransformerMixin):
    def __init__(self, vocabulary_size=100):
        self.vocabulary_size = vocabulary_size

    def fit(self, X, Y=None):
        total_count = Counter()
        for word_count in X:
            for word, count in word_count.items():           # X是上个函数内的字典，不是X数据集
                total_count[word] += min(count, 10)            # 次数超过10的存10
        # most_common 字典里面出现次数最多的.当most_common没有参数时，返回字典所有的item，从大到小排列
        # 查看前vocabulaty_size个出现次数最多的
        most_common = total_count.most_common()[:self.vocabulary_size]
        self.most_common_ = most_common
        # most_commoon [('number', 15), ('i', 7), ('the', 7), ('url', 7), ('to', 4), ('chri', 3), ('wa', 3), ('from', 3), ('list', 3), ('of', 3)]
        # 将most_common里面的出现频率最多的词从多到少依次排序，返回{(单词,序号)}
        self.vocabulary_ = {word: index + 1 for index,
                            (word, count) in enumerate(most_common)}
        # vocabulary {'number': 1, 'i': 2, 'the': 3, 'url': 4, 'to': 5, 'chri': 6, 'wa': 7, 'from': 8, 'list': 9, 'of': 10}
        return self

    def transform(self, X, Y=None):
        rows = []
        cols = []
        data = []
        for row, word_count in enumerate(X):
            for word, count in word_count.items():
                rows.append(row)
                cols.append(self.vocabulary_.get(word, 0))
                data.append(count)
        return csr_matrix((data, (rows, cols)), shape=(len(X), self.vocabulary_size+1))

vocab_transformer = WordCounterToVectorTransformer(vocabulary_size=10)
X_few_vectors = vocab_transformer.fit_transform(X_few_wordcounts)
print(X_few_wordcounts)
print(X_few_vectors.toarray())
print(vocab_transformer.vocabulary_)

10.训练分类器

from sklearn.pipeline import Pipeline                       # 创建流水线处理
preprocess_pipeline = Pipeline([
    ("email_to_wordcount", EmailToWordCounterTransformer()),
    ("wordcount_to_vector", WordCounterToVectorTransformer()),
])
X_train_transformed = preprocess_pipeline.fit_transform(x_train)

from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import cross_val_score
log_clf = LogisticRegression()
score = cross_val_score(log_clf, X_train_transformed,
                        y_train, cv=3, verbose=3, n_jobs=-1)
print(score.mean())

11.评估分类器

from sklearn.metrics import precision_score, recall_score
X_test_transformed = preprocess_pipeline.transform(x_test)
# solver 优化算法的参数，包括newton-cg,lbfgs,liblinear,sag,saga,对损失的优化的方法
log_clf2 = LogisticRegression(solver="liblinear", random_state=42)
log_clf2.fit(X_train_transformed, y_train)
y_pred = log_clf2.predict(X_test_transformed)
print(precision_score(y_test, y_pred))
print(recall_score(y_test, y_pred))

from sklearn.naive_bayes import MultinomialNB
mnb = MultinomialNB()
mnb.fit(X_train_transformed, y_train)
mnb_y_pred = mnb.predict(X_test_transformed)
print(precision_score(y_test, mnb_y_pred))
print(recall_score(y_test, mnb_y_pred))