对爬取京东商品按照标题为其进行自动分类---基于逻辑回归的文本分类

1.首先先从京东爬取商品作为训练数据集（以京东的商品分类作为分类的基础）

 

 

 

 复制该类别的链接，写入JD_type文件中（文件如下图所示）

 

 

 爬取训练集代码如下：

import requests
from lxml import etree
import time
import csv
import random
from User_agent import  User_Agent
class Foo:
    cou=0  # 一个user_agent爬取了的页数
    i=0
    agents =User_Agent
    #正在爬取所用'user-agent'
    using_agent='Mozilla/5.0 (Windows NT 6.1; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/66.0.3359.139 Safari/537.36'
# 定义函数抓取第一页以及总页数和第二页的url
def crow_first(url,name):
    print('-------'+name+'-------crow_first---------------------')
    if Foo.cou>=14:#如果爬取的页数等于15页数，则换一个user_agent
        if Foo.i==35:#如果所有的agent都用过来了，则再重新用一遍
            Foo.i=0
        Foo.using_agent=Foo.agents[Foo.i]
        Foo.cou=0
        Foo.i=Foo.i+1
    # 获取当前的Unix时间戳，并且保留小数点后5位
    a = time.time()
    b = '%.5f' % a
    #url = 'https://list.jd.com/list.html?cat=737,1276,739'
    print(url)
    head = {'authority': 'search.jd.com',
            'method': 'GET',
            'scheme': 'https',
            'user-agent': Foo.using_agent,
            'x-requested-with': 'XMLHttpRequest',
            'Cookie': 'qrsc=3; pinId=RAGa4xMoVrs; xtest=1210.cf6b6759; ipLocation=%u5E7F%u4E1C; _jrda=5; TrackID=1aUdbc9HHS2MdEzabuYEyED1iDJaLWwBAfGBfyIHJZCLWKfWaB_KHKIMX9Vj9_2wUakxuSLAO9AFtB2U0SsAD-mXIh5rIfuDiSHSNhZcsJvg; shshshfpa=17943c91-d534-104f-a035-6e1719740bb6-1525571955; shshshfpb=2f200f7c5265e4af999b95b20d90e6618559f7251020a80ea1aee61500; cn=0; 3AB9D23F7A4B3C9B=QFOFIDQSIC7TZDQ7U4RPNYNFQN7S26SFCQQGTC3YU5UZQJZUBNPEXMX7O3R7SIRBTTJ72AXC4S3IJ46ESBLTNHD37U; ipLoc-djd=19-1607-3638-3638.608841570; __jdu=930036140; user-key=31a7628c-a9b2-44b0-8147-f10a9e597d6f; areaId=19; __jdv=122270672|direct|-|none|-|1529893590075; PCSYCityID=25; mt_xid=V2_52007VwsQU1xaVVoaSClUA2YLEAdbWk5YSk9MQAA0BBZOVQ0ADwNLGlUAZwQXVQpaAlkvShhcDHsCFU5eXENaGkIZWg5nAyJQbVhiWR9BGlUNZwoWYl1dVF0%3D; __jdc=122270672; shshshfp=72ec41b59960ea9a26956307465948f6; rkv=V0700; __jda=122270672.930036140.-.1529979524.1529984840.85; __jdb=122270672.1.930036140|85.1529984840; shshshsID=f797fbad20f4e576e9c30d1c381ecbb1_1_1529984840145'
            }
    r = requests.get(url, headers=head)
    r.encoding = 'utf-8'
    html1 = etree.HTML(r.text)
    datas = html1.xpath('//li[contains(@class,"gl-item")]')
    with open('洗发护发/JD_'+name+'.csv', 'a', newline='', encoding='utf-8-sig')as f:
        write = csv.writer(f)
        for data in datas:
            p_name = data.xpath('div/div[@class="p-name"]/a/em/text()')
            if p_name!=[]:
                cc=''.join(p_name[0]).split()
                n_name=""
                for c in cc:
                    n_name=n_name+c
                p_urll = data.xpath('div/div[@class="p-name"]/a/@href')
                print(n_name)
                if n_name!="":
                  write.writerow([n_name,name])
    f.close()
    count=html1.xpath('//div[contains(@class,"f-pager")]//span[contains(@class,"fp-text")]//i//text()')#总页数
    next_page=html1.xpath('//a[contains(@class,"pn-next")]//@href')#下一页的url
    pages=''+count[0]
    next_url="https://list.jd.com"+next_page[0]
    Foo.cou = Foo.cou + 1
    return pages+'  '+next_url


# 定义函数抓取第二页以及以后的商品信息
def crow_then(url,name):
    # 获取当前的Unix时间戳，并且保留小数点后5位
    a = time.time()
    b = '%.5f' % a
    if Foo.cou>=14:#如果爬取的页数等于15页数，则换一个user_agent
        if Foo.i == 35:  # 如果所有的agent都用过来了，则再重新用一遍
            Foo.i = 0
        Foo.using_agent=Foo.agents[Foo.i]
        Foo.cou=0
        Foo.i=Foo.i+1
    #url = 'https://list.jd.com/list.html?cat=737,1276,739&page=3&sort=sort_rank_asc&trans=1&JL=6_0_0#J_main'
    head = {'authority': 'search.jd.com',
            'method': 'GET',
            'scheme': 'https',
            'user-agent': 'Mozilla/5.0 (Windows NT 6.1; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/66.0.3359.139 Safari/537.36',
            'x-requested-with': 'XMLHttpRequest',
            'Cookie': 'qrsc=3; pinId=RAGa4xMoVrs; xtest=1210.cf6b6759; ipLocation=%u5E7F%u4E1C; _jrda=5; TrackID=1aUdbc9HHS2MdEzabuYEyED1iDJaLWwBAfGBfyIHJZCLWKfWaB_KHKIMX9Vj9_2wUakxuSLAO9AFtB2U0SsAD-mXIh5rIfuDiSHSNhZcsJvg; shshshfpa=17943c91-d534-104f-a035-6e1719740bb6-1525571955; shshshfpb=2f200f7c5265e4af999b95b20d90e6618559f7251020a80ea1aee61500; cn=0; 3AB9D23F7A4B3C9B=QFOFIDQSIC7TZDQ7U4RPNYNFQN7S26SFCQQGTC3YU5UZQJZUBNPEXMX7O3R7SIRBTTJ72AXC4S3IJ46ESBLTNHD37U; ipLoc-djd=19-1607-3638-3638.608841570; __jdu=930036140; user-key=31a7628c-a9b2-44b0-8147-f10a9e597d6f; areaId=19; __jdv=122270672|direct|-|none|-|1529893590075; PCSYCityID=25; mt_xid=V2_52007VwsQU1xaVVoaSClUA2YLEAdbWk5YSk9MQAA0BBZOVQ0ADwNLGlUAZwQXVQpaAlkvShhcDHsCFU5eXENaGkIZWg5nAyJQbVhiWR9BGlUNZwoWYl1dVF0%3D; __jdc=122270672; shshshfp=72ec41b59960ea9a26956307465948f6; rkv=V0700; __jda=122270672.930036140.-.1529979524.1529984840.85; __jdb=122270672.1.930036140|85.1529984840; shshshsID=f797fbad20f4e576e9c30d1c381ecbb1_1_1529984840145'

            }
    r = requests.get(url, headers=head)
    r.encoding = 'utf-8'
    html1 = etree.HTML(r.text)
    datas = html1.xpath('//li[contains(@class,"gl-item")]')
    with open('洗发护发/JD_' + name + '.csv', 'a', newline='', encoding='utf-8-sig')as f:
        write = csv.writer(f)
        for data in datas:
            p_name = data.xpath('div/div[@class="p-name"]/a/em/text()')
            cc=''.join(p_name[0]).split()
            n_name = ""
            for c in cc:
                n_name = n_name + c
            p_urll = data.xpath('div/div[@class="p-name"]/a/@href')
            if n_name != "":
              print(n_name)
              write.writerow([n_name,name])
    f.close()
    Foo.cou=Foo.cou+1

def getMessage(urll,namee):
    pages_url=crow_first(urll,namee)
    str=pages_url.split('  ')
    pages=int(str[0])
    next_url=str[1]
    next=next_url.split("&")
    for i in range(2,pages):
        try:
           url=next[0]+'&page='+'%d'%i+'&'+next[2]+'&'+next[3]+'&'+next[4]
           print(namee+'   url:   ' +url)
           crow_then(url,namee)
           print('   Finish')
        except Exception as e:
           print(e)
        i=i+1
if __name__ == '__main__':
    f = open(r"JD_type.txt", 'r',encoding='utf-8')
    s = f.read()
    f.close()
    # 切割文件中的字符串
    zifuchuan = s.split("\n");  # 按行分割
    i = 0
    url = []  # 第一页网址
    name = []  # type
    for ss in zifuchuan:
        if ss != '':  # 去掉空行
            zifu = ss.split("\t")
            url.append(zifu[0])
            name.append(zifu[1])
            #print(i,":","https:"+zifu[0] + "   " + zifu[1])
            getMessage(zifu[0],zifu[1])
            i=i+1

该爬取算法会按每个类别生成一个csv文件，总的需要自己将所有csv文件合成一个训练文件（此处不做赘述了）

2.对数据进行预处理（去停用词、分词）训练集和测试集都要进行预处理。

import csv
import re
import codecs
import string
title=[]
with open('good_test.csv', 'r', encoding='utf-8') as file_handler:
    # 返回csv迭代器file_reader，用于迭代得到样本
    file_reader = csv.reader(file_handler)
    # 通过迭代器，遍历csv文件中的所有样本
    for sample in file_reader:
        title.append(sample)
# 将所有商品标题转换为list

# 对每个标题进行分词，使用jieba分词
import jieba
import pandas as pd
title_s = []
for line in title:

    #去除数字
    line[1]=result = re.sub(r'\d+', '', line[1])

    title_cut = jieba.lcut(line[1])
    title_s.append([line[0],title_cut])

# 导入停用此表
stopwords = [line.strip() for line in open('stop_words.txt', 'r', encoding='utf-8').readlines()]

# 剔除停用词


'''
  数据预处理之后将测试集存入good_test1.csv文件
'''
file_csv = codecs.open("good_test1.csv", 'w', encoding='utf-8-sig')  # 追加
writer = csv.writer(file_csv)
title_clean = []
x=0
for line in title_s:
    line_clean = []
    after_title=""
    print(line[0])
    for word in line[1]:
        if word not in stopwords:
            line_clean.append(word)
            after_title=after_title+" "+word
    title_clean.append(line_clean)
    writer.writerow([line[0],after_title])
    x=x+1

3.基于逻辑回归的自动分类（逻辑回归的原理在此不做赘述）

import pandas as pd  # 是python的一个数据分析包，该工具是为了解决数据分析任务而创建的。Pandas 纳入了大量库和一些标准的数据模型，提供了高效地操作大型数据集所需的工具。
from sklearn.linear_model import LogisticRegression
from sklearn.feature_extraction.text import CountVectorizer

df_train = pd.read_csv('good_train4.csv')
df_test = pd.read_csv('good_test1.csv')
# df_train.drop(columns=[' article', 'id'],
#               inplace=True)  # .drop()删除指定列，inplace可选参数。如果手动设定为True（默认为False），那么原数组直接就被替换，不需要进行重新赋值。
# df_test.drop(columns=['article'], inplace=True)

vectorizer = CountVectorizer(ngram_range=(1, 1), min_df=0.0, max_df=1.0, max_features=100000)
# x_train=vectorizer.fit_transform(df_train['word_seg'])等同于以下两行代码。注：fit_transform(X)拟合模型，并返回文本矩阵
vectorizer.fit(df_train['word_seg'])
x_train = vectorizer.transform(df_train['word_seg'])
x_test = vectorizer.transform(df_test['word_seg'])
y_train = df_train['class'] - 1


lg = LogisticRegression(penalty='l2', dual=False, tol=0.0001, C=1.0,fit_intercept=True, intercept_scaling=1, class_weight=None, random_state=None,solver='liblinear', max_iter=100, multi_class='ovr', verbose=0,warm_start=False, n_jobs=1)
lg.fit(x_train, y_train)

"""根据训练好的分类型对测试集的样本进行预测"""
y_test = lg.predict(x_test)
print(y_test)
# test_accuracy = lg.score(x_test, y_test)
# print(test_accuracy)
"""保存预测结果至本地"""
df_test['class'] = y_test.tolist()  # tolist()将数组或者矩阵转换成列表
df_test['class'] = df_test['class'] + 1
df_result = df_test.loc[:, ['id', 'class']]
df_result.to_csv('result2.csv', index=False)
print("finish..................")

在此就完成了，准确率为81.5%