【實戰(zhàn)】用Python進行10w+QQ說說數(shù)據(jù)分析

• 開發(fā)環(huán)境：win7下的python3.5、MySQL5.7

• 編輯器：pycharm2017.1、ipython，Navicat for mysql

• 需要的python第三方庫：selenium、PIL、Requests、MySQLdb、csv、pandas、numpy、matplotlib、jieba、wordcloud另外還用到了無頭瀏覽器PhantomJS。

• 通過selenium+phantomjs模擬登錄qq空間取到cookies和g_qzonetoken，并算出gtk

• 通過Requests庫利用前面得到的url參數(shù)，構造http請求

• 分析請求得到的響應，是一個json，利用正則表達式提取字段

• 設計數(shù)據(jù)表，并將提取到的字段插入到數(shù)據(jù)庫中?
  

• 通過qq郵箱中的導出聯(lián)系人功能，把好友的qq號導出到一個csv文件，遍歷所有的qq號爬取所有的說說

• 通過sql查詢和ipython分析數(shù)據(jù)，并將數(shù)據(jù)可視化

• 通過python的第三方庫jieba、wordcloud基于說說的內(nèi)容做一個詞云?

import?re
from?selenium import?webdriver
from?time import?sleep
from?PIL import?Image

#定義登錄函數(shù)
def?QR_login():
????def?getGTK(cookie):
????????""" 根據(jù)cookie得到GTK """
????????hashes = 5381
????????for?letter in?cookie['p_skey']:
????????????hashes += (hashes << 5) + ord(letter)

return?hashes & 0x7fffffff
????browser=webdriver.PhantomJS(executable_path="D:\phantomjs.exe")#這里要輸入你的phantomjs所在的路徑
????url="https://qzone.qq.com/"#QQ登錄網(wǎng)址
????browser.get(url)
????browser.maximize_window()#全屏
????sleep(3)#等三秒
????browser.get_screenshot_as_file('QR.png')#截屏并保存圖片
????im = Image.open('QR.png')#打開圖片
????im.show()#用手機掃二維碼登錄qq空間
????sleep(20)#等二十秒，可根據(jù)自己的網(wǎng)速和性能修改
????print(browser.title)#打印網(wǎng)頁標題
????cookie = {}#初始化cookie字典
????for?elem in?browser.get_cookies():#取cookies
????????cookie[elem['name']] = elem['value']
print('Get the cookie of QQlogin successfully!(共%d個鍵值對)'?% (len(cookie)))
????html = browser.page_source#保存網(wǎng)頁源碼
????g_qzonetoken=re.search(r'window\.g_qzonetoken = \(function\(\)\{ try\{return (.*?);\} catch\(e\)',html)#從網(wǎng)頁源碼中提取g_qzonetoken
????gtk=getGTK(cookie)#通過getGTK函數(shù)計算gtk
????browser.quit()
return?(cookie,gtk,g_qzonetoken.group(1))
if?__name__=="__main__":
????QR_login()

def?parse_mood(i):
????'''從返回的json中，提取我們想要的字段'''
????text = re.sub('"commentlist":.*?"conlist":', '', i)
if?text:
????????myMood = {}
????????myMood["isTransfered"] = False
????????tid = re.findall('"t1_termtype":.*?"tid":"(.*?)"', text)[0] # 獲取說說ID
????????tid = qq + '_'?+ tid
????????myMood['id'] = tid
????????myMood['pos_y'] = 0
????????myMood['pos_x'] = 0
????????mood_cont = re.findall('\],"content":"(.*?)"', text)
if?re.findall('},"name":"(.*?)",', text):
????????????name = re.findall('},"name":"(.*?)",', text)[0]
????????????myMood['name'] = name
if?len(mood_cont) == 2: # 如果長度為2則判斷為屬于轉載
????????????myMood["Mood_cont"] = "評語:"?+ mood_cont[0] + "--------->轉載內(nèi)容:"?+ mood_cont[1] # 說說內(nèi)容
????????????myMood["isTransfered"] = True
????????elif?len(mood_cont) == 1:
????????????myMood["Mood_cont"] = mood_cont[0]
else:
????????????myMood["Mood_cont"] = ""
????????if?re.findall('"created_time":(\d+)', text):
????????????created_time = re.findall('"created_time":(\d+)', text)[0]
????????????temp_pubTime = datetime.datetime.fromtimestamp(int(created_time))
????????????temp_pubTime = temp_pubTime.strftime("%Y-%m-%d %H:%M:%S")
????????????dt = temp_pubTime.split(' ')
????????????time = dt[1]
????????????myMood['time'] = time
????????????date = dt[0]
????????????myMood['date'] = date
if?re.findall('"source_name":"(.*?)"', text):
????????????source_name = re.findall('"source_name":"(.*?)"', text)[0] # 獲取發(fā)表的工具（如某手機）
????????????myMood['tool'] = source_name
if?re.findall('"pos_x":"(.*?)"', text):#獲取經(jīng)緯度坐標
????????????pos_x = re.findall('"pos_x":"(.*?)"', text)[0]
????????????pos_y = re.findall('"pos_y":"(.*?)"', text)[0]
if?pos_x:
????????????????myMood['pos_x'] = pos_x
if?pos_y:
????????????????myMood['pos_y'] = pos_y
????????????idname = re.findall('"idname":"(.*?)"', text)[0]
????????????myMood['idneme'] = idname
????????????cmtnum = re.findall('"cmtnum":(.*?),', text)[0]
????????????myMood['cmtnum'] = cmtnum
return?myMood#返回一個字典

CREATE?TABLE?`mood`?(
`name`?varchar(80) DEFAULT?NULL,
`date`?date?DEFAULT?NULL,
`content`?text,
`comments_num`?int(11) DEFAULT?NULL,
`time`?time?DEFAULT?NULL,
`tool`?varchar(255) DEFAULT?NULL,
`id`?varchar(255) NOT?NULL,
`sitename`?varchar(255) DEFAULT?NULL,
`pox_x`?varchar(30) DEFAULT?NULL,
`pox_y`?varchar(30) DEFAULT?NULL,
`isTransfered`?double?DEFAULT?NULL,
PRIMARY KEY?(`id`)
) ENGINE=InnoDB?DEFAULT?CHARSET=utf8;

#從csv文件中取qq號，并保存在一個列表中
csv_reader = csv.reader(open('qq.csv'))
friend=[]
for?row in?csv_reader:
????friend.append(row[3])
friend.pop(0)
friends=[]
for?f in?friend:
????f=f[:-7]
????friends.append(f)
headers={
'Host': 'h5.qzone.qq.com',
????'User-Agent': 'Mozilla/5.0 (Windows NT 6.1; WOW64; rv:53.0) Gecko/20100101 Firefox/53.0',
????'Accept': '*/*',
????'Accept-Language':'zh-CN,zh;q=0.8,en-US;q=0.5,en;q=0.3',
????'Accept-Encoding': 'gzip, deflate, br',
????'Referer': 'https://user.qzone.qq.com/790178228?_t_=0.22746974226377736',
????'Connection':'keep-alive'
}#偽造瀏覽器頭
conn = MySQLdb.connect('localhost', 'root', '123456', 'qq_mood', charset="utf8", use_unicode=True)#連接mysql數(shù)據(jù)庫
cursor = conn.cursor()#定義游標
cookie,gtk,qzonetoken=QRlogin#通過登錄函數(shù)取得cookies，gtk，qzonetoken
s=requests.session()#用requests初始化會話
for?qq in?friends:#遍歷qq號列表
????for?p in?range(0,1000):
????????pos=p*20
????????params={
'uin':qq,
????????'ftype':'0',
????????'sort':'0',
????????'pos':pos,
????????'num':'20',
????????'replynum':'100',
????????'g_tk':gtk,
????????'callback':'_preloadCallback',
????????'code_version':'1',
????????'format':'jsonp',
????????'need_private_comment':'1',
????????'qzonetoken':qzonetoken
????????}

????????response=s.request('GET','https://h5.qzone.qq.com/proxy/domain/taotao.qq.com/cgi-bin/emotion_cgi_msglist_v6',params=params,headers=headers,cookies=cookie)
print(response.status_code)#通過打印狀態(tài)碼判斷是否請求成功
????????text=response.text#讀取響應內(nèi)容
????????if?not?re.search('lbs', text):#通過lbs判斷此qq的說說是否爬取完畢
????????????print('%s說說下載完成'% qq)
break
????????textlist = re.split('\{"certified"', text)[1:]
for?i in?textlist:
????????????myMood=parse_mood(i)
'''將提取的字段值插入mysql數(shù)據(jù)庫，通過用異常處理防止個別的小bug中斷爬蟲，開始的時候可以先不用異常處理判斷是否能正常插入數(shù)據(jù)庫'''
????????????try:
????????????????insert_sql = '''
???????????????????????????insert into mood(id,content,time,sitename,pox_x,pox_y,tool,comments_num,date,isTransfered,name)
???????????????????????????VALUES (%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s)
????????????????????????'''
????????????????cursor.execute(insert_sql, (myMood['id'],myMood["Mood_cont"],myMood['time'],myMood['idneme'],myMood['pos_x'],myMood['pos_y'],myMood['tool'],myMood['cmtnum'],myMood['date'],myMood["isTransfered"],myMood['name']))
????????????????conn.commit()
except:
????????????????pass
print('說說全部下載完成！')

#coding:utf-8
import?matplotlib.pyplot as?plt
from?wordcloud import?WordCloud,ImageColorGenerator,STOPWORDS
import?jieba
import?numpy as?np
from?PIL import?Image

#讀入背景圖片
abel_mask = np.array(Image.open("qq.jpg"))

#讀取要生成詞云的文件
text_from_file_with_apath = open('mood.txt',encoding='utf-8').read()

#通過jieba分詞進行分詞并通過空格分隔
wordlist_after_jieba = jieba.cut(text_from_file_with_apath, cut_all = True)
stopwords = {'轉載','內(nèi)容','em','評語','uin','nick'}
seg_list = [i for?i in?wordlist_after_jieba if?i not?in?stopwords]
wl_space_split = " ".join(seg_list)
#my_wordcloud = WordCloud().generate(wl_space_split) 默認構造函數(shù)
my_wordcloud = WordCloud(
background_color='black', # 設置背景顏色
????????????mask = abel_mask, # 設置背景圖片
????????????max_words = 250, # 設置最大現(xiàn)實的字數(shù)
????????????stopwords = STOPWORDS, # 設置停用詞
????????????font_path = 'C:/Windows/fonts/simkai.ttf',# 設置字體格式，如不設置顯示不了中文
????????????max_font_size = 42, # 設置字體最大值
????????????random_state = 40, # 設置有多少種隨機生成狀態(tài)，即有多少種配色方案
????????????????scale=1.5,
????????????mode='RGBA',
????????????relative_scaling=0.6
????????????????).generate(wl_space_split)

# 根據(jù)圖片生成詞云顏色
#image_colors = ImageColorGenerator(abel_mask)
#my_wordcloud.recolor(color_func=image_colors)

# 以下代碼顯示圖片
plt.imshow(my_wordcloud)
plt.axis("off")
plt.show()

my_wordcloud.to_file("cloud.jpg")

• 爬蟲沒有采用多線程和異步IO導致效率太低。（主要是twisted這個庫不太懂，后面我可能會結合scapy這個框架，重寫這個爬蟲，利用他的twisted模塊加上異步IO的功能）

• 對于python中的關于繪圖的，和數(shù)據(jù)分析的這幾個庫了解的不好，導致數(shù)據(jù)可視化這塊做的不好。