随着OpenAI大火,从事AI开发的人趋之若鹜,这次使用Python selenium抓取了领英上几万条岗位薪资数据,并使用Pandas、matplotlib、seaborn等库进行可视化探索分析。
但领英设置了一些反爬措施,对IP进行限制封禁,因此会用到IP代理,用不同的IP进行访问,我这里用的是亮数据的IP代理。
亮数据是一家提供网络数据采集解决方案的网站,它拥有全球最大的代理IP网络,覆盖超过195个国家和地区,拥有超过7200万个不重复的真人IP地址。
这些IP地址可以用于匿名浏览网页、绕过IP封锁、抓取网页数据等。
亮数据官网地址:
https://get.brightdata.com/weijun
另外,亮数据提供各种数据采集工具,帮助企业轻松采集网页数据。这些工具包括Web Scraper IDE、亮数据浏览器、SERP API等等。
下面是关于Python爬取领英的步骤和代码。
- 1、爬虫采集AI岗位数据-selenium&亮数据
- 2、处理和清洗数据-pandas
- 3、可视化数据探索-matplotlib seaborn
1、爬虫采集AI岗位数据-selenium&亮数据
python
# 导入相关库
import random
from selenium import webdriver
from selenium.webdriver.common.by import By
import time
import requests
import pandas as pd
from scripts.helpers import strip_val, get_value_by_path
# 选择Edge浏览器
BROWSER = 'edge'
# 创建网络会话,登录Linkedin
# create_session函数用于创建一个自动化的浏览器会话,并使用提供的电子邮件和密码登录LinkedIn。
# 它首先根据BROWSER变量选择相应的浏览器驱动程序(Chrome或Edge),然后导航到LinkedIn的登录页面,自动填写登录表单,并提交。
# 登录成功后,它会获取当前会话的cookies,并创建一个requests.Session对象来保存这些cookies,以便后续的HTTP请求可以保持登录状态。最后,它返回这个会话对象。
def create_session(email, password):
if BROWSER == 'chrome':
driver = webdriver.Chrome()
elif BROWSER == 'edge':
driver = webdriver.Edge()
# 登录信息
driver.get('https://www.linkedin.com/checkpoint/rm/sign-in-another-account')
time.sleep(1)
driver.find_element(By.ID, 'username').send_keys(email)
driver.find_element(By.ID, 'password').send_keys(password)
driver.find_element(By.XPATH, '//*[@id="organic-div"]/form/div[3]/button').click()
time.sleep(1)
input('Press ENTER after a successful login for "{}": '.format(email))
driver.get('https://www.linkedin.com/jobs/search/?')
time.sleep(1)
cookies = driver.get_cookies()
driver.quit()
session = requests.Session()
for cookie in cookies:
session.cookies.set(cookie['name'], cookie['value'])
return session
# 获取登录账号和密码
def get_logins(method):
logins = pd.read_csv('logins.csv')
logins = logins[logins['method'] == method]
emails = logins['emails'].tolist()
passwords = logins['passwords'].tolist()
return emails, passwords
# JobSearchRetriever类用于检索LinkedIn上的职位信息。
# 它初始化时设置了一个职位搜索链接,并获取登录凭证来创建多个会话。
# 它还定义了一个get_jobs方法,该方法通过会话发送HTTP GET请求到LinkedIn的职位搜索API,获取职位信息,并解析响应以提取职位ID和标题。
# 如果职位被标记为赞助(即广告),它也会记录下来。
class JobSearchRetriever:
def __init__(self):
self.job_search_link = 'https://www.linkedin.com/voyager/api/voyagerJobsDashJobCards?decorationId=com.linkedin.voyager.dash.deco.jobs.search.JobSearchCardsCollection-187&count=100&q=jobSearch&query=(origin:JOB_SEARCH_PAGE_OTHER_ENTRY,selectedFilters:(sortBy:List(DD)),spellCorrectionEnabled:true)&start=0'
emails, passwords = get_logins('search')
self.sessions = [create_session(email, password) for email, password in zip(emails, passwords)]
self.session_index = 0
self.headers = [{
'Authority': 'www.linkedin.com',
'Method': 'GET',
'Path': 'voyager/api/voyagerJobsDashJobCards?decorationId=com.linkedin.voyager.dash.deco.jobs.search.JobSearchCardsCollection-187&count=25&q=jobSearch&query=(origin:JOB_SEARCH_PAGE_OTHER_ENTRY,selectedFilters:(sortBy:List(DD)),spellCorrectionEnabled:true)&start=0',
'Scheme': 'https',
'Accept': 'application/vnd.linkedin.normalized+json+2.1',
'Accept-Encoding': 'gzip, deflate, br',
'Accept-Language': 'en-US,en;q=0.9',
'Cookie': "; ".join([f"{key}={value}" for key, value in session.cookies.items()]),
'Csrf-Token': session.cookies.get('JSESSIONID').strip('"'),
# 'TE': 'Trailers',
'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36',
# 'X-Li-Track': '{"clientVersion":"1.12.7990","mpVersion":"1.12.7990","osName":"web","timezoneOffset":-7,"timezone":"America/Los_Angeles","deviceFormFactor":"DESKTOP","mpName":"voyager-web","displayDensity":1,"displayWidth":1920,"displayHeight":1080}'
'X-Li-Track': '{"clientVersion":"1.13.5589","mpVersion":"1.13.5589","osName":"web","timezoneOffset":-7,"timezone":"America/Los_Angeles","deviceFormFactor":"DESKTOP","mpName":"voyager-web","displayDensity":1,"displayWidth":360,"displayHeight":800}'
} for session in self.sessions]
# self.proxies = [{'http': f'http://{proxy}', 'https': f'http://{proxy}'} for proxy in []]
# 添加亮数据代理IP
# get_jobs函数用于发送HTTP请求到LinkedIn的职位搜索API,获取职位信息
# 它使用当前会话索引来选择一个会话,并发送带有相应请求头的GET请求。如果响应状态码是200(表示请求成功)
# 它将解析JSON响应,提取职位ID、标题和赞助状态,并将这些信息存储在一个字典中。
def get_jobs(self):
results = self.sessions[self.session_index].get(self.job_search_link, headers=self.headers[self.session_index]) #, proxies=self.proxies[self.session_index], timeout=5)
self.session_index = (self.session_index + 1) % len(self.sessions)
if results.status_code != 200:
raise Exception('Status code {} for search\nText: {}'.format(results.status_code, results.text))
results = results.json()
job_ids = {}
for r in results['included']:
if r['$type'] == 'com.linkedin.voyager.dash.jobs.JobPostingCard' and 'referenceId' in r:
job_id = int(strip_val(r['jobPostingUrn'], 1))
job_ids[job_id] = {'sponsored': False}
job_ids[job_id]['title'] = r.get('jobPostingTitle')
for x in r['footerItems']:
if x.get('type') == 'PROMOTED':
job_ids[job_id]['sponsored'] = True
break
return job_ids
2、处理和清洗数据-pandas
python
# 导入相关库
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from wordcloud import WordCloud
# 导入职位数据
job_postings = pd.read_csv('./archive/job_postings.csv')
job_postings
# 根据AI岗位关键词筛选AI相关岗位
keywords = ['data scientist', 'machine learning', 'data science', 'data analyst', 'ml engineer',' data engineer','ai engineer','ai/ml','ai/nlp','ai reasearcher','ai consultant','artificial intelligence','computer vision','deep learning']
# 新增一列,标注职位是否包含关键字
def check_keywords(description):
for keyword in keywords:
if keyword in str(description).lower():
return 'AI岗位'
return '非AI岗位'
job_postings['is_programmer'] = job_postings['description'].apply(check_keywords)
# 保存AI岗位新表
job_ai = job_postings[(job_postings['is_programmer']=='AI岗位') & (job_postings['pay_period']=='YEARLY') & (job_postings['max_salary']>10000) ]
job_others = job_postings[(job_postings['is_programmer']=='非AI岗位') & (job_postings['pay_period']=='YEARLY') & (job_postings['max_salary']>10000) & (job_postings['max_salary']<200000) ]
job_ai
处理好的数据如下:
3、可视化数据探索-matplotlib seaborn
AI岗位中位数年薪18W美金,最高50w以上
python
# 设置Seaborn样式和调色板
sns.set_style("whitegrid")
palette = ["skyblue"]
# palette = ["#87CEEB"] # 使用颜色代码或者其他有效的颜色名称,这里使用天蓝色的颜色代码
# 箱线图
plt.figure(figsize=(8, 6))
sns.boxplot(y='max_salary', data=job_ai, palette=palette)
plt.ylabel('Yearly Salary')
plt.title('AI Yearly Salary Boxplot')
# 添加分位数标注
quantiles = job_ai['max_salary'].quantile([0.25, 0.5, 0.75])
for q, label in zip(quantiles, ['Q1', 'Median', 'Q3']):
plt.text(0, q, f'{label}: {int(q)}', horizontalalignment='center', verticalalignment='bottom', fontdict={'size': 10})
# 添加平均值、最大最小值标注
avg_value = job_ai['max_salary'].mean()
max_value = job_ai['max_salary'].max()
min_value = job_ai['max_salary'].min()
plt.text(0.2, avg_value, f'Avg: {int(avg_value)}', ha='left', va='bottom', fontdict={'size': 10})
plt.text(0, max_value, f'Max: {int(max_value)}', ha='center', va='bottom', fontdict={'size': 10})
plt.text(0, min_value, f'Min: {int(min_value)}', ha='center', va='top', fontdict={'size': 10})
# 显示图形
plt.show()
AI岗位年薪主要集中在15-30w美金
python
# 1. 直方图
plt.figure(figsize=(10, 6))
plt.hist(job_ai['max_salary'], bins=30, color='skyblue', edgecolor='black')
plt.xlabel('Yearly Salary')
plt.ylabel('Frequency')
plt.title('Yearly Salary Distribution')
plt.show()
AI大多需要高级岗,对软件开发、机器学习、数据科学要求较多
python
# 词云
stopwords = set(["Manager"])
job_titles_text = ' '.join(job_ai['title'])
wordcloud = WordCloud(width=800, height=400, background_color='white',stopwords=stopwords).generate(job_titles_text)
# 显示词云
plt.figure(figsize=(10, 6))
plt.imshow(wordcloud, interpolation='bilinear')
plt.title('AI Job Title Word Cloud')
plt.axis('off')
plt.tight_layout()
plt.show()
数据发现,AI岗位平均年薪竟高达18万美金,远超普通开发岗,而且AI岗位需求也在爆发性增长。
这次使用的是亮数据IP服务,质量还是蛮高的,大家可以试试。
亮数据官网地址: