使用 Flask 的 Twitter 情绪分析 WebApp

这是一个使用Python和 Flask 框架制作的网络应用程序。它有一个注册系统和一个仪表板。用户可以输入关键字来检索基于关键字的实时 Twitter 文本，并对其进行分析以获取客户的感受和情绪。这些数据可以在图表中可视化。特别是，该项目使用流行的“Tweepy”API 挖掘数据。 Tweepy API 实时连接到 Twitter，并从 Twitter 平台收集元数据和文本。

目的：

• 帮助公司研究围绕特定产品的客户情绪。
• 帮助系统用户快速高效地分析海量数据。
• 以 -1 到 1 的比例区分客户情绪，其中 -1 表示对关键字的强烈负面情绪，1 表示强烈的正面反应。
• 清晰有效地可视化收集到的数据。

详细项目实施视频：

项目中使用的工具和技术：

硬件：

• 4GB 内存
• 视窗 10
• 用于部署的 Google Chrome 网络浏览器
• 用于连接到 Twitter 的 Internet 连接

本项目设计是基于软件的，因此硬件设计要求包括一台运行良好的 64 位 PC，具有至少 4 GB 的 RAM。

软件：

• Python- Python是一种解释型、高级和通用的编程语言。 Python了代码的可管理性和可读性，使其成为用于机器学习的顶级应用程序之一。
• Flask Web 框架（与Python）- Flask 是一个 Web 框架。这意味着 Flask 为我们提供了工具、库和技术，使我能够构建 Web 应用程序和网页。 Flask 是一个后端微框架，它使数据处理变得干净和简单。
• Tweepy（ Python 的Twitter API）- Tweepy 是一个开源Python包，它为您提供了一种使用Python访问 Twitter API 的非常方便的方法。 Tweepy 包括一组表示 Twitter 模型和 API 端点的类和方法，它透明地处理各种实现细节，例如数据编码和解码。
• 适用于 Windows 的 Xampp 服务器上的 Apache SQL - SQL 服务器在本地主机内的 phpMyAdmin 上运行。该数据库用于存储、验证和检索用户的登录凭据。
• Bootstrap - 在 Bootstrap 的帮助下增强了 UI，它有助于构建现代、直观和响应式的网页。
• JQuery - JQuery 用于从 Tweepy API 来回发送和检索数据，并将结果显示在网页上。
• HTML/CSS- HTML 和 CSS 是网站前端设计的基础。

构建项目所需的技能：

• 机器学习 - 中等
• 编程技巧
• Python- 高级。
• HTML/CSS/JQUERY/BOOTSTRAP-基本到中等。
• SQL/DBMS-基本到中等
• 调试/部署 – 中等
• 使用 API 的能力。

实施项目：

按照以下步骤实施项目：

步骤 1：在您的系统上下载并安装 Xampp 服务器。

第 2 步：启动 Apache Xampp 和 MySQL。在 MySQL 中创建一个包含 3 列（用户名、电子邮件 ID 和密码）的数据库。

第 3 步：下载并安装 PyCharm。单击“创建”->“新建项目”。为您的项目命名。

第 4 步：在 Pycharm 中键入以下代码。有关详细的项目文件和层次结构，请参阅此 GitHub 存储库。

main.py 文件：

在这个文件中，我们首先初始化我们的项目。我们使用“conn”对象建立到我们的 SQL 数据库的连接。

我们设置了一个用户 cookie 变量，在将他/她重定向到主页之前，它会检查用户是否已登录。该脚本还处理登录/注册页面上的用户输入。

Python3

from flask import Flask, render_template, request, redirect, session
import mysql.connector
from sentiments import second
import os
 
 
app = Flask(__name__)
 
# initializing the user cookie
app.secret_key = os.urandom(24)
 
# blueprint to call the second python file in the project.
app.register_blueprint(second)
 
# establishing a connection with mysql database made in xampp
try:
    conn = mysql.connector.connect(
        host="localhost", user="root", password="", database="users")
    cursor = conn.cursor()
except:
    print("An exception occured")
 
# call the login template when the url is http://localhost:5000/
@app.route('/')
def login():
    return render_template('login.html')
 
# call the register template when the url is http://localhost:5000/register
@app.route('/register')
def register():
    return render_template('register.html')
 
 
@app.route('/home')
def home():
    if 'user_id' in session:
        return render_template('home.html')
    else:
        return redirect('/')
 
 
@app.route('/login_validation', methods=['POST'])
def login_validation():
    email = request.form.get('email')
    password = request.form.get('password')
 
    cursor.execute(
        """SELECT * from `users` WHERE `email` LIKE '{}' AND `password` LIKE '{}'""".format(email, password))
    users = cursor.fetchall()
    # check if a user has already logged in
    if len(users) > 0:
        session['user_id'] = users[0][0]
        return redirect('/home')
    else:
        return redirect('/login')
 
 
@app.route('/add_user', methods=['POST'])
def add_user():
   
  # get user login data and pass the data to database
    name = request.form.get('uname')
    email = request.form.get('uemail')
    password = request.form.get('upassword')
    cursor.execute("""INSERT INTO `users` (`name`,`email`,`password`) VALUES ('{}','{}','{}')""".format(
        name, email, password))
    conn.commit()
    cursor.execute(
        """SELECT * from `users` WHERE `email` LIKE '{}'""".format(email))
    myuser = cursor.fetchall()
    session['user_id'] = myuser[0][0]
    return redirect('/home')
 
 
@app.route('/logout')
def logout():
  # close the session
    session.pop('user_id')
    return redirect('/')
 
 
if __name__ == "__main__":
    app.run(debug=True)

情绪.py

这是我们项目中的第二个Python文件，它在我们的“main.py”文件中注册为蓝图。

此代码从 HTML 页面获取用户输入，该页面指定要搜索的关键字以及要查看的推文数量。然后它连接到 Tweepy API，后者检索推文文本。将文本清理干净，然后将其分类为不同的情绪（-1 是愤怒/悲伤的情绪，1 是快乐的情绪）。

在此基础上，创建一个饼图并保存为图像。此图像稍后会在其他 HTML 页面中调用。

蟒蛇3

from flask import Blueprint, render_template, request
import matplotlib.pyplot as plt
import os
 
import tweepy
import csv
import re
from textblob import TextBlob
import matplotlib
matplotlib.use('agg')
 
# register this file as a blueprint
second = Blueprint("second", __name__, static_folder="static",
                   template_folder="template")
 
# render page when url is called
@second.route("/sentiment_analyzer")
def sentiment_analyzer():
    return render_template("sentiment_analyzer.html")
 
 
# class with main logic
class SentimentAnalysis:
 
    def __init__(self):
        self.tweets = []
        self.tweetText = []
 
    # This function first connects to the Tweepy API using API keys
    def DownloadData(self, keyword, tweets):
 
        # authenticating
        consumerKey = '//get from Tweepy'
        consumerSecret = '//get from Tweepy'
        accessToken = '//insert your access token here'
        accessTokenSecret = '//Tweepy AccessToken secret here'
        auth = tweepy.OAuthHandler(consumerKey, consumerSecret)
        auth.set_access_token(accessToken, accessTokenSecret)
        api = tweepy.API(auth, wait_on_rate_limit=True)
 
        # input for term to be searched and how many tweets to search
        # searchTerm = input("Enter Keyword/Tag to search about: ")
        # NoOfTerms = int(input("Enter how many tweets to search: "))
        tweets = int(tweets)
 
        # searching for tweets
        self.tweets = tweepy.Cursor(
            api.search, q=keyword, lang="en").items(tweets)
 
        # Open/create a file to append data to
        csvFile = open('result.csv', 'a')
 
        # Use csv writer
        csvWriter = csv.writer(csvFile)
 
        # creating some variables to store info
        polarity = 0
        positive = 0
        wpositive = 0
        spositive = 0
        negative = 0
        wnegative = 0
        snegative = 0
        neutral = 0
 
        # iterating through tweets fetched
        for tweet in self.tweets:
           
            # Append to temp so that we can store in csv later. I use encode UTF-8
            self.tweetText.append(self.cleanTweet(tweet.text).encode('utf-8'))
             
            # print (tweet.text.translate(non_bmp_map))    #print tweet's text
            analysis = TextBlob(tweet.text)
             
            # print(analysis.sentiment)  # print tweet's polarity
            # adding up polarities to find the average later
            polarity += analysis.sentiment.polarity
 
            # adding reaction of how people are reacting to find average later
            if (analysis.sentiment.polarity == 0):
                neutral += 1
            elif (analysis.sentiment.polarity > 0 and analysis.sentiment.polarity <= 0.3):
                wpositive += 1
            elif (analysis.sentiment.polarity > 0.3 and analysis.sentiment.polarity <= 0.6):
                positive += 1
            elif (analysis.sentiment.polarity > 0.6 and analysis.sentiment.polarity <= 1):
                spositive += 1
            elif (analysis.sentiment.polarity > -0.3 and analysis.sentiment.polarity <= 0):
                wnegative += 1
            elif (analysis.sentiment.polarity > -0.6 and analysis.sentiment.polarity <= -0.3):
                negative += 1
            elif (analysis.sentiment.polarity > -1 and analysis.sentiment.polarity <= -0.6):
                snegative += 1
 
        # Write to csv and close csv file
        csvWriter.writerow(self.tweetText)
        csvFile.close()
 
        # finding average of how people are reacting
        positive = self.percentage(positive, tweets)
        wpositive = self.percentage(wpositive, tweets)
        spositive = self.percentage(spositive, tweets)
        negative = self.percentage(negative, tweets)
        wnegative = self.percentage(wnegative, tweets)
        snegative = self.percentage(snegative, tweets)
        neutral = self.percentage(neutral, tweets)
 
        # finding average reaction
        polarity = polarity / tweets
 
        # printing out data
        #  print("How people are reacting on " + keyword + " by analyzing " + str(tweets) + " tweets.")
        #  print()
        #  print("General Report: ")
 
        if (polarity == 0):
            htmlpolarity = "Neutral"
 
        # print("Neutral")
        elif (polarity > 0 and polarity <= 0.3):
            htmlpolarity = "Weakly Positive"
            # print("Weakly Positive")
        elif (polarity > 0.3 and polarity <= 0.6):
            htmlpolarity = "Positive"
        elif (polarity > 0.6 and polarity <= 1):
            htmlpolarity = "Strongly Positive"
        elif (polarity > -0.3 and polarity <= 0):
            htmlpolarity = "Weakly Negative"
        elif (polarity > -0.6 and polarity <= -0.3):
            htmlpolarity = "Negative"
        elif (polarity > -1 and polarity <= -0.6):
            htmlpolarity = "strongly Negative"
 
        self.plotPieChart(positive, wpositive, spositive, negative,
                          wnegative, snegative, neutral, keyword, tweets)
        print(polarity, htmlpolarity)
        return polarity, htmlpolarity, positive, wpositive, spositive, negative, wnegative, snegative, neutral, keyword, tweets
 
    def cleanTweet(self, tweet):
        # Remove Links, Special Characters etc from tweet
        return ' '.join(re.sub("(@[A-Za-z0-9]+)|([^0-9A-Za-z \t]) | (\w +:\ / \ / \S +)", " ", tweet).split())
 
    # function to calculate percentage
    def percentage(self, part, whole):
        temp = 100 * float(part) / float(whole)
        return format(temp, '.2f')
 
    # function which sets and plots the pie chart. The chart is saved in an img file every time the project is run.
    # The previous image is overwritten. This image is called in the html page.
 
    def plotPieChart(self, positive, wpositive, spositive, negative, wnegative, snegative, neutral, keyword, tweets):
        fig = plt.figure()
        labels = ['Positive [' + str(positive) + '%]', 'Weakly Positive [' + str(wpositive) + '%]',
                  'Strongly Positive [' + str(spositive) +
                  '%]', 'Neutral [' + str(neutral) + '%]',
                  'Negative [' + str(negative) +
                  '%]', 'Weakly Negative [' + str(wnegative) + '%]',
                  'Strongly Negative [' + str(snegative) + '%]']
        sizes = [positive, wpositive, spositive,
                 neutral, negative, wnegative, snegative]
        colors = ['yellowgreen', 'lightgreen', 'darkgreen',
                  'gold', 'red', 'lightsalmon', 'darkred']
        patches, texts = plt.pie(sizes, colors=colors, startangle=90)
        plt.legend(patches, labels, loc="best")
        plt.axis('equal')
        plt.tight_layout()
        strFile = r"C:\Users\LENOVO\PycharmProjects\SentimentAnalysis\static\images\plot1.png"
        if os.path.isfile(strFile):
            os.remove(strFile)  # Opt.: os.system("rm "+strFile)
        plt.savefig(strFile)
        plt.show()
 
 
@second.route('/sentiment_logic', methods=['POST', 'GET'])
def sentiment_logic():
   
  # get user input of keyword to search and number of tweets from html form.
    keyword = request.form.get('keyword')
    tweets = request.form.get('tweets')
    sa = SentimentAnalysis()
     
    # set variables which can be used in the jinja supported html page
    polarity, htmlpolarity, positive, wpositive, spositive, negative, wnegative, snegative, neutral, keyword1, tweet1 = sa.DownloadData(
        keyword, tweets)
    return render_template('sentiment_analyzer.html', polarity=polarity, htmlpolarity=htmlpolarity, positive=positive, wpositive=wpositive, spositive=spositive,
                           negative=negative, wnegative=wnegative, snegative=snegative, neutral=neutral, keyword=keyword1, tweets=tweet1)
 
 
@second.route('/visualize')
def visualize():
    return render_template('PieChart.html')

SentimentAnalyzer.py 模板。 （为程序的主要逻辑呈现结果的模板）

此模板要求用户输入他们感兴趣的主题/词，以及用户想要分析的基于该主题的推文数量。

请注意，Twitter 有每日和每小时的速率限制，不能超过。

此表单将数据传递到 second.py 文件，该文件计算输出并设置“jinja”变量。 （Jinja 允许在Python和 HTML 之间传递值）。

然后这些变量将输出显示给用户。

HTML

   
      
      
      
      
      
      
      
      
      Home
   
   
      
      
         Twitter Sentiment Analysis
         
         
         
         
            Get started
            Home
            Visualize
            Help
         
         
            
               Sign out
            
         
      
      
         
            
               
                  
                      Enter your search keyword 
                      
                      Enter your number of tweets to analyze 
                     
                     
                  
                  
                   
 
 Need help?  Click here 
 
 
               
            
         
      
      
      
      
      
      
      -----------------------------------------------------------------------------------
      
         
            {% if polarity %}
            How are people reacting on {{keyword}} by analyzing {{tweets}} Tweets
            {% endif %}
         
      
      
      
         
            
            
               
                  General Report
                  
                     {% if polarity %}
                      The Average Sentiment is {{htmlpolarity}}  
                     {%endif%}
                  
               
            
            
            
            
               
                  Sentiment Polarity
                  
                     {% if polarity %}
                      The sentiment polarity is {{polarity}} 
                     {%endif%}
                  
               
            
            
         
         
      
      
         Detailed Report
         
            {% if polarity %}
             {{spositive}} "% people thought it was strongly positive"
             {{positive}} "% people thought it was positive"
             {{wpositive}} "% people thought it was weakly positive"
             {{neutral}} "% people thought it was neutral"
             {{negative}} "% people thought it was negative"
             {{wnegative}} "% people thought it was weakly negative"
             {{snegative}} "% people thought it was strongly negative"
            {%endif%}
         
      
      
       Generate Visualization
      
      
      
      
      
      
   

登录模板：

显示登录表单的页面的 HTML 代码。用户输入具有验证的用户名和密码。验证数据后，这些变量将传递到 main.py 文件，该文件连接到 SQL 数据库以检查登录是否有效。

HTML

   
      
      
      
      
      
      
      
      
      
      
      
      Login
   
   
   
      
         Sentiment Analysis
      
      
      
         
            
                Welcome to this Twitter Sentiment Analysis! 
            
            
               
                  
                     
                     
                        
                            Email 
                            
                           
                           Your email must be a valid email
                           
                        
                        
                            Password 
                           
                           
                              **Please Fill the password
                           
                        
                        
                     
                     
                      
 
 Not a member?  Create Account
 
 
                  
               
            
         
      
      
      
   

注册模板：

显示注册表单的页面的 HTML 代码。用户输入具有验证的用户名和密码。此外，用户名成为在 main.py 文件中设置的 cookie 的值。

验证数据后，这些变量将传递到 main.py 文件，该文件连接到 SQL 数据库以检查登录是否有效。

HTML

   
      
      
      
      
      
      
      Registration
   
   
      
         Sentiment Analysis App
      
      
         
            
                Welcome to this Twitter Sentiment Analysis! 
            
            
               
                  
                     
                         Name 
                         
                         Email 
                         
                         Password 
                        
                        
                     
                     
                      
 
 Already a member?  Login 
 
 
                  
               
            
         
      
      
   

主页模板：

HTML 主页。该页面有 2 个主要部分

• 导航栏- 导航栏包含所有其他页面的链接，它从 cookie 中检索用户名并显示“欢迎 Sam”[用户名]。此栏还有一个注销按钮。
• Bootstrap carousel -这个 carousel 是基本 carousel 的定制版本，可以在这里找到。

HTML

   
      
      
      
      
      
      
      Home
   
   
      
      
         
            {% if session['user_id'] %}
            Welcome {{session['user_id']}}
            {%endif%}
         
         
         
         
         
            Get started
            Home
            Visualize
            Help
         
         
            
               Sign out
            
         
      
      
         .carousel-item {
         height: 34rem;
         background: black;
         color: white;
         position: relative;
         background-position: center;
         background-size: cover;
         }
         .carousel-caption{
         position: absolute;
         bottom: 0;
         left: 0;
         right: 0;
         padding-bottom: 10px;
         }
         .overlay-image{
         position: absolute;
         bottom: 0;
         left: 0;
         right: 0;
         padding-bottom: 0px;
         background-position: center;
         background-size: cover;
         opacity: 0.4;
         }
      
      
      
         
            
            
            
         
         
            
               
                  
                     
                  
                  
                     
                        Let's get started!
                     
                     
                      Click Here to Analyze Some Tweets
                  
               
               
                  
                     
                  
                  
                     
                         How does this work? 
                     
                     
                      Click Here to know more about this project
                  
               
               
                  
                     
                  
                  
                     
                         WorkFlow
                     
                     
                     
                        Enter a keyword of your choice 
                        Enter the no of tweets you want to search 
                        Voila! See instant Results! 
                        Click on Generate to View Detailed Data Visualization.
                     
                  
               
            
            
            
            Previous
            
            
            
            Next
            
         
      
      
      
      
   

饼图生成模板：

此 HTML 模板仅显示保存在我们项目中的图像文件。这个图像文件是在 second.py 文件代码中生成的。图像是一个饼图，它直观地表示了情感分析的结果。每次运行代码时都会覆盖图像。

HTML

   
      
      Data Visualization
      
      
      
      
   
   
   
      
         Twitter Sentiment Analysis
         
         
         
         
            Get started
            Home
            Visualize
            Help
         
         
            
               Sign out
            
         
      
      
         
      
       ** Last generated visual. Press CTRL+ Refresh to reload
       Back 
      
      
   

详细项目实施指南：

1. 用户通过单击“创建帐户”注册页面验证创建一个具有所有必需验证的新帐户。通过使用 cookie 可以防止 URL 重写。
2. 成功注册/登录后，用户将被定向到主页。
3. 主页有一个引导轮播滑块。在左上角，检索 cookie 并显示登录用户的名称。此 cookie 还可以防止网站劫持（如果用户未正确登录，则用户无法直接访问 /home URL）。
4. 主页欢迎用户使用他们的注册名称，使用 cookie，
5. 用户可以使用顶部的导航栏进行导航。单击“开始”以转到主模块屏幕。
6. 用户输入关键字和要分析的推文数量。这两个字段都需要填写。用户点击分析。系统连接到 Tweepy，获取最新的推文，分析它们，并在相应的字段中显示结果。
7. 用户可以单击报告底部的生成可视化按钮。这将根据报告生成饼图。
8. 当用户点击“这个项目是如何工作的”时，他们被重定向到 tweepy 文档。
9. 用户可以单击“注销”以注销并结束他们的会话。

用例图

现实生活中的项目应用：

Twitter 每月有 3.3 亿活跃用户，这使企业能够在没有中介的情况下接触到广泛的受众并与客户建立联系。不利的一面是，信息太多，品牌很难快速发现可能损害其业务的负面社交提及。

这就是为什么情感分析（涉及监控社交媒体平台上对话中的情绪）已成为社交媒体营销的关键策略。倾听客户在 Twitter 上的感受可以让公司了解他们的受众，掌握关于他们的品牌和竞争对手的评论，并发现行业的新趋势。公司可以迅速采取行动。这个想法甚至可以扩展到普通用户，在那里可以使用本地关键字和主题标签来分析 Twitter 上文本的情绪。