Dependencies

import tweepy import numpy as np import pandas as pd import matplotlib.pyplot as plt from numpy.random import rand from itertools import cycle

Import and Initialize Sentiment Analyzer

from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer analyzer = SentimentIntensityAnalyzer()

Twitter API Keys

from config import (consumer_key, consumer_secret, access_token, access_token_secret)

Setup Tweepy API Authentication

auth = tweepy.OAuthHandler(consumer_key, consumer_secret) auth.set_access_token(access_token, access_token_secret) api = tweepy.API(auth, parser=tweepy.parsers.JSONParser())

Target User Account

target_user1 = "@bbcworld"

Variables for holding sentiments

compound_list = [] positive_list = [] negative_list = [] neutral_list = []

Loop through 10 pages of tweets (total 100 tweets)

for x in range(5):

# Get all tweets from home feed
public_tweets = api.user_timeline(target_user1)

# Loop through all tweets
for tweet in public_tweets:

    # Run Vader Analysis on each tweet
    results = analyzer.polarity_scores(tweet["text"])
    compound = results["compound"]
    pos = results["pos"]
    neu = results["neu"]
    neg = results["neg"]

    # Add each value to the appropriate list
    compound_list.append(compound)
    positive_list.append(pos)
    negative_list.append(neg)
    neutral_list.append(neu)
    
    # Print the Averages

print(f"User: {target_user1}") print(f"Compound: {np.mean(compound_list):.3f}") print(f"Positive: {np.mean(positive_list):.3f}") print(f"Neutral: {np.mean(neutral_list):.3f}") print(f"Negative: {np.mean(negative_list):.3f}")

Convert sentiments to DataFrame

bbc_news = pd.DataFrame( {'Compound': compound_list, 'Positive': positive_list, 'Neutral': neutral_list, 'Negative': negative_list }) bbc_news["Tweets Ago"] = range(len(bbc_news))
bbc_news["name"] = "BBC" bbc_news.head()

Target User Account

target_user2 = "@cbsnews"

Variables for holding sentiments

compound_list2 = [] positive_list2 = [] negative_list2 = [] neutral_list2 = []

Loop through 10 pages of tweets (total 100 tweets)

for x in range(5):

# Get all tweets from home feed
public_tweets2 = api.user_timeline(target_user2)

# Loop through all tweets
for tweet in public_tweets:

    # Run Vader Analysis on each tweet
    results2 = analyzer.polarity_scores(tweet["text"])
    compound = results2["compound"]
    pos = results2["pos"]
    neu = results2["neu"]
    neg = results2["neg"]

    # Add each value to the appropriate list
    compound_list2.append(compound)
    positive_list2.append(pos)
    negative_list2.append(neg)
    neutral_list2.append(neu)
    
    # Print the Averages

print(f"User: {target_user2}") print(f"Compound: {np.mean(compound_list2):.3f}") print(f"Positive: {np.mean(positive_list2):.3f}") print(f"Neutral: {np.mean(neutral_list2):.3f}") print(f"Negative: {np.mean(negative_list2):.3f}")

Convert sentiments to DataFrame

cbs_news = pd.DataFrame( {'Compound': compound_list2, 'Positive': positive_list2, 'Neutral': neutral_list2, 'Negative': negative_list2 }) cbs_news["Tweets Ago"] = range(len(cbs_news))
cbs_news["name"] = "CBS" cbs_news.head()

Target User Account

target_user3 = "@cnn"

Variables for holding sentiments

compound_list3 = [] positive_list3 = [] negative_list3 = [] neutral_list3 = []

Loop through 10 pages of tweets (total 100 tweets)

for x in range(5):

# Get all tweets from home feed
public_tweets3 = api.user_timeline(target_user3)

# Loop through all tweets
for tweet in public_tweets3:

    # Run Vader Analysis on each tweet
    results3 = analyzer.polarity_scores(tweet["text"])
    compound = results3["compound"]
    pos = results3["pos"]
    neu = results3["neu"]
    neg = results3["neg"]

    # Add each value to the appropriate list
    compound_list3.append(compound)
    positive_list3.append(pos)
    negative_list3.append(neg)
    neutral_list3.append(neu)
    
    # Print the Averages

print(f"User: {target_user3}") print(f"Compound: {np.mean(compound_list3):.3f}") print(f"Positive: {np.mean(positive_list3):.3f}") print(f"Neutral: {np.mean(neutral_list3):.3f}") print(f"Negative: {np.mean(negative_list3):.3f}")

Convert sentiments to DataFrame

cnn_news = pd.DataFrame( {'Compound': compound_list3, 'Positive': positive_list3, 'Neutral': neutral_list3, 'Negative': negative_list3 }) cnn_news["Tweets Ago"] = range(len(cnn_news))
cnn_news["name"] = "CNN" cnn_news.head()

Target User Account

target_user4 = "@foxnews"

Variables for holding sentiments

compound_list4 = [] positive_list4 = [] negative_list4 = [] neutral_list4 = []

Loop through 10 pages of tweets (total 100 tweets)

for x in range(5):

# Get all tweets from home feed
public_tweets4 = api.user_timeline(target_user4)

# Loop through all tweets
for tweet in public_tweets4:

    # Run Vader Analysis on each tweet
    results4 = analyzer.polarity_scores(tweet["text"])
    compound = results4["compound"]
    pos = results4["pos"]
    neu = results4["neu"]
    neg = results4["neg"]

    # Add each value to the appropriate list
    compound_list4.append(compound)
    positive_list4.append(pos)
    negative_list4.append(neg)
    neutral_list4.append(neu)
    
    # Print the Averages

print(f"User: {target_user4}") print(f"Compound: {np.mean(compound_list4):.3f}") print(f"Positive: {np.mean(positive_list4):.3f}") print(f"Neutral: {np.mean(neutral_list4):.3f}") print(f"Negative: {np.mean(negative_list4):.3f}")

Convert sentiments to DataFrame

fox_news = pd.DataFrame( {'Compound': compound_list4, 'Positive': positive_list4, 'Neutral': neutral_list4, 'Negative': negative_list4 }) fox_news["Tweets Ago"] = range(len(fox_news))
fox_news["name"] = "Fox News" fox_news.head()

Target User Account

target_user5 = "@nytimes"

Variables for holding sentiments

compound_list5 = [] positive_list5 = [] negative_list5 = [] neutral_list5 = []

Loop through 10 pages of tweets (total 100 tweets)

for x in range(5):

# Get all tweets from home feed
public_tweets5 = api.user_timeline(target_user5)

# Loop through all tweets
for tweet in public_tweets5:

    # Run Vader Analysis on each tweet
    results5 = analyzer.polarity_scores(tweet["text"])
    compound = results5["compound"]
    pos = results5["pos"]
    neu = results5["neu"]
    neg = results5["neg"]

    # Add each value to the appropriate list
    compound_list5.append(compound)
    positive_list5.append(pos)
    negative_list5.append(neg)
    neutral_list5.append(neu)
    
    # Print the Averages

print(f"User: {target_user5}") print(f"Compound: {np.mean(compound_list5):.3f}") print(f"Positive: {np.mean(positive_list5):.3f}") print(f"Neutral: {np.mean(neutral_list5):.3f}") print(f"Negative: {np.mean(negative_list5):.3f}")

Convert sentiments to DataFrame

nytimes_news = pd.DataFrame( {'Compound': compound_list5, 'Positive': positive_list5, 'Neutral': neutral_list5, 'Negative': negative_list5 }) nytimes_news["Tweets Ago"] = range(len(nytimes_news))
nytimes_news["name"] = "New York Times" nytimes_news.head()

Target User Account

target_user6 = "@theeconomist"

Variables for holding sentiments

compound_list6 = [] positive_list6 = [] negative_list6 = [] neutral_list6 = []

Loop through 10 pages of tweets (total 100 tweets)

for x in range(5):

# Get all tweets from home feed
public_tweets6 = api.user_timeline(target_user6)

# Loop through all tweets
for tweet in public_tweets6:

    # Run Vader Analysis on each tweet
    results6 = analyzer.polarity_scores(tweet["text"])
    compound = results6["compound"]
    pos = results6["pos"]
    neu = results6["neu"]
    neg = results6["neg"]

    # Add each value to the appropriate list
    compound_list6.append(compound)
    positive_list6.append(pos)
    negative_list6.append(neg)
    neutral_list6.append(neu)
    
    # Print the Averages

print(f"User: {target_user6}") print(f"Compound: {np.mean(compound_list6):.3f}") print(f"Positive: {np.mean(positive_list6):.3f}") print(f"Neutral: {np.mean(neutral_list6):.3f}") print(f"Negative: {np.mean(negative_list6):.3f}")

Convert sentiments to DataFrame

economist_news = pd.DataFrame( {'Compound': compound_list6, 'Positive': positive_list6, 'Neutral': neutral_list6, 'Negative': negative_list6 }) economist_news["Tweets Ago"] = range(len(economist_news))
economist_news["name"] = "The Economist" economist_news.head()

Target User Account

target_user7 = "@msnbc"

Variables for holding sentiments

compound_list7 = [] positive_list7 = [] negative_list7 = [] neutral_list7 = []

Loop through 10 pages of tweets (total 100 tweets)

for x in range(5):

# Get all tweets from home feed
public_tweets7 = api.user_timeline(target_user7)

# Loop through all tweets
for tweet in public_tweets7:

    # Run Vader Analysis on each tweet
    results7 = analyzer.polarity_scores(tweet["text"])
    compound = results7["compound"]
    pos = results7["pos"]
    neu = results7["neu"]
    neg = results7["neg"]

    # Add each value to the appropriate list
    compound_list7.append(compound)
    positive_list7.append(pos)
    negative_list7.append(neg)
    neutral_list7.append(neu)
    
    # Print the Averages

print(f"User: {target_user7}") print(f"Compound: {np.mean(compound_list7):.3f}") print(f"Positive: {np.mean(positive_list7):.3f}") print(f"Neutral: {np.mean(neutral_list7):.3f}") print(f"Negative: {np.mean(negative_list7):.3f}")

Convert sentiments to DataFrame

msnbc_news = pd.DataFrame( {'Compound': compound_list7, 'Positive': positive_list7, 'Neutral': neutral_list7, 'Negative': negative_list7 }) msnbc_news["Tweets Ago"] = range(len(msnbc_news))
msnbc_news["name"] = "MSNBC" msnbc_news.head()

Target User Account

target_user8 = "@abc"

Variables for holding sentiments

compound_list8 = [] positive_list8 = [] negative_list8 = [] neutral_list8 = []

Loop through 10 pages of tweets (total 100 tweets)

for x in range(5):

# Get all tweets from home feed
public_tweets8 = api.user_timeline(target_user8)

# Loop through all tweets
for tweet in public_tweets8:

    # Run Vader Analysis on each tweet
    results8 = analyzer.polarity_scores(tweet["text"])
    compound = results8["compound"]
    pos = results8["pos"]
    neu = results8["neu"]
    neg = results8["neg"]

    # Add each value to the appropriate list
    compound_list8.append(compound)
    positive_list8.append(pos)
    negative_list8.append(neg)
    neutral_list8.append(neu)
    
    # Print the Averages

print(f"User: {target_user8}") print(f"Compound: {np.mean(compound_list8):.3f}") print(f"Positive: {np.mean(positive_list8):.3f}") print(f"Neutral: {np.mean(neutral_list8):.3f}") print(f"Negative: {np.mean(negative_list8):.3f}")

Convert sentiments to DataFrame

abc_news = pd.DataFrame( {'Compound': compound_list8, 'Positive': positive_list8, 'Neutral': neutral_list8, 'Negative': negative_list8 }) abc_news["Tweets Ago"] = range(len(abc_news))
abc_news["name"] = "ABC" abc_news.head()

Target User Account

target_user9 = "@newsweek"

Variables for holding sentiments

compound_list9 = [] positive_list9 = [] negative_list9 = [] neutral_list9 = []

Loop through 10 pages of tweets (total 100 tweets)

for x in range(5):

# Get all tweets from home feed
public_tweets9 = api.user_timeline(target_user9)

# Loop through all tweets
for tweet in public_tweets9:

    # Run Vader Analysis on each tweet
    results9 = analyzer.polarity_scores(tweet["text"])
    compound = results9["compound"]
    pos = results9["pos"]
    neu = results9["neu"]
    neg = results9["neg"]

    # Add each value to the appropriate list
    compound_list9.append(compound)
    positive_list9.append(pos)
    negative_list9.append(neg)
    neutral_list9.append(neu)
    
    # Print the Averages

print(f"User: {target_user9}") print(f"Compound: {np.mean(compound_list9):.3f}") print(f"Positive: {np.mean(positive_list9):.3f}") print(f"Neutral: {np.mean(neutral_list9):.3f}") print(f"Negative: {np.mean(negative_list9):.3f}")

Convert sentiments to DataFrame

newsweek_news = pd.DataFrame( {'Compound': compound_list9, 'Positive': positive_list9, 'Neutral': neutral_list9, 'Negative': negative_list9 }) newsweek_news["Tweets Ago"] = range(len(newsweek_news))
newsweek_news["name"] = "NewsWeek" newsweek_news.head()

Target User Account

target_userz = "@washingtonpost"

Variables for holding sentiments

compound_listz = [] positive_listz = [] negative_listz = [] neutral_listz = []

Loop through 10 pages of tweets (total 100 tweets)

for x in range(5):

# Get all tweets from home feed
public_tweetsz = api.user_timeline(target_userz)

# Loop through all tweets
for tweet in public_tweetsz:

    # Run Vader Analysis on each tweet
    resultsz = analyzer.polarity_scores(tweet["text"])
    compound = resultsz["compound"]
    pos = resultsz["pos"]
    neu = resultsz["neu"]
    neg = resultsz["neg"]

    # Add each value to the appropriate list
    compound_listz.append(compound)
    positive_listz.append(pos)
    negative_listz.append(neg)
    neutral_listz.append(neu)

    
    # Print the Averages

print(f"User: {target_userz}") print(f"Compound: {np.mean(compound_listz):.3f}") print(f"Positive: {np.mean(positive_listz):.3f}") print(f"Neutral: {np.mean(neutral_listz):.3f}") print(f"Negative: {np.mean(negative_listz):.3f}")

Convert sentiments to DataFrame

washpost_news = pd.DataFrame( {'Compound': compound_listz, 'Positive': positive_listz, 'Neutral': neutral_listz, 'Negative': negative_listz }) washpost_news["Tweets Ago"] = range(len(washpost_news))
washpost_news["name"] = "Washington Post" washpost_news

#create list of all dataframes all_dfs = [bbc_news, cbs_news, cnn_news, fox_news, nytimes_news, economist_news, msnbc_news, abc_news, newsweek_news, washpost_news]

Give all df's common column names

for df in all_dfs: df.columns = ['Compound', 'Negative', 'Neutral', 'Positive', 'Tweets Ago', 'Name']

#create new dataframe that combines all dataframes
Major_News_Sentiment = pd.concat(all_dfs).reset_index(drop=True) Major_News_Sentiment.head() Major_News_Sentiment.to_csv("MajorNewsSentiment", encoding='utf-8', index=False)

mns_group = Major_News_Sentiment.groupby('Name') mns_group.mean()

Split up our data into groups based upon 'name'

news_names = Major_News_Sentiment.groupby('Name')

Find out how many of each gender took bike trips

news_score = news_names['Compound'].mean()

Chart our data, give it a title, and label the axes

news_chart = news_score.plot(kind="bar", title="Major Network Tweet Sentiment") news_chart.set_xlabel("News Network") news_chart.set_ylabel("Compound Sentiment Score")

fig.savefig('AggregateTweetSentiment.png') plt.show()

Plot and Formatting

x = Major_News_Sentiment['Tweets Ago'] y = Major_News_Sentiment['Compound']

fig, ax = plt.subplots() for news in ['BBC', 'CBS', 'CNN', 'Fox', 'NYTimes', 'Economist', 'MSNBC', 'ABC', 'NewsWeek', 'WashPost']: ax.scatter(x, y, s=20, c=np.random.rand(3,), label=news, alpha=0.5, edgecolors='none') plt.xlabel('Tweets Ago') plt.ylabel('Compound Score') ax.legend() ax.grid(True)

fig2.savefig('TweetSentiment.png') plt.show()

3 Observable Trends

1. New York Times' twitter account is relatively more negative than other news outlets

2. News Outlet twitter accounts tend to be negative in nature but to varying extents

3. ABC and The Economist represent the least negative news outlets when considering their last 100 tweets