Analyze Tweets of the French Presidential Election

Posted on Dim 23 septembre 2018 in Data Analysis

French Presidential Election : Tweets (03/02/2017 - 26/02/2017)

This Dataset comes from the Twitter API, these data were collected during 10 minutes every hour, during three weeks. The stream filtered tweets by the mention of the name of some candidates. You can find the dataset here : https://www.kaggle.com/jeanmidev/french-presidential-election

Let's go to Analyze these data

In [1]:
%matplotlib inline
import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
import sqlite3
import datetime

Converting Sqlite database to Pandas dataframe

In [2]:
connection = sqlite3.connect("database.sqlite")
df = pd.read_sql_query("select * from data;", connection)
connection.close()
df.head()
Out[2]:
index lang mention_Fillon mention_Hamon mention_Le Pen mention_Macron mention_Mélenchon text timestampms
0 0 fr 1 0 0 0 0 RT @AllezFillon: Bonjour @jccambadelis, pensez... 1486481509767
1 1 fr 1 0 0 0 0 @Framilon @GeoffroyDidier @FrancoisFillon @enm... 1486481509866
2 2 fr 1 0 0 0 0 RT @PaulAveline: Le Canard Enchaîné révèle que... 1486481509939
3 3 fr 0 1 0 0 0 RT @chirpichirp: @RTenfrancais @FredElNacional... 1486481510001
4 4 fr 0 0 0 0 0 RT @beaulieu37: Ha ha. La fameuse "transparenc... 1486481510041
In [3]:
lines = df.shape[0]
print("Number of Tweets : " + str(lines))

Number of Tweets : 1067316

Parsing Dates

In [4]:
# Disable Warning 'SettingWithCopyWarning'
pd.options.mode.chained_assignment = None
df_tweets = df[["timestampms","mention_Fillon","mention_Hamon",
         "mention_Le Pen","mention_Macron",
         "mention_Mélenchon"]]

df_tweets["datetime"] = pd.to_numeric(df_tweets["timestampms"])
df_tweets["datetime"] = pd.to_datetime(df_tweets["datetime"], unit='ms')

df_tweets['year'] = df_tweets.datetime.dt.year
df_tweets['month'] = df_tweets.datetime.dt.month
df_tweets['day'] = df_tweets.datetime.dt.day

df_tweets["datetime"] = pd.to_datetime(df_tweets[['day', 'month', 'year']])

df_tweets = df_tweets.drop(["timestampms"], axis = 1)
df_tweets = df_tweets[["datetime","mention_Fillon","mention_Hamon",
         "mention_Le Pen","mention_Macron", "mention_Mélenchon"]]
df_tweets.head()
Out[4]:
datetime mention_Fillon mention_Hamon mention_Le Pen mention_Macron mention_Mélenchon
0 2017-02-07 1 0 0 0 0
1 2017-02-07 1 0 0 0 0
2 2017-02-07 1 0 0 0 0
3 2017-02-07 0 1 0 0 0
4 2017-02-07 0 0 0 0 0

Reindexing

In [5]:
df_tweets = df_tweets.set_index('datetime')
# Remove the row of the index name
df_tweets.index.name = None

Group by Dates

In [6]:
group_tweets = df_tweets.groupby(df_tweets.index).sum()
group_tweets
Out[6]:
mention_Fillon mention_Hamon mention_Le Pen mention_Macron mention_Mélenchon
2017-02-03 9063 1586 2001 3575 731
2017-02-04 17873 3406 6143 23855 1520
2017-02-05 17055 9265 12938 19952 6027
2017-02-06 60475 5018 10701 13374 3694
2017-02-07 40861 4027 6906 11048 1605
2017-02-08 25195 3592 6344 10463 1051
2017-02-09 22497 4056 10199 12924 1063
2017-02-10 17653 2639 10035 9894 880
2017-02-11 13544 2968 5048 8020 875
2017-02-12 13013 2980 5064 9055 1281
2017-02-13 11400 3272 5736 11849 973
2017-02-14 16318 2902 5171 12539 1318
2017-02-15 19308 3525 6965 13836 2193
2017-02-16 19412 3309 8006 19640 2031
2017-02-17 16048 3814 7001 16814 1635
2017-02-18 11221 4824 4529 18935 1260
2017-02-19 11085 5838 4890 11042 1902
2017-02-20 10157 4465 7860 9541 2344
2017-02-21 14725 5866 22097 15741 2367
2017-02-22 4899 2113 9043 7446 894
2017-02-26 305 205 371 740 61
In [7]:
group_tweets.plot()
axes = plt.gca()

plt.axvline(x= datetime.datetime(year = 2017, month = 2, day = 6), linestyle="dashed", color="r")
plt.text(datetime.datetime(year = 2017, month = 2, day = 6), y = axes.get_ylim()[1],
         s="06/02: Press Conference of Fillon,\nHe apologized to hired his wife", fontsize=9)

plt.axvline(x= datetime.datetime(year = 2017, month = 2, day = 21), linestyle="dashed", color="r")
plt.text(datetime.datetime(year = 2017, month = 2, day = 21), y = axes.get_ylim()[1],
         s="21/02: Marine Le Pen refuse \n to wear Muslim headscarf", fontsize=9)
         
plt.show()

Data Analysis