
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import string
import seaborn as sns
import nltk
from nltk.stem import WordNetLemmatizer
from nltk.tokenize import RegexpTokenizer
from nltk.corpus import stopwords
import missingno as msno
import re
import miceforest as mf
from wordcloud import WordCloud
from textblob import TextBlob
from scipy import stats
from sklearn.metrics.pairwise import cosine_similarity
from sklearn.metrics import silhouette_score
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.cluster import KMeans
import param
import panel as pn
pn.extension()

df = pd.read_csv("2019-02-25-16-09-36-EST-Historical-Report-2016-01-01--2019-02-26 - 2019-02-25-16-09-36-EST-Historical-Report-2016-01-01--2019-02-26.csv")
df.head()


list(df.columns)

['Page Name',
 'User Name',
 'Page Id',
 'Page Likes at Posting',
 'Created',
 'Type',
 'Likes',
 'Comments',
 'Shares',
 'Love',
 'Wow',
 'Haha',
 'Sad',
 'Angry',
 'Thankful',
 'Video Share Status',
 'Post Views',
 'Total Views',
 'Total Views for all Crossposts',
 'URL',
 'Message',
 'Link',
 'Final Link',
 'Link Text',
 'Description',
 'Sponsor Id',
 'Sponsor Name',
 'Score',
 'Yes']


df = df.drop(['User Name', 'Page Id', 'Link', 'Final Link', 'Sponsor Id', 'Yes', 'Post Views', 'Total Views'], axis = 1)
df.head()


msno.bar(df)

<AxesSubplot:>


df.dtypes

Page Name                          object
Page Likes at Posting              object
Created                            object
Type                               object
Likes                               int64
Comments                            int64
Shares                              int64
Love                                int64
Wow                                 int64
Haha                                int64
Sad                                 int64
Angry                               int64
Thankful                           object
Video Share Status                 object
Total Views for all Crossposts     object
URL                                object
Message                            object
Link Text                          object
Description                        object
Sponsor Name                       object
Score                             float64
dtype: object


misplaced_df = df[df['Created'].str.match('^[0-9]{4}-[0-9]{2}-[0-9]{2} ([0-9]+(:[0-9]+)+) [a-zA-Z]+$')== False]
misplaced_df


missplaced_indices = misplaced_df.index.tolist()
df.loc[missplaced_indices,'Page Likes at Posting'] = misplaced_df.loc[missplaced_indices,'Type']
df.loc[missplaced_indices,'Created'] = misplaced_df.loc[missplaced_indices,'Page Likes at Posting']
df.loc[missplaced_indices,'Type'] = misplaced_df.loc[missplaced_indices,'Created']
df.loc[missplaced_indices,'Score'] = misplaced_df.loc[missplaced_indices,'Sponsor Name']
df.loc[missplaced_indices,'Sponsor Name'] = np.NaN
df.loc[missplaced_indices,'Message'] = misplaced_df.loc[missplaced_indices,'URL']
df.loc[missplaced_indices,'Total Views for all Crossposts'] = 0
df.loc[[3278, 6695, 7208, 9379],'URL'] = misplaced_df.loc[[3278, 6695, 7208, 9379],'Total Views for all Crossposts']
df.loc[[1154, 1928, 3439],'URL'] = misplaced_df.loc[[1154, 1928, 3439],'Message']
df.loc[missplaced_indices,'Message'] = np.NaN
df['Created'] = pd.to_datetime(df['Created']).dt.strftime('%d-%m-%Y')
df.dtypes

C:\Users\tanke\.virtualenvs\test-facebook-A5e32CXY\lib\site-packages\dateutil\parser\_parser.py:1213: UnknownTimezoneWarning: tzname EST identified but not understood.  Pass `tzinfos` argument in order to correctly return a timezone-aware datetime.  In a future version, this will raise an exception.
  warnings.warn("tzname {tzname} identified but not understood.  "
C:\Users\tanke\.virtualenvs\test-facebook-A5e32CXY\lib\site-packages\dateutil\parser\_parser.py:1213: UnknownTimezoneWarning: tzname EDT identified but not understood.  Pass `tzinfos` argument in order to correctly return a timezone-aware datetime.  In a future version, this will raise an exception.
  warnings.warn("tzname {tzname} identified but not understood.  "

Page Name                         object
Page Likes at Posting             object
Created                           object
Type                              object
Likes                              int64
Comments                           int64
Shares                             int64
Love                               int64
Wow                                int64
Haha                               int64
Sad                                int64
Angry                              int64
Thankful                          object
Video Share Status                object
Total Views for all Crossposts    object
URL                               object
Message                           object
Link Text                         object
Description                       object
Sponsor Name                      object
Score                             object
dtype: object


df['Page Likes at Posting'] = pd.to_numeric(df['Page Likes at Posting'])
df['Total Views for all Crossposts'] = pd.to_numeric(df['Total Views for all Crossposts'])
df['Score'] = pd.to_numeric(df['Score'])
df.dtypes

Page Name                          object
Page Likes at Posting             float64
Created                            object
Type                               object
Likes                               int64
Comments                            int64
Shares                              int64
Love                                int64
Wow                                 int64
Haha                                int64
Sad                                 int64
Angry                               int64
Thankful                           object
Video Share Status                 object
Total Views for all Crossposts      int64
URL                                object
Message                            object
Link Text                          object
Description                        object
Sponsor Name                       object
Score                             float64
dtype: object


df[df['Thankful'].str.match('^\d$')== False]['Thankful']

6          10
185       691
239        39
272        38
314        33
973       123
1154    share
1255       12
1700       12
1928    share
2195       63
2848       22
3048       42
3311       10
3739       15
4035       15
4857       70
5844      100
6183       26
8191       37
8482       87
8535       10
9357       26
9502       38
Name: Thankful, dtype: object


df.loc[df.Thankful == 'share', 'Thankful'] = np.NaN
df['Thankful'] = pd.to_numeric(df['Thankful'])
df.dtypes

Page Name                          object
Page Likes at Posting             float64
Created                            object
Type                               object
Likes                               int64
Comments                            int64
Shares                              int64
Love                                int64
Wow                                 int64
Haha                                int64
Sad                                 int64
Angry                               int64
Thankful                          float64
Video Share Status                 object
Total Views for all Crossposts      int64
URL                                object
Message                            object
Link Text                          object
Description                        object
Sponsor Name                       object
Score                             float64
dtype: object


msno.bar(df)

<AxesSubplot:>


impute_columns = ['Page Likes at Posting',
 'Likes',
 'Comments',
 'Shares',
 'Love',
 'Wow',
 'Haha',
 'Sad',
 'Angry',
 'Thankful',
 'Total Views for all Crossposts']
df_imputation = df[impute_columns]
df_imputation.head()


df_amp = mf.ampute_data(df_imputation,perc=0.25,random_state=666)

kernel = mf.MultipleImputedKernel(
  df_amp,
  datasets=4,
  save_all_iterations=True,
  random_state=666
)

kernel.mice(3)

kernel.plot_correlations()


kernel.plot_feature_importance(annot=True,cmap="YlGnBu",vmin=0, vmax=1)

AxesSubplot(0.125,0.125;0.62x0.755)


kernel_df = pd.concat([kernel.complete_data(0), kernel.complete_data(1), kernel.complete_data(2), kernel.complete_data(3)]).groupby(level=0).mean().round(decimals=0)
msno.bar(kernel_df)

<AxesSubplot:>


kernel_df.head()


df = df.drop(impute_columns, axis = 1)
df = df.merge(kernel_df, left_index=True, right_index=True)
msno.bar(df)

<AxesSubplot:>


df['Video Share Status'].unique()

array(['crosspost', 'original', nan, 'share', '0'], dtype=object)


df['Video Share Status'] = np.where(df['Video Share Status'] == '0', np.NaN, df['Video Share Status'])
df['Video Share Status'].unique()

array(['crosspost', 'original', nan, 'share'], dtype=object)


df[df['Video Share Status']=="crosspost"]


df = df.drop(['Video Share Status'], axis = 1)
df['Sponsor Name'].unique()

array([nan, 'Gates Foundation', 'Great Big Story', 'NowThis Future',
       'World Health Organization African Region',
       'American Senior Communities', 'Ministry of Health Zambia',
       'World Health Organization (Eastern Mediterranean Regional Office)',
       'Daily Wire', 'Johnson & Johnson', 'SourceFed', 'Providr',
       'Icahn School of Medicine at Mount Sinai'], dtype=object)


df = df.drop(['Sponsor Name'], axis = 1)
df.head()


df['Message'].fillna("", inplace=True)
df['Link Text'].fillna("", inplace=True)
df['Description'].fillna("", inplace=True)
msno.bar(df)

<AxesSubplot:>


df['Text'] = df['Message'] + df['Link Text'] + df['Description']
df.head()


words = set(nltk.corpus.words.words())
def preprocess_text(text):
    # Tokenise words while ignoring punctuation
    text = re.sub(r'http\S+', '', text)
    text = ''.join([i for i in text if not i.isdigit()])
    tokeniser = RegexpTokenizer(r'\w+')
    tokens = tokeniser.tokenize(text)
    
    # Lowercase and lemmatise 
    lemmatiser = WordNetLemmatizer()
    lemmas = [lemmatiser.lemmatize(token.lower(), pos='v') for token in tokens]
    
    # Remove stopwords
    keywords= [lemma for lemma in lemmas if lemma not in stopwords.words('english')]
    keywords= " ".join(str(x) for x in keywords)
    return " ".join(w for w in nltk.wordpunct_tokenize(keywords) \
                    if w.lower() in words or not w.isalpha())

df.shape[0]

9541


chunk_size = int(df.shape[0] / 10)
chunks = list()
for start in range(0, df.shape[0], chunk_size):
    df_subset = df.iloc[start:start + chunk_size].copy()
    df_subset['Text'] = df_subset['Text'].apply(preprocess_text)
    chunks.append(df_subset)

df = pd.concat(chunks)
df.head()


df[df['Text'].str.len()==0]


df = df[df['Text'].str.len()>0]
tfIdfVectorizer=TfidfVectorizer(use_idf=True)
tfIdf = tfIdfVectorizer.fit_transform(df['Text'])
feature_names = tfIdfVectorizer.get_feature_names()
tfIdf_df = pd.DataFrame(tfIdf[0].T.todense(), index=feature_names, columns=["TF-IDF"])
tfIdf_df = tfIdf_df.sort_values('TF-IDF', ascending=False)
tfIdf_df.head()


dist = 1 - cosine_similarity(tfIdf)
dist

array([[-4.44089210e-16,  9.93218548e-01,  9.89330364e-01, ...,
         9.96673486e-01,  9.29683600e-01,  9.93216475e-01],
       [ 9.93218548e-01,  0.00000000e+00,  2.41630256e-01, ...,
         9.90077156e-01,  8.34622929e-01,  9.94305985e-01],
       [ 9.89330364e-01,  2.41630256e-01, -2.22044605e-16, ...,
         9.95606815e-01,  8.84105697e-01,  9.91041289e-01],
       ...,
       [ 9.96673486e-01,  9.90077156e-01,  9.95606815e-01, ...,
        -2.22044605e-16,  9.97136383e-01,  9.00124760e-01],
       [ 9.29683600e-01,  8.34622929e-01,  8.84105697e-01, ...,
         9.97136383e-01, -2.22044605e-16,  9.94160427e-01],
       [ 9.93216475e-01,  9.94305985e-01,  9.91041289e-01, ...,
         9.00124760e-01,  9.94160427e-01,  0.00000000e+00]])


def polarity(x):
    return TextBlob(x).sentiment[0]

def subjectivity(x):
    return TextBlob(x).sentiment[1]

df['Polarity'] = df['Text'].apply(polarity)
df['Subjectivity'] = df['Text'].apply(subjectivity)

df.head(3)


status_metrics = ['Score', 'Page Likes at Posting', 'Likes', 'Comments',
       'Shares', 'Love', 'Wow', 'Haha', 'Sad', 'Angry', 'Thankful',
       'Total Views for all Crossposts', 'Polarity', 'Subjectivity']
z_scores = np.abs(stats.zscore(df[status_metrics]))
z_scores

array([[5.52201234e+01, 1.44142455e-01, 1.27134407e+01, ...,
        2.90857353e+01, 7.18857996e-01, 7.92476531e-01],
       [4.35356545e+01, 4.47081767e-01, 2.31547289e+01, ...,
        6.19053515e-02, 1.48677659e+00, 8.65307710e-01],
       [3.59702556e+01, 3.01073183e-01, 1.43019694e+01, ...,
        7.17201182e+00, 2.14449418e+00, 1.31493491e+00],
       ...,
       [1.86200013e-01, 3.52983738e-01, 3.04248841e-01, ...,
        2.95152340e-02, 7.23117918e-01, 2.86147209e-01],
       [1.86200013e-01, 1.33031072e-01, 2.83026711e-01, ...,
        6.19053515e-02, 1.00725318e+00, 7.58045095e-01],
       [1.86200013e-01, 1.07252378e-02, 3.44605679e-01, ...,
        6.19053515e-02, 4.01034734e-01, 1.11028531e+00]])


df.shape[0]

9534


df_o = df[(z_scores < 3).all(axis=1)]
df_o.describe()


color = plt.get_cmap('RdYlGn').copy()
plt.figure(figsize=(10,8), dpi= 100)
color.set_bad('lightblue')
sns.heatmap(df_o[status_metrics].corr(), annot=True, cmap=color)

<AxesSubplot:>


class BoxDashboard2(param.Parameterized):
    
    metric = param.ObjectSelector(default='Page Likes at Posting', objects=status_metrics)
    plt.figure(figsize=(12,8), dpi= 100)

    def box_plot_1(self):
        plt.clf()
        sns.boxplot(x=df[self.metric])
        return plt.gcf()

    def box_plot_2(self):
        plt.clf()
        sns.boxplot(x=df_o[self.metric])
        return plt.gcf()

    def append_end(self):
        plt.clf()
        return ""

bd2 = BoxDashboard2(name='')

dashboard = pn.Column('Boxplot Dashboard',
                      bd2.param,
                      pn.Tabs(
                          ('Original', bd2.box_plot_1),
                          ('Outlier removed', bd2.box_plot_2),
                          ('',bd2.append_end)
                      )
                     )

dashboard.embed()

<Figure size 1728x1152 with 0 Axes>


sil = []
inertias = []
kmax = 10

for k in range(2, kmax+1):
    kmeans = KMeans(n_clusters = k).fit(tfIdf)
    labels = kmeans.labels_
    sil.append(silhouette_score(tfIdf, labels, metric = 'euclidean'))
    inertias.append(kmeans.inertia_)

plt.plot(range(2, kmax+1), inertias)
plt.title('Elbow method')
plt.xlabel('No of clusters')
plt.ylabel('Error')
plt.show()


plt.plot(range(2, kmax+1), sil)
plt.xlabel('No of clusters')
plt.ylabel('Silhouette score')
plt.show()


km = KMeans(n_clusters=6, random_state=666)
km.fit(tfIdf)
df['Cluster'] = km.labels_.tolist()
df.head()


ax = sns.countplot(x="Cluster", data=df)


date_post_count = pd.DataFrame({'Count' : df.groupby( [ "Cluster", "Created"] ).size()}).reset_index()
date_post_count['Created'] = pd.to_datetime(date_post_count['Created'])
date_post_count = date_post_count.sort_values(by=['Created'])
date_post_count.head()


class DateDashboard(param.Parameterized):
    
    Cluster = param.ObjectSelector(default=0, objects=sorted(list(df.Cluster.unique())))
    plt.figure(figsize=(12,8), dpi= 100)
    
    def line_plot(self):
        plt.clf()  
        plt.plot('Created', 'Count', data=date_post_count[date_post_count['Cluster']==self.Cluster], linestyle = 'solid')
        return plt.gcf()

    def append_end(self):
        plt.clf()
        return ""

dd = DateDashboard(name='')

dashboard = pn.Column('Date Dashboard',
                      dd.param,
                      dd.line_plot,
                      dd.append_end
                     )

dashboard.embed()

<Figure size 1728x1152 with 0 Axes>


dashboard_cluster = 1

class RatingsDashboard(param.Parameterized):
    plt.figure(figsize=(12,8), dpi= 100)
    Cluster = param.ObjectSelector(default=dashboard_cluster, objects=sorted(list(df.Cluster.unique())))

    def get_data(self):
        class_df = df[(df.Cluster==self.Cluster)].copy()
        return class_df
    
    def get_tfidf(self):
        df_temp = self.get_data()
        tfIdf_temp = tfIdfVectorizer.fit_transform(df_temp['Text'])
        tfIdf_df_temp = pd.DataFrame(tfIdf_temp[0].T.todense(), index=tfIdfVectorizer.get_feature_names(), columns=["TF-IDF"])
        tfIdf_df_temp = tfIdf_df_temp.sort_values('TF-IDF', ascending=False)
        return(tfIdf_df_temp)

    def word_cloud(self):
        wordcloud = WordCloud(background_color="white", max_words=50).generate_from_frequencies(self.get_tfidf().sum(axis=1))
        plt.clf()
        plt.axis("off")
        im = plt.imshow(wordcloud, interpolation='bilinear')
        return plt.gcf()
    
    def table_view(self, x=10):
        return self.get_tfidf().head(x)
    
    def count_plot(self):
        plt.clf()
        ax = sns.countplot(y="Page Name", data=self.get_data(), order=pd.value_counts(self.get_data()['Page Name']).iloc[:10].index)
        return plt.gcf()
    
    def status_table(self):
        return self.get_data().mean()
        
    def corr_plot(self):
        plt.clf()
        color = plt.get_cmap('RdYlGn').copy()
        color.set_bad('lightblue')
        corrMatrix = self.get_data()[status_metrics].corr()
        sns.heatmap(corrMatrix, annot=True, cmap=color)
        return plt.gcf()
    
    def most_commented_posts(self):
        return self.get_data().sort_values(by=['Likes', 'Comments'], ascending=False).head(20)['URL']

    def append_end(self):
        plt.clf()
        return ""


rd = RatingsDashboard(name='')

dashboard = pn.Column('Cluster Dashboard',
                      rd.param,
                      pn.Tabs(
                          ('Most Common Words', rd.table_view),
                          ('Word Cloud', rd.word_cloud),
                          ('Most Common Page Names', rd.count_plot),
                          ('Average Status Metrics', rd.status_table),
                          ('Status Correlations', rd.corr_plot),
                          ('Target Posts', rd.most_commented_posts),
                          ('',rd.append_end))
                     )

dashboard.embed()

<Figure size 1728x1152 with 0 Axes>

	Page Name	User Name	Page Id	Page Likes at Posting	Created	Type	Likes	Comments	Shares	Love	...	URL	Message	Link	Final Link	Link Text	Description	Sponsor Id	Sponsor Name	Score	Yes
0	Planet Paws	PlanetPaws.ca	1.124379e+14	1578676	2017-02-26 09:50:24 EST	Native Video	75461	42879	894482	2406	...	https://www.facebook.com/PlanetPaws.ca/posts/1...	Over-vaccinating and the overdosing of pet vac...	https://www.facebook.com/PlanetPaws.ca/videos/...	NaN	The Dangers of Vaccine Overdosing	NaN	NaN	NaN	1091367.0	1
1	Natalie Bomke Fox 32 Chicago	NatalieBomkeFox32Chicago	1.668040e+14	112316	2018-02-02 09:30:00 EST	Native Video	135485	41642	627972	27306	...	https://www.facebook.com/NatalieBomkeFox32Chic...	CANCER VACCINE SUCCESSFUL IN TESTING... Stanfo...	https://www.facebook.com/NatalieBomkeFox32Chic...	NaN	CANCER VACCINE SUCCESSFUL IN TESTING...	NaN	NaN	NaN	861565.0	1
2	FOX 2 Detroit	WJBKFox2Detroit	3.636580e+11	819062	2018-02-05 14:45:02 EST	Native Video	84593	15447	580983	14496	...	https://www.facebook.com/WJBKFox2Detroit/posts...	CANCER VACCINE SUCCESSFUL IN TESTING: The vacc...	https://www.facebook.com/WJBKFox2Detroit/video...	NaN	CANCER VACCINE SUCCESSFUL	NaN	NaN	NaN	712774.0	1
3	Gizmodo	gizmodo	5.718759e+09	1555197	2017-11-20 17:40:46 EST	Native Video	122206	20150	433945	11158	...	https://www.facebook.com/gizmodo/posts/1015594...	Paul Alexander spends nearly every hour, of ev...	https://www.facebook.com/gizmodo/videos/101559...	NaN	The Last of the Iron Lungs	NaN	NaN	NaN	674298.0	1
4	Hashem Al-Ghaili	ScienceNaturePage	6.935046e+14	13426659	2018-06-22 11:08:29 EDT	Native Video	40464	3112	322544	3762	...	https://www.facebook.com/ScienceNaturePage/pos...	Cancer Vaccine Has Been Approved For Human Tri...	https://www.facebook.com/ScienceNaturePage/vid...	NaN	Cancer Vaccine Has Been Approved For Human Trials	NaN	NaN	NaN	373126.0	1

	Page Name	Page Likes at Posting	Created	Type	Likes	Comments	Shares	Love	Wow	Haha	...	Angry	Thankful	Video Share Status	Total Views for all Crossposts	URL	Message	Link Text	Description	Sponsor Name	Score
0	Planet Paws	1578676	2017-02-26 09:50:24 EST	Native Video	75461	42879	894482	2406	7310	106	...	36332	5	crosspost	38821399	https://www.facebook.com/PlanetPaws.ca/posts/1...	Over-vaccinating and the overdosing of pet vac...	The Dangers of Vaccine Overdosing	NaN	NaN	1091367.0
1	Natalie Bomke Fox 32 Chicago	112316	2018-02-02 09:30:00 EST	Native Video	135485	41642	627972	27306	28412	446	...	141	0	original	0	https://www.facebook.com/NatalieBomkeFox32Chic...	CANCER VACCINE SUCCESSFUL IN TESTING... Stanfo...	CANCER VACCINE SUCCESSFUL IN TESTING...	NaN	NaN	861565.0
2	FOX 2 Detroit	819062	2018-02-05 14:45:02 EST	Native Video	84593	15447	580983	14496	16707	319	...	139	0	crosspost	27529776	https://www.facebook.com/WJBKFox2Detroit/posts...	CANCER VACCINE SUCCESSFUL IN TESTING: The vacc...	CANCER VACCINE SUCCESSFUL	NaN	NaN	712774.0
3	Gizmodo	1555197	2017-11-20 17:40:46 EST	Native Video	122206	20150	433945	11158	34444	218	...	163	0	crosspost	38621026	https://www.facebook.com/gizmodo/posts/1015594...	Paul Alexander spends nearly every hour, of ev...	The Last of the Iron Lungs	NaN	NaN	674298.0
4	Hashem Al-Ghaili	13426659	2018-06-22 11:08:29 EDT	Native Video	40464	3112	322544	3762	2923	128	...	99	0	original	0	https://www.facebook.com/ScienceNaturePage/pos...	Cancer Vaccine Has Been Approved For Human Tri...	Cancer Vaccine Has Been Approved For Human Trials	NaN	NaN	373126.0

	Page Likes at Posting	Likes	Comments	Shares	Love	Wow	Haha	Sad	Angry	Thankful	Total Views for all Crossposts
0	1578676.0	75461	42879	894482	2406	7310	106	32391	36332	5.0	38821399
1	112316.0	135485	41642	627972	27306	28412	446	161	141	0.0	0
2	819062.0	84593	15447	580983	14496	16707	319	90	139	0.0	27529776
3	1555197.0	122206	20150	433945	11158	34444	218	52014	163	0.0	38621026
4	13426659.0	40464	3112	322544	3762	2923	128	94	99	0.0	0

	Page Likes at Posting	Likes	Comments	Shares	Love	Wow	Haha	Sad	Angry	Total Views for all Crossposts
0	1578676.0	75461.0	42879.0	894482.0	2406.0	7310.0	106.0	32391.0	2224.0	38821399.0
1	112316.0	135485.0	41642.0	627972.0	6685.0	28412.0	446.0	161.0	141.0	0.0
2	819062.0	84593.0	15447.0	580983.0	14496.0	12640.0	2973.0	1772.0	139.0	9634769.0
3	5631828.0	74635.0	20150.0	433945.0	11158.0	34444.0	218.0	52014.0	3388.0	38621026.0
4	13426659.0	40464.0	5602.0	24774.0	3762.0	2923.0	128.0	146.0	99.0	0.0

	Page Name	Created	Type	Video Share Status	URL	Message	Link Text	Description	Sponsor Name	Score	...	Likes	Comments	Shares	Love	Wow	Haha	Sad	Angry	Thankful	Total Views for all Crossposts
0	Planet Paws	26-02-2017	Native Video	crosspost	https://www.facebook.com/PlanetPaws.ca/posts/1...	Over-vaccinating and the overdosing of pet vac...	The Dangers of Vaccine Overdosing	NaN	NaN	1091367.0	...	75461.0	42879.0	894482.0	2406.0	7310.0	106.0	32391.0	2224.0	0.0	38821399.0
2	FOX 2 Detroit	05-02-2018	Native Video	crosspost	https://www.facebook.com/WJBKFox2Detroit/posts...	CANCER VACCINE SUCCESSFUL IN TESTING: The vacc...	CANCER VACCINE SUCCESSFUL	NaN	NaN	712774.0	...	84593.0	15447.0	580983.0	14496.0	12640.0	2973.0	1772.0	139.0	0.0	9634769.0
3	Gizmodo	20-11-2017	Native Video	crosspost	https://www.facebook.com/gizmodo/posts/1015594...	Paul Alexander spends nearly every hour, of ev...	The Last of the Iron Lungs	NaN	NaN	674298.0	...	74635.0	20150.0	433945.0	11158.0	34444.0	218.0	52014.0	3388.0	0.0	38621026.0
16	Fusion	20-11-2017	Native Video	crosspost	https://www.facebook.com/fusiontv/posts/101559...	Paul Alexander spends nearly every hour of eve...	The Last of the Iron Lungs	NaN	NaN	99116.0	...	54270.0	8285.0	68735.0	2630.0	4600.0	34.0	6463.0	24.0	0.0	38539076.0
27	NowThis	01-03-2017	Native Video	crosspost	https://www.facebook.com/NowThisNews/posts/146...	An experimental vaccine therapy left HIV patie...	A Vaccine Therapy Left HIV Patients Virus-Free	NaN	NaN	67578.0	...	34488.0	746.0	23102.0	1024.0	1103.0	74.0	7.0	2.0	0.0	3938467.0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
9181	J.B. Handley	19-11-2017	Native Video	crosspost	https://www.facebook.com/jbhandleyjr/posts/190...	#11: All-time most viewed videos. I started th...	Vaccine adjuvants cause autoimmunity	NaN	NaN	1729.0	...	737.0	112.0	784.0	78.0	53.0	1.0	7.0	7.0	0.0	224430.0
9235	Rockets Are Cool	10-07-2017	Native Video	crosspost	https://www.facebook.com/RocketsAreCool/posts/...	A tiny patch could replace painful needles.	Cutting-Edge Vaccine Patch	NaN	NaN	1720.0	...	658.0	93.0	828.0	40.0	72.0	3.0	0.0	3.0	0.0	0.0
9426	Gates Foundation	20-07-2017	Native Video	crosspost	https://www.facebook.com/gatesfoundation/posts...	How do you look for weak spots on viruses to t...	How 3D Can Build Better Vaccines	NaN	NaN	1691.0	...	1080.0	176.0	311.0	102.0	20.0	2.0	0.0	2.0	0.0	0.0
9474	Stop Mandatory Vaccination	12-11-2017	Native Video	crosspost	https://www.facebook.com/StopMandatoryVaccinat...	Tasha Dāvid has 6 vaccine injured kids and 2 h...	“My Family Was Devastated By Vaccines” by Tash...	NaN	NaN	1683.0	...	416.0	53.0	968.0	19.0	21.0	0.0	159.0	47.0	0.0	81718.0
9475	J.B. Handley	16-11-2017	Native Video	crosspost	https://www.facebook.com/jbhandleyjr/posts/190...	#13: All-time most viewed videos. I started th...	Four nurses tell the truth about vaccines	NaN	NaN	1683.0	...	730.0	55.0	743.0	103.0	33.0	1.0	2.0	3.0	2.0	164748.0

Import dataset & libraries¶

Drop columns¶

Update data types¶

Imputation¶

Natural Language Processing¶

EDA¶

Clustering TF-IDF¶

	Page Name	Page Likes at Posting	Created	Type	Likes	Comments	Shares	Love	Wow	Haha	...	Thankful	Total Views for all Crossposts	URL	Message	Link Text	Description	Sponsor Name	Score
1154	Rodney Habib	2018-06-17 19:22:32 EDT	Native Video	2096	312	3704	75	110	0	725	...	share	NaN	Over-vaccinating and the overdosing of pet vac...	https://www.facebook.com/PlanetPaws.ca/videos/...	Over-vaccinating and the overdosing of pet vac...	NaN	7878	NaN
1928	Rodney Habib	2018-07-15 09:43:36 EDT	Native Video	1827	250	2668	81	79	0	293	...	share	NaN	The day John Robb shook up the American Govern...	https://www.facebook.com/PlanetPaws.ca/videos/...	Over-vaccinating and the overdosing of pet vac...	NaN	5614	NaN
3278	Truong Huy San	2018-12-07 04:00:39 EST	Status	3248	0	269	106	27	196	22	...	NaN	https://www.facebook.com/Osinhuyduc/posts/1886...	GIÁO DỤC VỊ SINH VIÊN HAY VỊ NGƯỜI NGỒI TRÊN B...	NaN	NaN	NaN	3887	NaN
3439	Rodney Habib	2018-08-14 10:47:19 EDT	Photo	1146	203	2191	56	94	0	48	...	NaN	NaN	Do you know? By Rodney Habib	https://www.facebook.com/PlanetPaws.ca/photos/...	SO JUST HOW LONG DO YOUR PET’S SHOTS LAST?\n\n...	NaN	3762	NaN
6695	Mike Acebedo Lopez	2017-12-01 13:21:30 EST	Status	1268	0	357	8	11	5	222	...	NaN	https://www.facebook.com/MikeAcebedoLopez/post...	Noynoy Aquino's government potentially sentenc...	NaN	NaN	NaN	2264	NaN
7208	Trang Minh Nguyen	2018-10-04 08:29:38 EDT	Photo	654	223	1203	45	0	0	0	...	NaN	https://www.facebook.com/trangmnguyen132/posts...	Chia sẻ với các bố mẹ lịch tiêm chủng mà Mầm N...	https://www.facebook.com/photo.php?fbid=204533...	NaN	NaN	2125	NaN
9379	Yaya Trương Nhi	2018-08-21 23:59:26 EDT	Link	1483	12	119	78	3	1	0	...	NaN	https://www.facebook.com/yayatruongnhi/posts/1...	❤️❤️❤️ Cảm ơn những Tiến sỹ đã cứu cả thế giới...	http://vietbao.info/suc-khoe/tin-vui-my-da-ngh...	Giáo sư Ronald Levy – cha đẻ của vaccine chống...	NaN	1696	NaN

	Page Name	Created	Type	URL	Message	Link Text	Description	Score	Page Likes at Posting	Likes	Comments	Shares	Love	Wow	Haha	Sad	Angry	Total Views for all Crossposts
889	Good Health Journal	19-01-2016	Photo	https://www.facebook.com/390774211023677/posts...	ျဖည့္ထုိးႏုိင္သည့္ကာကြယ္ေဆးမ်ား --------------...	Timeline Photos		9119.0	3776566.0	3448.0	70.0	358.0	2.0	0.0	0.0	0.0	0.0	0.0
896	Yoyarlay	23-01-2016	Photo	https://www.facebook.com/YoYarLay/posts/101590...	ျဖည့္ထုိးႏုိင္သည့္ကာကြယ္ေဆးမ်ား --------------...	Timeline Photos		9081.0	3187808.0	5377.0	421.0	3622.0	0.0	0.0	0.0	0.0	2.0	0.0
3278	Truong Huy San	07-12-2018	Status	https://www.facebook.com/Osinhuyduc/posts/1886...				3887.0	3248.0	0.0	269.0	106.0	10.0	196.0	22.0	70.0	460.0	0.0
4293	China Xinhua News	01-07-2016	Link	https://www.facebook.com/XinhuaMyanmar/posts/6...	အဆုတ္ေရာင္ကာကြယ္ေဆးသစ္ ကေလးငယ္တစ္ဦး မက်န္ ထို...	အဆုတ္ေရာင္ကာကြယ္ေဆးသစ္ ကေလးငယ္တစ္ဦး မက်န္ ထို...	ေနျပည္ေတာ္၊ ဇူလိုင္ ၁ ရက္ (ဆင္ဟြာ) ကမၻာ့က်န္းမ...	3187.0	7439522.0	2914.0	21.0	252.0	0.0	2.0	0.0	0.0	0.0	0.0
6109	UNICEF Bangladesh	10-07-2017	Photo	https://www.facebook.com/unicef.bd/posts/16076...	ধন্যবাদ সকল মা-বাবাকে যারা সময়মতো টিকা দিয়ে বা...	Timeline Photos		2449.0	532837.0	4177.0	27.0	150.0	8.0	2.0	0.0	0.0	1.0	0.0
6695	Mike Acebedo Lopez	01-12-2017	Status	https://www.facebook.com/MikeAcebedoLopez/post...				2264.0	1268.0	0.0	357.0	8.0	4.0	5.0	222.0	393.0	1072.0	0.0
7208	Trang Minh Nguyen	04-10-2018	Photo	https://www.facebook.com/trangmnguyen132/posts...				2125.0	346663.0	223.0	1203.0	45.0	0.0	47.0	0.0	0.0	267.0	232550.0

	TF-IDF
overdose	0.598978
witness	0.270074
terrible	0.270074
vet	0.239843
global	0.216992

	Score	Page Likes at Posting	Likes	Comments	Shares	Love	Wow	Haha	Sad	Angry	Thankful	Total Views for all Crossposts	Polarity	Subjectivity
count	8819.000000	8.819000e+03	8819.000000	8819.000000	8819.000000	8819.000000	8819.000000	8819.000000	8819.000000	8819.000000	8819.000000	8.819000e+03	8819.000000	8819.000000
mean	3997.134029	1.673022e+06	1841.368976	190.766413	1452.849756	105.459009	73.275088	51.584647	103.016442	131.138791	0.026874	1.119640e+04	0.075021	0.379761
std	3495.881871	2.575138e+06	1968.704536	283.029199	1895.926428	169.605461	123.171708	146.648540	223.887119	238.699886	0.471293	1.034664e+05	0.176332	0.242108
min	1674.000000	3.950000e+02	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000e+00	-0.525000	0.000000
25%	2107.000000	2.213700e+05	700.000000	54.000000	448.000000	6.000000	6.000000	1.000000	2.000000	1.000000	0.000000	0.000000e+00	0.000000	0.214286
50%	2852.000000	7.362250e+05	1321.000000	101.000000	945.000000	40.000000	29.000000	4.000000	15.000000	11.000000	0.000000	0.000000e+00	0.052941	0.411111
75%	4467.000000	1.795225e+06	2209.000000	209.000000	1691.000000	132.000000	90.000000	19.000000	99.500000	165.000000	0.000000	0.000000e+00	0.166667	0.526523
max	60774.000000	1.668844e+07	19436.000000	3153.000000	39618.000000	1534.000000	1680.000000	1270.000000	2235.000000	1571.000000	22.000000	2.973804e+06	0.675000	1.000000

	Cluster	Created	Count
479	2	2016-01-01	1
986	4	2016-01-01	7
481	2	2016-01-02	1
990	4	2016-01-02	8
2134	5	2016-01-03	1