Extract and transform e-mail domain.

emailDomain.py

# Import libraries
import pandas as pd
import numpy as np
import seaborn as sns
from functions import aux_functions

# Extract e-mail domain
hardbounce_2['emailDomain'] = hardbounce_2['email'].apply(aux_functions.getEmailDomain)

# count by domain
group_df = pd.DataFrame(hardbounce_2.groupby('emailDomain').count()['email'].sort_values(ascending=False)).reset_index()

# plot top 20 domains
sns.set(style="whitegrid")
ax = sns.barplot(x="email", y="emailDomain", data=group_df.head(20))

# Classify emails different from gmail and hotmail in others
hardbounce_2["emailDomain_cat"] =  hardbounce_2["emailDomain"].apply(lambda x: 'others' if (x != 'gmail.com' and x != 'hotmail.com') else x)

# count by domain categorized
group_df = pd.DataFrame(hardbounce_2.groupby('emailDomain_cat').count()['email'].sort_values(ascending=False)).reset_index()

# plot domains
sns.set(style="whitegrid")
ax = sns.barplot(x="email", y="emailDomain_cat", data=group_df)

# Variable Y distribution by domain category
aux_functions.freqTable(
    [hardbounce_2["emailDomain_cat"]]
    ,[hardbounce_2["flgHardBounce_n"]]
    ,True
    ,"index"
)

emailDomainPiece1.py

# Import libraries
import pandas as pd
import numpy as np
import seaborn as sns
from functions import aux_functions

# Split first domain piece
hardbounce_2['emailDomainPiece1'] = hardbounce_2['emailDomain'].apply(aux_functions.getPiece1EmailDomain)

# count by domain piece
group_df = pd.DataFrame(hardbounce_2.groupby('emailDomainPiece1').count()['email'].sort_values(ascending=False)).reset_index()

# plot top 20 domains
sns.set(style="whitegrid")
ax = sns.barplot(x="email", y="emailDomainPiece1", data=group_df.head(20))

# Create category with domains different from .com because this represents the majority of the dataset rows
hardbounce_2["emailDomainPiece1"] =  hardbounce_2["emailDomainPiece1"].apply(lambda x: 'others' if x != 'com' else x)

# Variable Y distribution by domain piece
aux_functions.freqTable(
    [hardbounce_2["emailDomainPiece1"]]
    ,[hardbounce_2["flgHardBounce_n"]]
    ,True
    ,"index"
)

emailDomainPiece2.py

# Import libraries
import pandas as pd
import numpy as np
import seaborn as sns
from functions import aux_functions

# Split second domain piece
hardbounce_2['emailDomainPiece2'] = hardbounce_2['emailDomain'].apply(aux_functions.getPiece2EmailDomain)

# count by domain piece
group_df = pd.DataFrame(hardbounce_2.groupby('emailDomainPiece2').count()['email'].sort_values(ascending=False)).reset_index()

# plot top 20 domains
sns.set(style="whitegrid")
ax = sns.barplot(x="email", y="emailDomainPiece2", data=group_df.head(20))

# Create category with domains different from .br and missing because they represents the majority of the dataset rows
hardbounce_2["emailDomainPiece2"] =  hardbounce_2["emailDomainPiece2"].apply(lambda x: 'others' if (x != 'missing' and x != 'br') else x)

# Variable Y distribution by domain piece
aux_functions.freqTable(
    [hardbounce_2["emailDomainPiece2"]]
    ,[hardbounce_2["flgHardBounce_n"]]
    ,True
    ,"index"
)