Types de données et fusion des données

Analyser des campagnes marketing avec pandas

Jill Rosok

Data Scientist

Types de données courants

Chaînes (objets)
Nombres (floats, entiers)
Booléens (True, False)
Dates

Type de données d'une colonne

# Print a data type of a single column
print(marketing['converted'].dtype)

dtype('object')

Modifier le type de données d'une colonne

# Change the data type of a column
marketing['converted'] = marketing['converted']\
                          .astype('bool')

print(marketing['converted'].dtype)

dtype('bool')

Créer de nouvelles colonnes booléennes

marketing['is_house_ads'] = np.where(
    marketing['marketing_channel'] == 'House Ads', 
    True, False
)

print(marketing.is_house_ads.head(3))

0    True
1    False
2    True
Name: is_house_ads, dtype: bool

Mapper des valeurs à des colonnes existantes

channel_dict = {"House Ads": 1, "Instagram": 2, 
                "Facebook": 3, "Email": 4, "Push": 5}

marketing['channel_code'] = marketing['marketing_channel']\
                              .map(channel_dict)

print(marketing['channel_code'].head(3))

0    1
1    1
2    1
Name: channel_code, dtype: int64

Colonnes de dates

# Read date columns using parse_dates
marketing = pd.read_csv('marketing.csv', 
                        parse_dates=['date_served', 
                                     'date_subscribed', 
                                     'date_canceled'])

# Or
# Convert already existing column to datetime column
marketing['date_served'] = pd.to_datetime(
    marketing['date_served']
)

Colonnes de dates

# Or convert each column individually
# Convert already existing column to datetime column
marketing['date_served'] = pd.to_datetime(
    marketing['date_served']
)

Colonnes de dates

marketing['day_served'] = marketing['date_served']\
                       .dt.dayofweek

Passons à la pratique !

Analyser des campagnes marketing avec pandas