Convert data frame for profit curve

user_dataset_profit_curve.py

def transform_column(df,column_name,fn):
    """
    Transforms a column with the given function
    """
    df[column_name] = df[column_name].apply(fn)
 


def vectorize_feature_index(df,label_column):
    feature_names = []
    global X
    for feature_name in df.columns.values:
        print feature_name
        if feature_name != label_column:
            if feature_name not in feature_names:
                feature_names.append(feature_name)
    
    X = df[feature_names].index
    scaler = StandardScaler()
    X = scaler.fit_transform(X)
    train_index,test_index = train_test_split(df.index)
    X = df[feature_names].as_matrix().astype(np.float)
    y = df[label_column].index
    y_test = y[test_index].astype(float)

q = session.query(Users).join(Event).add_entity(Event)
df= query_to_df(session,q)
df = df.drop(['Users_id','Event_id','Event_User_Id','Event_Meal_Id','Users_Created_Date'],1)
print df.columns
print df.Users_date.dtype
def to_epoch(time_input):
    return (time_input - datetime(1970,1,1)).total_seconds()

transform_column(df,'Event_Type',event_to_num.get)
transform_column(df,'Users_Campaign_ID',campaign_to_num.get)
transform_column(df,'Users_date',to_epoch)
transform_column(df,'Event_date',to_epoch)
print df
vectorize_feature_index(df,'Event_Type')

print X