shawlu95

def load_dataset(split="trn_set", limit=None, ignore_categorical=False):
    sql = """
    SELECT o.*, f1.*, f2.*, f3.*, f4.*,
    EXTRACT(MONTH FROM o.dt) AS month
    FROM %s AS t 
    JOIN Online AS o 
        ON t.index = o.index 
    JOIN features_group_1 AS f1
        ON t.index = f1.index
    JOIN features_group_2 AS f2

shawlu95

mysql> show tables;
+----------------------+
| Tables_in_shutterfly |
+----------------------+
| Online               |
| Purchase             |
| features_group_1     |
| features_group_2     |
| features_group_3     |
| features_group_4     |

shawlu95

USE Shutterfly;

DROP TABLE IF EXISTS features_group_1;

CREATE TABLE IF NOT EXISTS features_group_1
SELECT o.index
  ,LEFT(o.dt, 10) AS day
  ,COUNT(*) AS order_count
  ,SUM(p.revenue) AS revenue_sum
  ,MAX(p.revenue) AS revenue_max

shawlu95

mysql> select count(*) from trn_set;
+----------+
| count(*) |
+----------+
|   859296 |
+----------+
1 row in set (0.61 sec)

mysql> select count(*) from tst_set;
+----------+

shawlu95

sql = "SELECT `index`, event2 FROM Online;"
df = pd.read_sql_query(sql, engine).set_index('index')

# shuffle dataset, preserving index
df = df.sample(frac=1)

train_frac = 0.9
test_frac = 1 - train_frac

trn_cutoff = int(len(df) * train_frac)

shawlu95

mysql> use shutterfly;
Reading table information for completion of table and column names
You can turn off this feature to get a quicker startup with -A

Database changed

mysql> show tables;
+----------------------+
| Tables_in_shutterfly |
+----------------------+

shawlu95

from sqlalchemy import create_engine
import pandas as pd

username = "root"
password = "1234567"
port = 3306
database = "Shutterfly"

engine = create_engine('mysql+mysqldb://%s:%s@localhost:%i/%s'
                       %(username, password, port, database))

shawlu95

# two-tailed t-test
h_0 = 300
h_1 = 290

n = 50
se = 20 / np.sqrt(n)

power = compute_power(h_0, h_1, se, tail="two")
print("power: %.3f, beta: %.3f"%(power, 1 - power))

shawlu95

# one-tailed z-test
h_0 = 0.8
h_1 = 0.75

n = 100
se = np.sqrt(h_0 * (1 - h_0) / n)

power = compute_power(h_0, h_1, se, tail="left")
print("power: %.3f, beta: %.3f"%(power, 1 - power))

shawlu95

def critical_z(alpha=0.05, tail="two"):
    """
    Given significance level, compute critical value.
    """
    if tail == "two":
        p = 1 - alpha / 2
    else:
        p = 1 - alpha
        
    return norm.ppf(p)