import pandas as pd
import numpy as np
import sklearn
from sklearn.model_selection import train_test_split
from tpot import TPOTRegressor
#import data
from olist.order import Order
from olist.data import Olist
data = Olist().get_data()
training_orders = Order().get_training_data()
orders = data['olist_orders_dataset']
orders['estimate_wait_time'] = (pd.to_datetime(orders['order_estimated_delivery_date'])\
- pd.to_datetime(orders['order_purchase_timestamp'])) / np.timedelta64(24, 'h')
training_orders =\
training_orders.merge(orders[['estimate_wait_time', 'order_id']], on='order_id')
X = training_orders.drop(['order_id', 'wait_time', 'delay_vs_expected'],
axis=1)
y = training_orders['wait_time']
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.33,
random_state=42)
tpot = TPOTRegressor(generations=5, population_size=50, verbosity=2)
def __init__(self):
# Import only data once
olist = Olist()
self.data = olist.get_data()
self.matching_table = olist.get_matching_table()
self.order = Order()
def __init__(self):
self.data = Olist().get_data()
def get_distance_seller_customer(self):
"""
02-01 > Returns a DataFrame with order_id
and distance between seller and customer
"""
# Optional
# Hint: you can use the haversine_distance logic coded in olist/utils.py
data = self.data
matching_table = Olist().get_matching_table()
# Since one zipcode can map to multiple (lat, lng), take first one
geo = data['geolocation']
geo = geo.groupby('geolocation_zip_code_prefix',
as_index=False).first()
# Select sellers and customers
sellers = data['sellers']
customers = data['customers']
# Merge geo_location for sellers
sellers_mask_columns = ['seller_id', 'seller_zip_code_prefix',
'seller_city', 'seller_state',
'geolocation_lat', 'geolocation_lng']
sellers_geo = sellers.merge(geo,
how='left',
left_on='seller_zip_code_prefix',
right_on='geolocation_zip_code_prefix')[sellers_mask_columns]
# Merge geo_location for customers
customers_mask_columns = ['customer_id', 'customer_zip_code_prefix',
'customer_city', 'customer_state',
'geolocation_lat', 'geolocation_lng']
customers_geo = customers.merge(geo,
how='left',
left_on='customer_zip_code_prefix',
right_on='geolocation_zip_code_prefix')[customers_mask_columns]
# Use the matching table and merge customers and sellers
matching_geo = matching_table.merge(sellers_geo,
on='seller_id')\
.merge(customers_geo,
on='customer_id',
suffixes=('_seller',
'_customer'))
# Remove na()
matching_geo = matching_geo.dropna()
matching_geo.loc[:, 'distance_seller_customer'] =\
matching_geo.apply(lambda row:
haversine_distance(row['geolocation_lng_seller'],
row['geolocation_lat_seller'],
row['geolocation_lng_customer'],
row['geolocation_lat_customer']),
axis=1)
# Since an order can have multiple sellers,
# return the average of the distance per order
order_distance =\
matching_geo.groupby('order_id',
as_index=False).agg({'distance_seller_customer':
'mean'})
return order_distance
def __init__(self):
# Assign an attribute ".data" to all new instances of Order
self.data = Olist().get_data()