In [158]:

    

import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt

%matplotlib inline


    

In [159]:

    

aisles = pd.read_csv('data/aisles.csv', index_col='aisle_id').to_dict()['aisle']
depart = pd.read_csv('data/departments.csv', index_col='department_id').to_dict()['department']
products_df = pd.read_csv('data/products.csv', index_col='product_id')
products_df['aisle'] = products_df.aisle_id.map(lambda x: aisles[x])
products_df['department'] = products_df.department_id.map(lambda x: depart[x])
products_df.drop(['aisle_id', 'department_id'], axis=1, inplace=True)


    

In [160]:

    

prior_df = pd.read_csv('data/order_products__prior.csv', index_col='order_id')
train_df = pd.read_csv('data/order_products__train.csv', index_col='order_id')
orders_df = pd.read_csv('data/orders.csv', index_col='order_id')


    

In [161]:

    

products_df.head()


    

    
Out[161]:






  

    

      

      
product_name
      
aisle
      
department
    
    

      
product_id
      

      

      

    
  
  

    

      
1
      
Chocolate Sandwich Cookies
      
cookies cakes
      
snacks
    
    

      
2
      
All-Seasons Salt
      
spices seasonings
      
pantry
    
    

      
3
      
Robust Golden Unsweetened Oolong Tea
      
tea
      
beverages
    
    

      
4
      
Smart Ones Classic Favorites Mini Rigatoni Wit...
      
frozen meals
      
frozen
    
    

      
5
      
Green Chile Anytime Sauce
      
marinades meat preparation
      
pantry
    
  

In [162]:

    

prior_df.head()


    

    
Out[162]:






  

    

      

      
product_id
      
add_to_cart_order
      
reordered
    
    

      
order_id
      

      

      

    
  
  

    

      
2
      
33120
      
1
      
1
    
    

      
2
      
28985
      
2
      
1
    
    

      
2
      
9327
      
3
      
0
    
    

      
2
      
45918
      
4
      
1
    
    

      
2
      
30035
      
5
      
0
    
  

In [163]:

    

orders_df.head()


    

    
Out[163]:






  

    

      

      
user_id
      
eval_set
      
order_number
      
order_dow
      
order_hour_of_day
      
days_since_prior_order
    
    

      
order_id
      

      

      

      

      

      

    
  
  

    

      
2539329
      
1
      
prior
      
1
      
2
      
8
      
NaN
    
    

      
2398795
      
1
      
prior
      
2
      
3
      
7
      
15.0
    
    

      
473747
      
1
      
prior
      
3
      
3
      
12
      
21.0
    
    

      
2254736
      
1
      
prior
      
4
      
4
      
7
      
29.0
    
    

      
431534
      
1
      
prior
      
5
      
4
      
15
      
28.0
    
  

In [235]:

    

def get_order_to_products(df):
    tmp = df.product_id * df.reordered
    return tmp.groupby(tmp.index).aggregate({'products': lambda x: tuple(e for e in x if e > 0)})

def get_user_to_last_order(orders_df):
    return orders_df[orders_df.eval_set=='prior'].reset_index().groupby('user_id').last()['order_id'].to_frame()

def calc_f1_score(exp, act):
    exp = set(exp)
    act = set(act)
    com = exp & act
    p = 1.0 * len(com) / len(act) if act else 1.0
    r = 1.0 * len(com) / len(exp) if exp else 1.0
    return 2.0 * p * r / (p + r) if (p + r) > 0 else 0.0

def get_score(expected, actual):
    score_df = pd.merge(
        expected, actual,
        left_index=True, right_index=True,
        suffixes=['_exp', '_act']
    )
    scores = score_df.apply(lambda x: calc_f1_score(x.get('products_exp'), x.get('products_act')), axis=1)   
    return scores.mean()


    

In [226]:

    

# Expensive computations.
order_to_products = get_order_to_products(prior_df)


    

In [236]:

    

def solution_1(order_to_products, prior_df, orders_df, label):
    """Make the same basket prediction as the last purchase of a user."""
    user_to_last_order = get_user_to_last_order(orders_df)
    user_to_last_products = pd.merge(
        user_to_last_order, order_to_products,
        left_on='order_id', right_index=True, how='inner'
    )
    
    predict_df = orders_df[orders_df.eval_set==label]['user_id'].to_frame()
    predictions = pd.merge(
        predict_df, user_to_last_products,
        left_on='user_id', right_index=True, how='inner'
    )
    return predictions['products'].to_frame()

train_actual = solution_1(order_to_products, prior_df, orders_df, 'train')
train_expected = get_order_to_products(train_df)
print('Solution 1 train validation score: {}'.format(get_score(train_expected, train_actual)))

test_actual = solution_1(order_to_products, prior_df, orders_df, 'test')
test_actual = test_actual.products.apply(lambda t: ' '.join(str(e) for e in t) if t else 'None').to_frame()
test_actual.to_csv("test_output.csv")


    

    




Solution 1 train validation score: 0.32558651711

In [ ]: