Apriori Algorithm

Association Rules

S.Reza.Eftekhary 1400/11/12

1 - Importing the required libraries

import numpy as np import pandas as pd from mlxtend.frequent_patterns import apriori, association_rules, fpgrowth

2 - Loading and exploring the data

Import Data UCI

http://archive.ics.uci.edu/ml/datasets/Online+Retail

data = pd.read_csv('OnlineRetail.csv') data.head()

InvoiceNoobject

StockCodeobject

536365

85123A

536365

71053

536365

84406B

536365

84029G

536365

84029E

3 - Exploring the columns of the data

data.columns print (data.columns)

Index(['InvoiceNo', 'StockCode', 'Description', 'Quantity', 'InvoiceDate',
       'UnitPrice', 'CustomerID', 'Country'],
      dtype='object')

4 - Exploring the different regions of transactions

data.Country.unique()

5 - Cleaning the Data

data['Description'] = data['Description'].str.strip() # Dropping the rows without any invoice number data.dropna(axis = 0, subset =['InvoiceNo'], inplace = True) data['InvoiceNo'] = data['InvoiceNo'].astype('str')

6 - Splitting the data according to the region of transaction

# Dropping all transactions which were done on credit data = data[~data['InvoiceNo'].str.contains('C')] # Transactions done in France basket_France = (data[data['Country'] =="France"] .groupby(['InvoiceNo', 'Description'])['Quantity'] .sum().unstack().reset_index().fillna(0) .set_index('InvoiceNo')) # Transactions done in the United Kingdom basket_UK = (data[data['Country'] =="United Kingdom"] .groupby(['InvoiceNo', 'Description'])['Quantity'] .sum().unstack().reset_index().fillna(0) .set_index('InvoiceNo')) # Transactions done in Portugal basket_Por = (data[data['Country'] =="Portugal"] .groupby(['InvoiceNo', 'Description'])['Quantity'] .sum().unstack().reset_index().fillna(0) .set_index('InvoiceNo')) basket_Sweden = (data[data['Country'] =="Sweden"] .groupby(['InvoiceNo', 'Description'])['Quantity'] .sum().unstack().reset_index().fillna(0) .set_index('InvoiceNo'))

7 - Hot encoding the Data

# Defining the hot encoding function to make the data suitable # for the concerned libraries def hot_encode(x): if(x<= 0): return 0 if(x>= 1): return 1 # Encoding the datasets basket_encoded = basket_France.applymap(hot_encode) basket_France = basket_encoded basket_encoded = basket_UK.applymap(hot_encode) basket_UK = basket_encoded basket_encoded = basket_Por.applymap(hot_encode) basket_Por = basket_encoded basket_encoded = basket_Sweden.applymap(hot_encode) basket_Sweden = basket_encoded

8 - Building the models and analyzing the results

a) France:

# Building the model frq_items = apriori(basket_France, min_support = 0.05, use_colnames = True) # Collecting the inferred rules in a dataframe rules = association_rules(frq_items, metric ="lift", min_threshold = 1) rules = rules.sort_values(['confidence', 'lift'], ascending =[False, False]) print(rules.head())

                                           antecedents  \
45                        (JUMBO BAG WOODLAND ANIMALS)   
258  (RED TOADSTOOL LED NIGHT LIGHT, PLASTERS IN TI...   
272  (RED TOADSTOOL LED NIGHT LIGHT, PLASTERS IN TI...   
300  (SET/6 RED SPOTTY PAPER CUPS, SET/20 RED RETRO...   
302  (SET/20 RED RETROSPOT PAPER NAPKINS, SET/6 RED...   

                         consequents  antecedent support  consequent support  \
45                         (POSTAGE)            0.076531            0.765306   
258                        (POSTAGE)            0.051020            0.765306   
272                        (POSTAGE)            0.053571            0.765306   
300  (SET/6 RED SPOTTY PAPER PLATES)            0.102041            0.127551   
302    (SET/6 RED SPOTTY PAPER CUPS)            0.102041            0.137755   

      support  confidence      lift  leverage  conviction  
45   0.076531       1.000  1.306667  0.017961         inf  
258  0.051020       1.000  1.306667  0.011974         inf  
272  0.053571       1.000  1.306667  0.012573         inf  
300  0.099490       0.975  7.644000  0.086474   34.897959  
302  0.099490       0.975  7.077778  0.085433   34.489796

From the above output, it can be seen that paper cups and paper and plates are bought together in France. This is because the French have a culture of having a get-together with their friends and family atleast once a week. Also, since the French government has banned the use of plastic in the country, the people have to purchase the paper-based alternatives. 

b) United Kingdom:

#frq_items = apriori(basket_UK, min_support = 0.01, use_colnames = True) #rules = association_rules(frq_items, metric ="lift", min_threshold = 1) #rules = rules.sort_values(['confidence', 'lift'], ascending =[False, False]) #print(rules.head())

If the rules for British transactions are analyzed a little deeper, it is seen that the British people buy different colored tea-plates together. A reason behind this may be because typically the British enjoy tea very much and often collect different colored tea-plates for different occasions.

c) Portugal:

frq_items = apriori(basket_Por, min_support = 0.05, use_colnames = True) rules = association_rules(frq_items, metric ="lift", min_threshold = 1) rules = rules.sort_values(['confidence', 'lift'], ascending =[False, False]) print(rules.head())

                             antecedents                          consequents  \
1170  (SET 12 COLOUR PENCILS DOLLY GIRL)     (SET 12 COLOUR PENCILS SPACEBOY)   
1171    (SET 12 COLOUR PENCILS SPACEBOY)   (SET 12 COLOUR PENCILS DOLLY GIRL)   
1172  (SET OF 4 KNICK KNACK TINS LONDON)   (SET 12 COLOUR PENCILS DOLLY GIRL)   
1173  (SET 12 COLOUR PENCILS DOLLY GIRL)   (SET OF 4 KNICK KNACK TINS LONDON)   
1174  (SET 12 COLOUR PENCILS DOLLY GIRL)  (SET OF 4 KNICK KNACK TINS POPPIES)   

      antecedent support  consequent support   support  confidence       lift  \
1170            0.051724            0.051724  0.051724         1.0  19.333333   
1171            0.051724            0.051724  0.051724         1.0  19.333333   
1172            0.051724            0.051724  0.051724         1.0  19.333333   
1173            0.051724            0.051724  0.051724         1.0  19.333333   
1174            0.051724            0.051724  0.051724         1.0  19.333333   

      leverage  conviction  
1170  0.049049         inf  
1171  0.049049         inf  
1172  0.049049         inf  
1173  0.049049         inf  
1174  0.049049         inf

On analyzing the association rules for Portuguese transactions, it is observed that Tiffin sets (Knick Knack Tins) and color pencils. These two products typically belong to a primary school going kid. These two products are required by children in school to carry their lunch and for creative work respectively and hence are logically make sense to be paired together.

d) Sweden:

frq_items = apriori(basket_Sweden, min_support = 0.05, use_colnames = True) rules = association_rules(frq_items, metric ="lift", min_threshold = 1) rules = rules.sort_values(['confidence', 'lift'], ascending =[False, False]) print(rules.head())

                           antecedents                     consequents  \
0        (12 PENCILS SMALL TUBE SKULL)   (PACK OF 72 SKULL CAKE CASES)   
1        (PACK OF 72 SKULL CAKE CASES)   (12 PENCILS SMALL TUBE SKULL)   
4              (36 DOILIES DOLLY GIRL)  (ASSORTED BOTTLE TOP  MAGNETS)   
5       (ASSORTED BOTTLE TOP  MAGNETS)         (36 DOILIES DOLLY GIRL)   
180  (CHILDRENS CUTLERY CIRCUS PARADE)  (CHILDRENS CUTLERY DOLLY GIRL)   

     antecedent support  consequent support   support  confidence  lift  \
0              0.055556            0.055556  0.055556         1.0  18.0   
1              0.055556            0.055556  0.055556         1.0  18.0   
4              0.055556            0.055556  0.055556         1.0  18.0   
5              0.055556            0.055556  0.055556         1.0  18.0   
180            0.055556            0.055556  0.055556         1.0  18.0   

     leverage  conviction  
0    0.052469         inf  
1    0.052469         inf  
4    0.052469         inf  
5    0.052469         inf  
180  0.052469         inf

On analyzing the above rules, it is found that boys’ and girls’ cutlery are paired together. This makes practical sense because when a parent goes shopping for cutlery for his/her children, he/she would want the product to be a little customized according to the kid’s wishes.

.css-ytumd6{-webkit-text-decoration:none;text-decoration:none;}Apriori Algorithm

Association Rules

1 - Importing the required libraries

2 - Loading and exploring the data

3 - Exploring the columns of the data

4 - Exploring the different regions of transactions

5 - Cleaning the Data

6 - Splitting the data according to the region of transaction

7 - Hot encoding the Data

8 - Building the models and analyzing the results

a) France:

b) United Kingdom:

c) Portugal:

d) Sweden:

Apriori Algorithm