Module 2: Association Rule Mining

Code Along

Code Along

Association rule mining helps discover interesting relationships between items in large datasets. A common example is market basket analysis, where we find which products are frequently bought together.

Let’s implement association rule mining using Python and the Apriori algorithm from the mlxtend library:

Import the libraries

# Install necessary libraries (run this if needed)
# pip install mlxtend pandas numpy

# Import the libraries
import pandas as pd
import numpy as np
from mlxtend.frequent_patterns import apriori
from mlxtend.frequent_patterns import association_rules

Step 1: Create a sample dataset

# Create a sample transaction dataset
data = {
    'TransactionID': [1, 1, 1, 2, 2, 3, 3, 3, 3, 4, 4, 5, 5, 5, 5],
    'Item': ['Bread', 'Milk', 'Eggs', 'Bread', 'Milk', 'Bread', 'Milk', 'Eggs', 'Beer', 'Milk', 'Eggs', 'Bread', 'Milk', 'Eggs', 'Beer']
}

# Convert to DataFrame
df = pd.DataFrame(data)
print(df.head(10))

Step 2: One-hot encoded format

# Transform the data into a one-hot encoded format
basket = pd.crosstab(df['TransactionID'], df['Item'])

# Convert to binary values (1 for presence, 0 for absence)
basket_sets = basket.applymap(lambda x: 1 if x > 0 else 0)
print(basket_sets)

Step 3: Apply the Apriori algorithm

# Find frequent itemsets using the Apriori algorithm
# min_support is a threshold for how frequently an itemset appears in the dataset
frequent_itemsets = apriori(basket_sets, min_support=0.4, use_colnames=True)
print(frequent_itemsets)

Step 4: Generate association rules

# Generate association rules
# min_threshold is a threshold for how strong the rules are
rules = association_rules(frequent_itemsets, metric="lift", min_threshold=1.0)
print(rules)

Step 5: Interpret the results

# Sort rules by confidence (descending)
rules = rules.sort_values(['confidence', 'lift'], ascending=[False, False])
print(rules[['antecedents', 'consequents', 'support', 'confidence', 'lift']])

Understanding the metrics:

• Support: Frequency of itemsets in the dataset

• Confidence: How often the rule has been found to be true

• Lift: How likely item Y is purchased when item X is purchased, compared to how likely item Y is purchased in general