Finding Association Rules for Market Basket Analysis

Finding Association Rules for Market Basket Analysis

The aim of a market basket analysis is to analyze whether there are combinations of products that frequently co-occur in transactions. This workflow illustrates how to build a recommendation engine for market basket analysis.

Task

Build a recommendation engine for market basket analysis using the Borgelt version of the Apriori algorithm.

Data Reading

2 files:

• Transactions

• Products info

Transactions are in the shape of concatenated Strings. The Association Rule learner node requires a collection of items as input. The Cell Splitter node here transforms string rows into item collections.

Another way of getting item collections is to use the aggregation method "List" in a GroupBy node.

Both files are located in TheData/Baskets

Build Association Rules

The Association Rule Learner (Borgelt) node implements the "A priori" algorithm to find the association rules.

• min. set size = 1

• min rule confidence = 10%

• min support (in %) is controlled by the Double Configuration node

Data Processing

• Create RuleID from RowID

• Ungroup antecedent item collection

• Join prices and product names to antecedent items

• Join consequent item to product name

• Add price to consequent item

• Calculate price stats and rule revenue

Save the Model

The model here is a table of rules.

We write the association rules to a file.