Understanding customer behavior is critical for businesses trying to stay competitive. Market basket analysis is a powerful technique for gaining insights into customer preferences. This data mining method allows us to discover which products are frequently purchased together, providing valuable information for cross-selling, marketing strategies and inventory management.
In this tutorial, we will start to enter the world of market basket analysis using Python. The dataset we chose is the Online Retail II dataset, derived from the UCI Machine Learning Repository. This real-world dataset captures two years of transactions from 1 December 2009 to 9 December 2011 for a UK online retail business. The dataset is rich in features, including multivariate, sequential, and time series. Properties, which contain both textual and numeric features.
Dataset overview:
The following is a snapshot of the key attributes and characteristics of the Online Retail II dataset:
Example: 1,067,371
Features: Multiple, including integers and real values.
Dataset information:
The Online Retail II dataset encapsulates all transactions for a UK-based registered non-store online retail company that specializes in unique all-occasion gift items. Although this dataset contains diverse customer segments, it primarily caters to wholesalers.
Variable Overview:
To start our market basket analysis journey, let’s understand the core variables in the dataset:
InvoiceNo: This is the invoice number, nominal attribute. It is a 6-digit integer that uniquely identifies each transaction. If the invoice number begins with the letter "c", it indicates a cancellation.
StockCode: The code of the product (commodity), also nominal. Each different product is assigned a unique 5-digit integer.
Description: The name of the product (item), another nominal attribute.
Quantity: Essentially a number, this attribute represents the quantity of each product in each transaction