Funny item co-occurrences in 3.2M Instacart orders
Summary
This article analyzes funny item co-occurrences in 3.2 million Instacart orders using a data-science workflow. It covers initial pair/triple/quad analyses, the problem of noise with 50,000 products, and the move to lift-based ranking to identify surprising combinations. It then introduces GS1 Global Product Classification (GPC) to reduce dimensionality, experiments with a humor index to boost ranking, and showcases humorous pairings and small-cart examples.